Expert Review of the Vieques Heart Study
Summary Report for the
Vieques Heart Study Expert Panel Review

(Resumen Ejecutivo en Español)
A MEETING ON JULY 12-13, 2001

SPONSORED BY
THE AGENCY FOR TOXIC SUBSTANCES & DISEASE REGISTRY
AND
THE PONCE SCHOOL OF MEDICINE

Draft prepared for:
The Agency for Toxic Substances and Disease Registry
Atlanta, Georgia

Contract No. 200-2000-10039
Task Order No. 9
by:
Eastern Research Group
110 Hartwell Avenue
Lexington, MA 02421-3136

August 17, 2001

This report was prepared by Eastern Research Group, Inc. (ERG), a contractor to the Agency for Toxic Substances and Disease Registry (ATSDR), as a general record of discussion for the expert panel review meeting on the Vieques Heart Study. The meeting was co-sponsored by ATSDR and the Ponce School of Medicine.

This report captures the main points of scheduled presentations and highlights discussions among the expert panelists and other participants. This report does not contain a verbatim transcript of all issues discussed during the meeting. Additionally, the report does not embellish, interpret, or enlarge upon matters that were incomplete or unclear. All panelists and participants received a draft copy of the report to verify that the contents of the report accurately reflect the content and tone of discussions at the meeting. Except as specifically noted, no statements in this report represent analyses or positions of ATSDR, CDC, or of ERG..

TABLE OF CONTENTS

List of Abbreviations
List of Tables

EXECUTIVE SUMMARY (English)
RESUMEN EJECUTIVO (Español)

1.0 INTRODUCTION

1.1 Background
1.2 The Expert Panel
1.3 The Expert Panel Review Meeting
1.4 Report Organization

2.0 TECHNICAL PRESENTATIONS ON THE VIEQUES HEART STUDY

2.1 Summary of Presentation by Dr. Oh, Mayo Clinic
2.2 Summary of Presentation by Dr. Ríos, Ponce School of Medicine

3.0 COMMENTS ON STUDY DESIGN AND DATA ASCERTAINMENT

4.0 COMMENTS ON ECHOCARDIOGRAPHIC MEASUREMENTS

4.1 Comments on How Images Were Collected
4.2 Comments on How Images Were Read
4.3 Comments on Limitations of Echocardiography (Resolution)
4.4 Comments on Differences Between the PSM and Mayo Clinic Data
4.5 Comments on Clinical Significance

5.0 COMMENTS ON STATISTICAL ANALYSIS

6.0 COMMENTS ON INTERPRETATION AND INFERENCE

7.0 REFERENCES

APPENDICES

Appendix A: List of Expert Panelists and Participants
Appendix B: Biographies of Expert Panelists and Participants
Appendix C: Charge to the Panelists
Appendix D: Meeting Agenda
Appendix E: Summary Statements Submitted by the Expert Panelists

LIST OF ABBREVIATIONS

ATSDR - Agency for Toxic Substances and Disease Registry
CDC - Centers for Disease Control and Prevention
CT - computerized tomography
EKG - electrocardiogram
EPA U.S. - Environmental Protection Agency
MR - magnetic resonance
PSM - Ponce School of Medicine
TGC - time gain compensation

LIST OF TABLES

Table 2-1 Average Pericardial Thicknesses Based on Echocardiographic Readings Conducted by PSM and by Mayo Clinic

Table 3-1 Data on Response Rates and Collection of Echocardiographic Images

Table 4-1 Paired Comparisons of Pericardial Thickness Measurements by PSM Investigators and by Mayo Clinic Investigators

Table 4-2 Differences Between the PSM Investigators' Readings and the Mayo Clinic Investigators' Readings (Paired Comparisons)

EXECUTIVE SUMMARY

BACKGROUND & INTRODUCTION

This report describes the meeting and include

s a synthesis of comments and written reports that were individually submitted by each of the external panel members. It does not represent consensus advice to ATSDR. ATSDR will take into account the individual views of the review participants and reach its own recommendations as requested by the White House.

In January 2001, a pilot study comparing the echocardiograms of residents of Vieques and Ponce, Puerto Rico reported substantial valvular abnormalities and pericardial thickening in a large proportion of Vieques residents - findings not seen among Ponce residents. The possible abnormalities noted in the Vieques residents were attributed to "vibro-acoustic disease" (VAD), which had been described in the medical literature by Portuguese investigators. VAD was said to be occurring as the result of noise and vibrations caused by naval exercises on the Island of Vieques. The White House asked the Department of Health and Human Services to investigate the issues raised by the study. The Department, in turn, referred this request to the Agency for Toxic Substances and Disease Registry (ATSDR), which was already investigating environmental public health issues in Vieques. ATSDR received considerable assistance in this work from the Cardiovascular Diseases Branch of the Centers for Disease Control & Prevention (CDC).

Concurrent with this request, the Ponce School of Medicine (PSM), led by President and Dean Dr. Manuel Martínez Maldonado, had begun a more definitive study of possible cardiac abnormalities among Vieques residents. This study sought to overcome methodological problems (e.g., sampling frame, lack of blinding) in the earlier pilot study. On March 29-30, 2001, scientists from ATSDR and CDC met with the PSM investigators and agreed to invite the assistance of recognized practitioners and scientists in reviewing and interpreting the findings. Reviewers were chosen by consensus. They were experts with international reputations in echocardiography and environmental or cardiovascular epidemiology. San Juan was chosen as the location of the meeting. Because of its extensive experience, the echocardiography "core" laboratory at Mayo Clinic, directed by Dr. Jae K. Oh, was selected to review the echocardiograms.

MEETING ARRANGEMENTS & REPORT ORGANIZATION

Eight accomplished physician-scientists were chosen as reviewers and accepted the invitation to participate. They are referred to herein as "panelists" (see appendices A & B) and are principally from academic institutions. Four panelists are from U.S. universities, two are from Mexico, and two are from Spain. Half of the panelists are specialists in cardiology and echocardiography; the rest are epidemiologists. Other meeting participants included personnel or consultants of PSM and ATSDR, Dr. Jae Oh of Mayo Clinic, and Dr. John Rullán, Secretary of Health of Puerto Rico.

The meeting took place during July 12-13, 2001, in a conference room in the Condado Plaza Hotel in San Juan. The co-chairs of the meeting were Dr. Martínez Maldonado and Dr. David Fleming, Deputy Administrator of ATSDR. The purpose of the meeting was to review the methods, results, and public health significance of the Vieques Heart Study, considering both the PSM and Mayo Clinic Data.

Before the meeting, panelists and other participants were provided with background materials including a specific charge (Appendix C). The meeting followed a prearranged agenda (Appendix D). Panelists provided verbal remarks and written comments on the study. They made individual recommendations on how the Vieques Heart Study data should be interpreted. Although there was broad agreement on many points, no effort was made to generate a consensus judgment. The panelists' individual written comments have been edited for style, translated (where necessary), and checked for accuracy by the persons who wrote them (Appendix E). A contractor recorded the meeting and summarized and organized the points made by the participants in the various sessions. The meeting minutes thus developed constitute the bulk of this report.

INTRODUCTORY PRESENTATIONS

Dr. Martínez Maldonado noted that Vieques is east-southeast of Puerto Rico, spans roughly 25 square miles, and has roughly 9,000 residents. For approximately the last 60 years, the eastern and western ends of the island have been Navy property. During this time, the Navy has conducted bombing exercises and other war games on the easternmost part of the island.

Dr. Rullán presented background information regarding health issues on Vieques. Changes in the health care system over time have limited the health care available on Vieques. Residents frequently travel to the main island for care. Vieques does less well than the main island on a number of indicators related to public health, including unemployment rates, teenage pregnancy rates, proportion of population receiving prenatal care during the first trimester of pregnancy, per capita income, mortality rates for specific diseases, and self-reported morbidity rates for various types of diseases. Between 1950 and 1991 the Puerto Rico cancer registry was an excellent resource. Following the health care reforms in 1991 and associated cutbacks in certain public health services, health officials have not been able to produce annual reports from the existing cancer registry data. However, an effort to update the cancer registry will soon be complete.

VIEQUES HEART STUDY METHODS & FINDINGS

Dr. Martínez Maldonado and Dr. Carlos Ríos presented the Vieques Heart Study. The study objective was to determine whether an association existed between place of residence (Vieques or Ponce Playa) and morphological cardiovascular changes among commercial fishermen. Investigators sampled randomly from the lists of licensed commercial fishermen from Vieques and from Ponce Playa to obtain 53 and 42 subjects from the two areas, respectively. Investigators measured height, weight, blood pressure and other physical parameters, collected questionnaire data on demographics and possible confounders, and recorded echocardiographic images of subjects. The echocardiograms were read "blindly" (i.e., without knowledge of the site of residence of the particular subject) for pericardial thickness by a group of several experienced PSM cardiologists, with caliper placement done by consensus and by using magnified images.

As noted above, the echocardiograms were re-read by Dr. Oh's group at the Mayo Clinic, who were also blind to the identities of study subjects. Dr. Oh and each of two experienced research sonographers read all studies for pericardial thickness. As with measurements of other parameters, the randomly-assigned, primary sonographer's reading of pericardial thickness was considered final, except in a very few cases in which a substantial discrepancy occurred between the readers. In these cases, Dr. Oh's readings were considered final.

For functional and structural measurements other than pericardial thickness, interobserver agreement among Mayo readers was strong (R-squares of 0.6 to 0.93). However, the interobserver variation on measures of pericardial thickness was weak (R-square only 0.22 for non-magnified images and similarly poor for magnified images). Although there was little intraobserver variation in most parameters measured, the R-square for pericardial thickness (non-magnified) was only 0.3.

For the anatomical and functional parameters measured by both groups, the Ponce and Mayo findings were virtually identical. Moreover, neither data set indicated cardiac pathology among either group of fishermen. Ponce and Mayo findings regarding pericardial thickness were also similar. In neither data set did any subject have an abnormally thick pericardium, based on an upper limit of normal of 2 mm.

By PSM's measurements, the average pericardial thickness was slightly greater among Vieques fishermen than among Ponce Playa fishermen (1.20 mm vs. 1.05 mm), and this difference was statistically significant (P = 0.03). The values for pericardial thickness measured by Mayo were within the same range as those measured by PSM, but did not achieve statistical significance when Vieques and Ponce fishermen were compared (0.78 mm vs. 0.82 mm, respectively).

PANEL CONCLUSIONS

The principal conclusion of the panel was that neither the Ponce nor the Mayo readings contained information indicating a cardiac health problem in the fishermen from either location. The initial report of gross valvular pathology from the pilot study was not replicated. All reviewers agreed that there was no clinically relevant difference between Vieques and Ponce Playa subjects in pericardial thickness as had been reported in the pilot study. Moreover, neither the PSM nor the Mayo measurements showed any subject's pericardial thickness to be larger than 2 mm - a reasonable value for the upper limit of normal, based on the published literature.

The PSM study got generally high marks from the panelists regarding study design and statistical analysis. The sampling frame (lists of registered fishermen) was regarded as appropriate, and reviewers generally felt that the response rate was adequate. The fact that reasonably clear-cut hypotheses had been developed beforehand largely obviated concerns about the problem of multiple comparisons. In general, panelists felt that the statistical tests used were appropriately chosen and employed. Panelists noted that echocardiographic readings were performed with appropriate blinding, including masking of dates, at both PSM and Mayo. More detailed comments and suggestions are described by individual panelists in Appendix E.

With regard to pericardial measurements, the panelists noted that there was substantial measurement error in the technique (i.e., low sensitivity of the echocardiography machine) for measurement of pericardial thickness within the normal range. The lower limit of resolution of trans-thoracic echocardiography was given by the panelists as 1 mm. This value is substantially larger than the average between-group difference of 0.15 mm found by PSM. Moreover, the interobserver and intraobserver reproducibility of the measurements at Mayo was low. Thus, it was not surprising that PSM group reported little correlation between their individual measurements of pericardial thickness and the Mayo measurements (R-square 0.04). With these facts in mind, the panelists opined that the difference between groups in pericardial thickness observed by PSM most likely represented measurement error inherent in the technique used. In any event, no panelist attributed clinical significance to a difference in thickness this small and within the range of values reported by PSM and Mayo.

Pericardial thickness measurements were slightly smaller in the Mayo than in the PSM readings. Nevertheless, it is impossible to state with certainty that either one is "correct," and it is similarly impossible to rule out that the small numerical differences measured by PSM between Ponce and Vieques fishermen exist. No significant cardiac function changes were detected in either of the populations studied (Vieques vs. Ponce). Thus, even if a small difference in pericardial thickness exists, Vieques fishermen do not appear to have the hemodynamic consequences of a thickened pericardium.

Reviewers noted that the Vieques Heart Study focused specifically on the heart, and it therefore could not rule out other health effects of the Naval exercises, including other potential effects of noise and vibration. Also, a study of commercial fishermen, a single occupational group, cannot be assumed to be representative of an entire population. In the words of one panelist, "further studies and ongoing monitoring are needed of … morbidity, mortality, and risk profiles of Vieques citizens as current plans are implemented to reduce their noise exposure and to improve the economy, health care, and public health of the area." That said, however, reviewers generally did not feel that further studies of pericardial thickness would be of value.

It was recommended that the Ponce and Mayo Clinic measurements be published jointly. This would address as thoroughly as possible the issues of appropriate methodology, assessment of pericardial thickness, and the clinical significance of the results.

SUMMARY CONCLUSION

The well-executed PSM study does not support the existence of cardiac pathology among Vieques fishermen. Because of the inability of trans-thoracic echocardiography to measure reliably the small differences found, the differences reported are likely due to measurement error (intrinsic to the technique, not the scientists who used it). This fact almost certainly accounts for the different results obtained when Mayo readings of pericardial thickness are used in place of PSM readings. The Vieques Heart Study represents a valuable contribution to scientific knowledge regarding the use of echocardiography and should be published in the peer-reviewed scientific literature..

RESUMEN EJECUTIVO

ANTECEDENTES E INTRODUCCIÓN

Este informe describe la reunión e incluye una síntesis de los comentarios e informes escritos sometidos por cada uno de los miembros del panel externo. Se debe aclarar que el informe no representa un consejo en consenso a la Agencia de Sustancia Tóxicas y Registro de Enfermedades (ATSDR por sus siglas en inglés). La ATSDR tomará en cuenta las opiniones de los participantes y hará sus propias recomendaciones de acuerdo a lo solicitado por la Casa Blanca.

En enero del 2001, un estudio piloto de comparación de ecocardiogramas de residentes de Vieques y Ponce, Puerto Rico, reportó anomalías valvulares y engrosamiento pericárdico sustancial en una gran proporción en los residentes de Vieques - no así con los residentes de Ponce. Estas posibles anomalías en los residentes de Vieques se atribuyeron a "enfermedad vibro acústica" (VAD por sus siglas en inglés), que ha sido descrita en la literatura médica por investigadores portugueses. Se sugirió que la VAD era el resultado del ruido y las vibraciones causadas por los ejercicios navales en la isla de Vieques. La Casa Blanca pidió al Departamento de Salud y Servicios Humanos que investigara la situación presentada en el estudio. El Departamento, a su vez, refirió la solicitud a la Agencia de Sustancias Tóxicas y Registro de Enfermedades (ATSDR), la cual se encontraba ya investigando situaciones de salud pública ambientales en Vieques. La ATSDR recibió asistencia considerable en este trabajo del Grupo de Enfermedades Cardiovasculares del Centro de Control y Prevención de Enfermedades (CDC por sus siglas en inglés).

Coincidentemente con esta solicitud, la Escuela de Medicina de Ponce (PSM), dirigida por su presidente y decano Dr. Manuel Martínez Maldonado, había comenzado un estudio más definitivo de las causas posibles de las anormalidades cardíacas entre los residentes de Vieques. Este estudio tenía la intención de reducir los problemas metodológicos (por ejemplo, enmarcado de las muestras, carencia de "ensayos a ciegas" ("blinding"), etc.) del estudio piloto anterior. El 29-30 de marzo del 2001, científicos de la ATSDR y del CDC se reunieron con los investigadores de la PSM y acordaron de pedir la colaboración de médicos practicantes y científicos reconocidos en la revisión y en la interpretación de los resultados de las investigaciones. Los revisores se escogieron por consenso. Estos revisores eran expertos de reputación internacional en ecocardiografía y epidemiología ambiental o cardiovascular. San Juan fue escogido como la sede de la reunión. Debido a su gran experiencia, el laboratorio ecocardiográfíco medular ("core") de la Clínica Mayo dirigido por el Dr. Jae K. Oh, fue seleccionado para revisar los ecocardiogramas.

ARREGLOS PARA LAS REUNIONES Y ORGANIZACIÓN

Ocho médicos-científicos experimentados fueron escogidos como revisores y aceptaron la invitación para participar. A partir de ahora nos referiremos a ellos como los "panelistas" (ver Apéndices A & B). En su mayoría, estos médicos pertenecen a instituciones académicas. Cuatro panelistas son de universidades de Estados Unidos, dos son de México, y dos son de España. La mitad de los panelistas son especialistas en cardiología y ecocardiografía, y el resto son epidemiólogos. Entre los otros participantes se encuentran incluidos parte del personal de la PSM
y la ATSDR, además del Dr. Jae Oh de la Clínica Mayo, y el Dr. John Rullán, Secretario de Salud de Puerto Rico.

La reunión se efectuó durante los días 12 y 13 de julio del 2001 en una sala de conferencias del Condado Plaza Hotel en San Juan. Los presidentes de la reunión fueron los doctores Martínez Maldonado y David Fleming, Administrador Suplente de la ATSDR. El propósito de la reunión fue la revisión de los métodos, resultados y el significado de salubridad pública en el Estudio Cardíaco de Vieques, considerando tanto los datos de PSM, como los de la Clínica Mayo.

Antes de la reunión, se proporcionó a los panelistas y a los otros participantes material de referencia, incluyendo instrucciones específicas (Apéndice C). La reunión estuvo sujeta a una agenda preparada con antelación (Apéndice D). Los panelistas proporcionaron observaciones verbales y comentarios escritos sobre el estudio. Los panelistas hicieron recomendaciones individuales acerca de la forma en la que los datos del Estudio del Corazón de Vieques deberían ser interpretados. Aunque hubo amplio acuerdo en una gran cantidad de tópicos, no se hizo esfuerzo por alcanzar un consenso. El estilo, la traducción (donde fue necesario) y la fidelidad de los comentarios individuales escritos de los panelistas fueron editados y comprobados por las personas que los escribieron. (Apéndice E). Un contratista grabó la reunión y resumió y organizó los planteamientos hechos por los participantes en las diferentes sesiones. Las actas de la reunión forman el grueso de este informe.

INTRODUCCIÓN Y PRESENTACIONES

El Dr. Martínez Maldonado señaló que Vieques se encuentra al este - sudeste de la isla de Puerto Rico, que abarca alrededor de 25 millas cuadradas y tiene aproximadamente 9,000 residentes. Durante aproximadamente los últimos 60 años, los extremos este y oeste de la isla han sido propiedad de la Marina. Durante la totalidad de este lapso, la Marina ha conducido ejercicios de bombardeos y otras tácticas de guerra en la parte más oriental de la isla.

El Dr. Rullán presentó el trasfondo de los problemas de Vieques. Cambios en el sistema de salud asistencial gubernamental en los últimos años han limitado los servicios de salud disponible en Vieques. Los residentes frecuentemente viajan hasta la isla principal para obtener atención médica. Vieques tiene una puntuación menor que la isla principal en varios de los indicadores relacionados con la salud pública, incluyendo tasas de desempleo, adolescentes embarazados, la proporción de la población que recibe cuidado prenatal durante el primer trimestre del embarazo, ingresos per cápita, tasas de mortalidad en relación con enfermedades específicas y tasas de morbilidad de enfermedades auto-reportables. Entre 1950 y 1991 el registro de cáncer de Puerto Rico fue una excelente fuente de información. Después de las reformas de salud de 1991 y los cortes asociados con ciertos servicios de salud pública, los oficiales de salud no han sido capaces de producir informes anuales de los datos de cáncer existentes. Sin embargo, se espera que pronto se pueda actualizar el registro de cáncer.

RESULTADOS Y MÉTODOS DEL ESTUDIO DEL CORAZÓN DE VIEQUES

Los doctores Martínez Maldonado y Carlos Ríos presentaron el Estudio del Corazón de Vieques. El objetivo del estudio fue determinar si existía alguna asociación entre el lugar de residencia (Vieques o Playa Ponce) y los cambios cardiovasculares morfológicos entre los pescadores comerciales. Los investigadores tomaron muestras al azar de la lista de pescadores comerciales licenciados de Vieques y Playa Ponce para obtener 53 y 42 muestras respectivamente de las dos áreas. Los investigadores midieron el peso, la presión sanguínea y otros parámetros físicos, reunieron datos de cuestionarios demográficos y grabaron ecocardiogramas de los sujetos de estudio. Los ecocardiogramas fueron leídos a ciegas (sin saber su procedencia) por un grupo de varios cardiólogos experimentados de la PSM en búsqueda de engrosamiento pericárdico, con la colocación de los calibradores llevada a cabo por consenso y usando imágenes aumentadas.

Como describimos anteriormente, los ecocardiogramas fueron leídos por el grupo del doctor Oh en la Clínica Mayo. Sin saber cuál era el lugar de procedencia de los sujetos, el Dr. Oh y cada uno de los otros dos experimentados técnicos sonográficos leyeron todos los estudios en búsqueda de engrosamiento pericárdico. Tal como se hizo con las medidas de los otros parámetros, la lectura inicial de los sonógrafos, asignadas al azar, en busca de engrosamiento pericárdico fue considerada final, excepto en limitadas ocasiones en las cuales ocurrió discrepancia substancial entre las lecturas. En tales casos, la lectura del Dr. Oh fue la considerada como final y válida.

En contraste con las medidas del engrosamiento pericárdico, las medidas funcionales y estructurales de la Mayo tuvieron una fuerte correlación interobservador (R-cuadrados de 0.6 a 0.93). Sin embargo, la variación interobservadora en medidas de engrosamiento pericárdico fue débil (R-cuadrado solamente 0.22 para imágenes no aumentadas y similarmente pobres para imágenes aumentadas). Aunque hubo poca variación intraobservadora en la mayoría de los parámetros medidos, el R-cuadrado para engrosamiento pericárdico (no-aumentado) fue solamente 0.3.

Para los parámetros anatómicos y funcionales medidos por ambos grupos, los resultados de Ponce y Mayo fueron virtualmente idénticos. Por otra parte, ninguno de los datos indicó patología cardiaca entre ninguno de los grupos de pescadores. Los resultados de Ponce y Mayo también fueron similares en relación con las lecturas del engrosamiento pericárdico. En ninguno de los datos se notó la existencia de engrosamiento anormal del pericardio, basado en un límite superior normal de 2 mm.

De acuerdo con las medidas de PSM, el promedio del grosor fue ligeramente mayor entre los pescadores de Vieques que entre la población de pescadores de Playa Ponce (1.20 mm vs. 1.05 mm), y esta diferencia fue estadísticamente significativa (P = 0.03). Los valores para el engrosamiento pericárdico medido por Mayo estuvieron dentro del mismo rango que los de PSM, pero no alcanzaron significado estadístico cuando los pescadores de Vieques y Ponce fueron comparados (0.78 mm vs. 0.82 mm, respectivamente).

CONCLUSIONES DEL PANEL

La conclusión principal del panel fue que los resultados de Ponce o de Mayo no contienen información que indique que existe al momento un problema de salud cardíaca en los pescadores de ninguna de las localizaciones estudiadas. El informe inicial de la patología gruesa valvular informada en el estudio piloto no fue corroborado. Todos los revisores estuvieron de acuerdo en que no existía ninguna diferencia de relevancia clínica entre los sujetos de Vieques y la Playa de Ponce en lo referente al engrosamiento pericárdico, tal y como fue reportado en el informe piloto. Además, ni las medidas de la PSM ni las medidas de Mayo mostraron ningún engrosamiento pericárdico en los sujetos estudiados mayor de 2 mm - un valor razonable para el límite normal, basado en la literatura publicada.

El estudio de la PSM recibió encomios substanciales de parte de los panelistas en relación con el análisis estadístico y el diseño del estudio. El marco de las muestras (lista de pescadores registrados) fue considerado como apropiado, y los revisores acordaron que la tasa de respuesta fue adecuada. El hecho de que una hipótesis netamente definida había sido desarrollada de antemano, en gran manera obviaba las preocupaciones acerca del problema de comparaciones múltiples. En general, los panelistas acordaron en que las pruebas estadísticas usadas fueron escogidas y usadas apropiadamente. Los panelistas aseveraron que las lecturas de ecocardiografía fueron llevadas a cabo con los parámetros apropiados, incluyendo la no-revelación de las fechas ni del lugar de procedencia del ecocardiograma, tanto al grupo de PSM como al de la Clínica Mayo. Mayores detalles y sugerencias se encuentran descritos por parte de los panelistas individuales en el Apéndice E.

En lo relacionado a las medidas pericárdicas, los panelistas aseveraron que había errores de medidas substanciales en las técnicas (la sensibilidad de la máquina) de medida del espesamiento pericardial dentro del índice normal. El límite menor de resolución de la ecocardiografía trans-toráxica fue proporcionado por los panelistas como 1 mm. Este valor es substancialmente mayor que el promedio entre la diferencia de grupos de 0.15 mm encontrado por la PSM. Por otra parte, la reproducibilidad del interobservador y del intraobservador de las medidas de Mayo fue baja. Por tanto, no fue sorprendente que el grupo PSM reportó una correlación pequeña entre sus medidas de espesamiento pericardial y las medidas de Mayo (R-cuadrado 0.04). Tomando esto en consideración, la opinión de los panelistas fue que la diferencia entre los grupos en lo referente al engrosamiento pericárdico observada por PSM posiblemente representaba errores de medida inherentes a la técnica usada. En cualquier caso, ningún panelista atribuyó significado clínico a la diferencia en un engrosamiento tan pequeño y dentro del índice de valores reportados por la PSM y Mayo.

Las lecturas en las medidas de engrosamiento pericárdico fueron ligeramente menores para Mayo que para la PSM. Sin embargo, es posible declarar con certidumbre que ninguna de las dos es incorrecta, y que es similarmente imposible descartar que la pequeña diferencia numérica entre las medidas del PSM entre Ponce y los pescadores de Vieques existe. No se detectó cambio alguno en las funciones cardiacas de las poblaciones estudiadas (Vieques vs. Ponce). Por lo tanto, aunque existiese la pequeña diferencia en el engrosamiento pericárdico, los pescadores de Vieques no aparentan tener las consecuencias hemodinámicas de un pericardio grueso.

Los revisores comentaron que el Estudio del Corazón de Vieques se concentró específicamente en el corazón, y que por lo tanto no se podía descartar que hubiese efectos de los ejercicios navales en otras áreas de la salud, incluyendo efectos potenciales debido al ruido y a las vibraciones. Además se declaró que no se puede asumir que un estudio de solamente los pescadores, un grupo ocupacional único, es representativo la totalidad de la población general de Vieques. En las palabras de uno de los panelistas: "...más estudios y monitoreo continuo son necesarios, además de estudios de... la morbilidad, y los perfiles de riesgo de los habitantes de Vieques, al tiempo que se implementan planes para reducir la exposición a ruidos y para mejorar la economía, el cuidado de la salud y la salud pública del área." Habiendo dicho esto, sin embargo, los revisores en general no consideraron que fueran necesarios más estudios para determinar el engrosamiento pericárdico.

Se recomendó que las medidas de Ponce y la Clínica Mayo fueran publicadas en conjunto para aclarar para todos los ecocardiografistas cual es la metodología apropiada para la evaluación del engrosamiento pericárdico, y el significado clínico de los resultados que usan esta tecnología.

CONCLUSIÓN DEL RESUMEN

El estudio bien ejecutado de la PSM no apoya la existencia de patologías cardíacas entre los pescadores de Vieques. Debido a la incapacidad del ecocardiograma trans-toráxico para medir confiablemente las pequeñas diferencias en los resultados encontrados, las diferencias reportadas son probablemente debidas a errores del método de medida utilizado (intrínseco a la técnica, no a los científicos que la usaron). Este hecho casi ciertamente explica la diferencia entre las lecturas de la PSM y la Mayo. El Estudio del Corazón de Vieques representa una contribución valiosa al conocimiento científico en lo relacionado con el uso de la ecocardiografía y debería ser publicado en una revista científica de revisión por pares ("peer review journal").

1.0 INTRODUCTION

This report is a record of the discussion during an expert panel review meeting held in San Juan, Puerto Rico, on July 12 - 13, 2001. The meeting was co-sponsored by the Agency for Toxic Substances and Disease Registry (ATSDR) and the Ponce School of Medicine (PSM). The Cardiovascular Diseases Branch of the Centers for Disease Control and Prevention (CDC) also contributed substantially to this effort.

The meeting was held to review possible cardiac abnormalities among commercial fishermen who reside on the island of Vieques. This section of the report provides background information on the expert panel review meeting. Later sections document the technical discussions that took place among the reviewers and other participants.

This introduction, Section 1.1, provides background information on Vieques and the research under review. Section 1.2 describes the process by which the expert panelists were selected. Section 1.3 reviews the meeting agenda, and Section 1.4 outlines the organization of the rest of this summary report.

1.1 Background

The island of Vieques lies east of the main island of Puerto Rico and has approximately 9,000 residents. For several decades, the United States Navy (Navy) has used the far eastern end of Vieques as a practice ground for training exercises. The exercises have involved the use of explosive ordnance (bombs, artillery shells, and other explosive devices), which cause noise and vibration evident to Vieques residents.

Concern has been expressed regarding possible adverse impact of Navy exercises on the health of Vieques residents. This concern has been particularly great during the last 2 years. One of the issues raised has been whether noise and vibration could have had an adverse health impact on island residents. Dr. N. Castelo-Branco and others in Portugal have described a syndrome of cardiologic, neurologic, and immunologic findings in aircraft workers that they labeled "vibroacoustic disease." Cardiac abnormalities noted by Dr. Castelo-Branco and colleagues included pericardial thickening and valvular abnormalities observable by echocardiography.

Acting on the hypothesis that Vieques residents might have developed abnormalities similar to those observed by the Portugese investigators, PSM investigators used commercial fishermen' trade association lists to recruit subjects in Vieques and a comparison group of fishermen in Ponce. Subjects were studied by echocardiography, with readings done by the PSM investigators. At the request of ATSDR and PSM, Dr. J. K. Oh of the Mayo Clinic in Rochester, Minnesota, performed blind independent readings of the same data. The PSM and Mayo Clinic investigators reported their findings at the expert panel review meeting (see Section 2.0). For the remainder of this report, the research conducted by PSM is referred to as "the Vieques Heart Study."

Earlier this year, the White House charged the Department of Health and Human Services (HHS) with scientifically evaluating reports of cardiovascular abnormalities occurring among residents of Vieques and potentially related to naval exercises. The Secretary of HHS delegated this task to ATSDR. The Governor of Puerto Rico also asked ATSDR to participate in this assessment. To respond to these requests, ATSDR has worked cooperatively with PSM investigators to evaluate the recent study of cardiac abnormalities - research that has raised a number of important questions on the frontiers of echocardiography, cardiology, and environmental epidemiology. The findings of this expert panel review meeting will be a critical input to ATSDR's response to the charge from the White House.

1.2 The Expert Panel

To organize a comprehensive and fair review, ATSDR and PSM each nominated candidates to serve as expert panelists for this meeting. The co-sponsors reviewed nominees and, by consensus, chose eight internationally recognized scientists to serve as the expert panelists. The final expert panelists have demonstrated expertise in one or more of the following fields: cardiology, echocardiography, cardiovascular epidemiology, or environmental epidemiology. These experts came from Mexico (2 panelists), Spain (2 panelists) and the United States (4 panelists) and had a variety of institutional affiliations (mostly academic). Although they were reimbursed for travel expenses, none of the co-sponsors paid the experts for the time spent during the meeting or reviewing reports. The remainder of this report refers to these eight individuals as panelists. Appendix A lists the eight expert panelists and their affiliations. Appendix B contains brief biographical sketches of the panelists.

Fourteen other individuals participated in the expert panel review meeting. These other participants included several representatives of the meeting co-sponsors: six individuals from ATSDR and five from PSM (including Dr. Julio Pérez, acting as consultant to PSM). They also included Dr. John Rullán, the Puerto Rico Secretary of Health, and Dr. J. K. Oh, Chief of the Echocardiography Core Lab at the Mayo Clinic. The remainder of this report refers to these 14 individuals as participants. Appendix A lists the names and affiliations of the 14 participants, and Appendix B presents brief biographical sketches of them.

Prior to the meeting, all panelists and participants received a package of background materials2,,, as well as a "charge to the reviewers" (Appendix C) that included both background information and a list of questions that the technical discussions at the meeting would be asked to address. The charge focused the reviewers' comments on four question areas: (1) study design and ascertainment, (2) echocardiographic measurements, (3) statistical analysis, and (4) interpretation and inference. The reviewers and participants did not receive written materials describing the PSM research on potential cardiovascular abnormalities among Vieques commercial fishermen (i.e., the Vieques Heart Study), because that work has not yet been published. For this reason, principal investigators from PSM and the Mayo Clinic gave detailed technical presentations on their research (see Section 2) before reviewers commented on the study.

1.3 The Expert Panel Review Meeting

The 2-day expert panel meeting took place at a conference room in the Hotel Condado Plaza in San Juan, Puerto Rico, on July 12 - 13, 2001. Presentations and discussions at the meeting generally followed the agenda (Appendix D). Two participants - Dr. David Fleming (ATSDR) and Dr. Manuel Martínez Maldonado (PSM) - acted as meeting co-chairs and moderators for the discussions among the panelists and participants.

As the agenda shows, the meeting began with introductory remarks and background presentations from four participants. Summaries of these presentations are provided below. The technical presentations on the Vieques Heart Study followed the opening remarks. For the remainder of the meeting, the panelists engaged in free-flowing discussions organized around the four question areas in the charge. During the discussions, reviewers and participants offered their individual perspectives on the Vieques Heart Study and associated topics. Although panelists agreed on many points, no effort was made to reach an absolute consensus on any issue. At the end of the meeting, the reviewers and some participants prepared written summaries of their comments (Appendix E).

Summaries of the four background presentations follow (refer to Section 2 for summaries of the technical presentations on the Vieques Heart Study):

Dr. Manuel Martínez Maldonado, President and Dean, PSM. Dr. Martínez Maldonado is the principal investigator of the Vieques Heart Study. He provided background information on how and why PSM initiated the Vieques Heart Study. He stated that PSM investigators decided to conduct the study after being approached by a group of physicians, including an ad honorem PSM faculty member, with preliminary data on possible health effects associated with noise and vibration. He noted that, in November 2000, a PSM-appointed committee of researchers designed a study to investigate further the preliminary data.

Dr. Martínez Maldonado reviewed in general how the study was implemented. He stressed, for instance, that his primary concern was to conduct the study following rigorous scientific methods. Accordingly, the researchers used a new Agilent echocardiogram machine. In addition, Dr. Martínez Maldonado said that while working on the project, he tried to ensure that all researchers remained objective and impartial.

Finally, Dr. Martínez Maldonado provided some background information on Vieques. The island lies east-southeast of Puerto Rico, spans roughly 25 square miles, and has roughly 9,000 residents. The eastern and western ends of the island have been Navy property, and the Navy has conducted bombing exercises and other war games on the easternmost part of the island for the last 60 years. Dr. Rullán provided additional background information on Vieques during his presentation, summarized below.

Dr. David Fleming, Deputy Administrator, ATSDR and Deputy Director for Science & Public Health, CDC. Dr. Fleming offered brief remarks on the political and scientific issues related to Vieques, encouraging participants and panelists to focus their discussions strictly on science. Although he acknowledged that the future of Vieques is the subject of intense political debate, he stressed that the panelists are not being asked to offer solutions on this matter. Dr. Fleming also stressed that the purpose of the meeting was strictly for the expert panelists to provide their individual opinions on the scientific issues relevant to the Vieques Heart Study, without attempting to reach consensus on any issue. After making these statements, at Dr. Fleming's invitation, panelists and participants introduced themselves, noting their affiliations and areas of expertise.

Dr. Edwin M. Kilbourne, Associate Administrator for Toxic Substances & Public Health, ATSDR. Dr. Kilbourne spoke briefly about the goal of the meeting and raised the subject of the earlier pilot study that preceded the Vieques Heart Study. He indicated that the purpose of the meeting was to solicit individual expert opinions on issues relevant to the Vieques Heart Study, whether or not those comments were consistent with others offered. Dr. Kilbourne encouraged the panelists and participants to provide verbal input throughout the meeting, and he added that panelists would be asked to submit written comments as well. Dr. Kilbourne asked that panelists and participants frame their discussions around the four question areas listed in the charge (Appendix C). Discussions on topics outside these question areas were permitted, but not encouraged.

Dr. Kilbourne then clarified some issues surrounding the earlier pilot study by Drs. Torres Aguiar, Castelo Branco and others involving echocardiographic data that led PSM to initiate the Vieques Heart Study.3 First, Dr. Kilbourne emphasized that the earlier investigation was not the focus of the meeting; the panelists' discussions should focus on the subsequent data generated by PSM. He asked, however, that the panelists submit written comments on the earlier study, because ATSDR must address the public health implications of both this earlier study and the Vieques Heart Study data when responding to the White House inquiry regarding cardiovascular abnormalities in Vieques.

Finally, regarding the availability of the Vieques Heart Study data, Dr. Kilbourne stated that ATSDR and PSM are still analyzing and reviewing the results. Thus, no written summaries of the study are available for distribution. However, he added that the PSM investigators might be able to supplement their presentation by generating selected data summaries during the meeting, if necessary.

Dr. John Rullán, Secretary of Health, Puerto Rico Department of Health. Dr. Rullán's presentation addressed three topics: (1) changes in the Puerto Rico health care system that have affected access to health care in Vieques, (2) findings from Puerto Rico's National Health Review Survey, and (3) cancer statistics for Puerto Rico. First, Dr. Rullán described how the public health care system in Puerto Rico has changed over the years, focusing on how these changes have affected access to health care on Vieques. Specifically, he noted that the 1993 privatization of health care in Puerto Rico caused primary care centers throughout the Commonwealth to cut many services. Dr. Rullán recalled that services at the primary care center in Vieques were cut dramatically throughout the 1990s. The primary care center now operates without a laboratory, delivery rooms, and other services it once provided, causing many residents to travel to the main island of Puerto Rico to seek certain types of medical care. However, Dr. Rullán indicated that services offered by the primary care center in Vieques are now expanding, and he listed specific examples of scheduled improvements.

Second, Dr. Rullán presented a series of demographic and statistical data comparing the population in Vieques to the entire population of Puerto Rico. These data came from several sources, including the 1990 United States Census and a 1998 health interview survey administered by the Puerto Rico School of Public Health. Dr. Rullán explained that this latter data source is similar to the National Health Interview Survey, but was tailored to the population of Puerto Rico. Data from the health survey are based on responses from 250 households on Vieques, which are felt to be representative of the entire island's population. Dr. Rullán then presented numerous statistics including, but not limited to, unemployment rates, teenage pregnancy rates, proportion of population receiving prenatal care during the first trimester, per capita income, mortality rates for specific diseases, and self-reported morbidity rates for different types of diseases.

Third, Dr. Rullán reviewed the Puerto Rico Department of Health's recent efforts to update its cancer registry. For background, Dr. Rullán said that between 1950 and 1991 the Puerto Rico cancer registry was an excellent resource. Following the 1991 health care reforms and associated cutbacks in certain public health services, health officials have not been able to produce annual reports from the existing cancer registry data, and the cancer registry for Puerto Rico is not yet updated. Recognizing the value of accurate cancer registry data, the Puerto Rico Department of Health, in collaboration with CDC, has worked to update the cancer registry for the years 1997 - 2000. Dr. Rullán reported the progress of the updates, indicating that he expects in August 2001 to report 1997 cancer incidence and mortality data for Vieques and all of Puerto Rico. Dr. Rullán indicated that preliminary results suggest that cancer mortality at Vieques is higher than that of Puerto Rico. Finally, Dr. Rullán acknowledged that interpreting these and other health outcome data for Vieques is complicated by various confounding factors, such as the limited access to, and the quality of, health care on the island.

At the end of his presentation, Dr. Rullán and another participant provided some background information on Navy operations at Vieques. Referring to a map of Vieques, the participant indicated where the Navy exercises take place and said that, prior to 1999, these exercises occurred from sea, air, and land 180 days per year. He indicated that Navy exercises temporarily ceased in 1999, when a watchman was killed by an errant bomb dropped on the island. Since 1999, exercises have been extremely limited. This participant offered additional information on legal issues related to noises generated by the Navy exercises.

Following his presentation, Dr. Rullán answered several questions about the health outcome data he presented. When one panelist asked if data are available on trends in cancer rates over time; Dr. Rullán responded that such data should be available after the cancer registries are updated. He added that data trends might determine whether further research into cancer incidence and mortality (e.g., case control studies) is warranted. Another panelist asked if the cancer registry data are organized in a fashion that will allow Puerto Rico Department of Health to compare data for Vieques with data for other communities with similar characteristics, such as per capita income and socioeconomic status. Dr. Rullán responded that such data are not currently available, but likely will be within several months. Another panelist asked if anyone could elaborate on the ancestry and primary occupations of the Vieques population. A participant noted that the population of Vieques is believed to be of the same heritage as the population of Puerto Rico; he added that 105 of the roughly 9,000 residents of Vieques are registered commercial fishermen.

1.4 Report Organization

The remainder of this report is organized according to the four question areas listed in the charge to the reviewers. First, Section 2 reviews two technical presentations given at the beginning of the meeting. Then, Sections 3, 4, 5, and 6 summarize the discussions on study design and data ascertainment, echocardiographic measurements, statistical analysis, and interpretation and inference, respectively. At the end of the meeting, the panelists were asked to summarize their overall impressions of the Vieques Heart Study in writing and then share their final thoughts with the group. Section 6 summarizes these final remarks along with the panelists' comments on interpretation and inference.

All tables cited in the text appear at the end of the section in which they are mentioned. Full citations for references are provided in Section 7..

2.0 TECHNICAL PRESENTATIONS ON THE VIEQUES HEART STUDY

This section summarizes the two technical presentations specifically addressing the Vieques Heart Study. Because no written materials were available on the Vieques Heart Study, the content of the technical presentations largely formed the basis for the panelists' reviews. Accordingly, this report summarizes the two presentations in detail, as well as the question and answer sessions that followed. The presentations are summarized in the order in which they were given.

Note that the content of this section is almost entirely based on the presentations of two participants. Sections 3 through 6 of this report summarize the panelists' comments and findings regarding the information provided.

2.1 Summary of Presentation by Dr. Oh, Mayo Clinic

Dr. Oh opened his presentation by reviewing his and the Mayo Clinic's experience with echocardiography. He indicated that the "Echo Lab" at the Mayo Clinic conducts 200 echocardiographic examinations daily and is currently staffed by 80 sonographers, including 7 research sonographers, who have performed at least 3 years of clinical sonography to receive that designation. In addition, Dr. Oh indicated that he manages the Mayo Clinic's Echo Core Lab, established specifically to read and interpret echocardiographic results for clinical trials. For the Vieques Heart Study, the Core Lab recruited three experienced research sonographers to read PSM's echocardiographic images, and all three operated under Dr. Oh's direct supervision. The following subsections review the main topics of Dr. Oh's presentation.

2.1.1 Background on Echocardiography

Dr. Oh briefly reviewed the application of echocardiography, focusing exclusively on transthoracic echocardiography - the type used by PSM investigators in the Vieques Heart Study. Dr. Oh explained that echocardiographic readings can be taken in various modalities, such as M-mode, 2-dimensional imaging, and Doppler echocardiography. For perspective, Dr. Oh displayed sample outputs from these three modalities, all of which were used in the Vieques Heart Study:

• 2-dimensional echocardiography. Dr. Oh showed several videos of 2-dimensional echocardiographic readings from both the parasternal and apical views, as well as different transducer orientations for these views. He identified the different heart structures visible from these views and listed various dimensions and parameters that can be calculated from these images.
• M-mode echocardiography. Dr. Oh displayed several sample images obtained from M-mode echocardiography. He indicated the dimensions that the Mayo researchers routinely measure using M-mode (e.g., aortic valve opening, left atrial dimension, left ventricle dimension). Dr. Oh noted that his research team typically does not measure pericardial thickness using M-mode echocardiography, primarily because pericardial disease typically manifests as hemodynamic problems and impaired diastolic function, which can be assessed directly with Doppler echocardiographic measurements or other tests.
• Doppler echocardiography. Dr. Oh explained how Doppler echocardiography is used to determine the velocity of blood through various heart structures and showed some sample outputs. Several hemodynamic parameters can then be calculated from the velocity measurements. These parameters include stroke volume, cardiac output, and intracardiac pressures.

2.1.2 Mayo Clinic's Readings of Echocardiographic Images from the Vieques Heart Study (See appendix E.9)

Following his background presentation on echocardiography, Dr. Oh commented specifically on how his research team read the echocardiograms from the Vieques Heart Study, (copies of which the PSM investigators provided on 7 CD-ROMs). The images did not include any reference to the subjects from which they were collected (i.e., the Mayo readers were "blinded" to the subjects). The Mayo Clinic was not involved with the original acquisition of the echocardiograms in the Vieques Heart Study. Dr. Oh emphasized that his team, comprised of himself and three research sonographers, employed a standardized approach to process the data. The approach was designed to generate the data requested by ATSDR and PSM, while also characterizing intra-observer and inter-observer variability in measurements, as summarized below:

• Approach for measuring parameters other than pericardial thickness. The research team included two primary sonographers and a secondary sonographer. The secondary sonographer received the echocardiographic images, randomly divided the images into two sets, and assigned the sets to the two primary sonographers for reading. The primary sonographers measured a large set of parameters, such as left ventricular dimension at different points in the cardiac cycle. To characterize intra-observer variability in these measurements, both sonographers re-read 10 randomly selected echocardiograms from their original set, but more than 10 days after the first readings were made. To characterize inter-observer variability, the secondary sonographer measured 30 randomly selected echocardiographic images, or roughly one-third of the total echocardiograms reviewed. Dr. Oh later presented the intra- and inter-observer variability data for measuring parameters other than pericardial thickness (see Section 2.1.3).

  Finally, Dr. Oh visually reviewed the sonographers' data, resolved any significant data discrepancies observed between the two sonographers' readings, and conducted random quality control checks. Dr. Oh entered the primary sonographers' measurements into the final data set, except in the few cases when significant discrepancies were observed between the primary and secondary sonographer. In these cases, Dr. Oh's measurements were used as the final result.

• Approach for measuring pericardial thickness. Given the emphasis on pericardial thickening, Dr. Oh implemented a more rigorous reading procedure for determining pericardial thickness than that used for measuring other structural and functional parameters. Rather than having each of the two research sonographers read half of the total images, Dr. Oh instructed both sonographers to measure pericardial thickness in every echocardiographic image provided by PSM. Even though two sonographers measured pericardial thickness on every image, all images had a randomly assigned "primary" and "secondary" sonographer. Additionally, Dr. Oh read every image for pericardial thickness.

  As with the measurements for other parameters (see previous bulleted item), the primary sonographer's readings of pericardial thickness were considered the final results, except in cases with significant discrepancies between the readers. As stated, in these cases Dr. Oh's readings were considered final. He indicated that this occurred in only a small subset of the recordings Mayo reviewed

• Measurement of mitral leaflet thickness. At the request of a consultant to PSM, Dr. Oh's research team also measured mitral leaflet thickness. These measurements were conducted after the measurements listed above were completed. Time constraints prevented Dr. Oh from having two sonographers make these measurements. He did not discuss these measures further.

• After describing the general approach used to read the echocardiographic images, Dr. Oh spoke more specifically on how sonographers read each individual recording. He indicated that sonographers made all readings by loading the echocardiographic images onto a Digisonic workstation, which was calibrated for each image being viewed. Dr. Oh noted that the calibration step was necessary because the echocardiograms provided were collected from different views (i.e., parasternal and apical) and different orientations. At PSM's request, the Mayo sonographers always measured pericardial thickness at end diastole, as determined by the output from the electrocardiogram (EKG) on the echocardiographic image. Dr. Oh instructed the sonographers to measure pericardial thickness in at least three different cardiac cycles for each image. The thicknesses were measured visually using calipers, both in unmagnified and magnified views.

2.1.3 Variability in Measuring Pericardial Thickness Using Echocardiography

To characterize the precision of the sonographers' measurements, Dr. Oh determined both inter-observer and intra-observer variability from data generated during duplicate measurements. First, Dr. Oh presented data on inter-observer variability; that is, comparisons of measurements that the two sonographers reported for the same echocardiographic images. These data suggested relatively strong agreement (R2 values ranging from 0.6 to 0.93) between the two sonographers for structural parameters (e.g., left ventricular mass) and functional parameters (e.g., deceleration time).However, the agreement for measures of pericardial thickness from nonmagnified images was weak (R2 = 0.22). Dr. Oh also indicated that the agreement was weak for reading pericardial thickness from magnified images, but he did not present any data characterizing the correlation.

Second, Dr. Oh presented data on intra-observer variability, but only for a limited set of measures. He indicated that agreement between sonographers on measurements of left ventricular end-diastolic dimension was excellent (R2 = 0.93), while the agreement between sonographers on measurements of pericardial thickness from nonmagnified images was relatively poor (R2 = 0.30). Dr. Oh noted that the intra-observer variability was also weak for measuring pericardial thickness in magnified images, although he did not provide the correlation statistics.

Some panelists asked Dr. Oh to comment on whether variability among pericardial thickness measurements might appear to be high because the observed values fall into a relatively small range. Dr. Oh and a panelist indicated that the range of pericardial thicknesses in the Vieques Heart Study is less than the ranges of the other functional and structural parameters measured, although the variability in measurements of pericardial thickness (in relation to its absolute value) is greater. Dr. Oh added that all measurements made from the Doppler echocardiograms appeared to be highly precise, while measurements of pericardial thickness were not. He concluded that the variability in pericardial thickness measurement was much greater than that of measuring other cardiac dimension or hemodynamic parameters.

Concerned about the implications of the variability data, a panelist asked Dr. Oh if any systematic differences were observed between the primary sonographers' measurements. A participant shared this concern, wondering why the measures were highly precise for all parameters except pericardial thickness. In response, Dr. Oh showed the distributions of measurements of pericardial thickness made by himself and by the primary sonographers. The medians of the distributions for the nonmagnified readings were approximately 1.3 mm, 1.7 mm, and 1.7 mm. Dr. Oh indicated that an 0.4 mm difference in medians should not be viewed as a significant error, given that he did not think echocardiography can achieve a resolution finer than approximately 1.0 mm (see Section 2.1.4).

2.1.4 Limitations of Using Echocardiography to Measure Pericardial Thickness

Dr. Oh acknowledged that echocardiography has many advantages, but he noted that it has several limitations, including limited precision in measuring structures at sub-millimeter dimensions. Following are some of Dr. Oh's specific comments on the limitations of using echocardiography for the Vieques Heart Study:

• Concerns about how image acquisition affects echocardiographic images. Dr. Oh was concerned that practices of the echocardiogram operator at the time of image acquisition might have influenced the quality of images acquired. For instance, he noted that inconsistent use of gain settings, particularly time-gain compensation (TGC), can affect subsequent measures of pericardial thickness - an issue the panelists revisited when discussing Question Area 2 (see Section 4.0). He added that the angle between the transducer and the chest wall might have varied from subject to subject, although he noted that the angle would have to change substantially before causing significant errors in measured pericardial thickness. Finally, a panelist asked if inconsistent placement of the transducer might bias results. Dr. Oh, however, was not very concerned about the placement, noting that all pericardial images were acquired from the posterior portion of the left ventricle.
• Difficulties in reading images after acquisition. Dr. Oh noted that even if no systematic biases affected image acquisition,, measuring dimensions of fine cardiac structures from echocardiographic images is a complicated task. First, he explained how M-mode echocardiographic images clearly show that pericardial thickness varies throughout the cardiac cycle. Thus, even though PSM and Mayo Clinic investigators both measured the thickness at "end diastole," the thickness can vary even over short time frames that can reasonably be considered as end diastole. Second, Dr. Oh indicated that the pericardial thickness evident from an echocardiographic image can vary significantly across cardiac cycles of an individual subject. His approach was to average the thicknesses observed in multiple cycles while rejecting any thicknesses that appear to be outliers. Third, Dr. Oh showed images for which sonographers had difficulty determining exactly where the leading edge of the pericardium begins and where the trailing edge ends. Overall, the Mayo Clinic investigators rejected 29 images due to difficulty in measuring the pericardium thickness, primarily due to blurring of the pericardial image at end diastole or overlap of the pericardial image with other structures (e.g., posterior wall). But even for the images of sufficient quality to read, Dr. Oh had concerns about the precision of the measured pericardial thickness, as Section 2.1.4 describes in greater detail.
• Scientific literature on how accurately echocardiographic images portray pericardial thickness. Dr. Oh referred to results from the only scientific publication he knew of comparing echocardiographic measurements of pericardial thickness in laboratory animals against actual anatomical thicknesses measured after the animals were sacrificed. The study found that pericardial thicknesses measured from echocardiographic images were consistently greater than the actual thicknesses, and that echocardiographic measurements and actual measurements were not strongly correlated (R2 = 0.34). Dr. Oh noted that this study also showed that changes to the gain settings in the ultrasound caused the dimensions of the pericardia to appear larger than they actually are.

Given these concerns, Dr. Oh said that he would not have recommended using transthoracic echocardiography to determine sub-millimeter differences in pericardial thickness. Transesophageal echocardiography and other imaging techniques (e.g., computed tomography) are known to achieve finer resolution. As discussed later in the report, however, the PSM investigators did not intend to use echocardiography to detect sub-millimeter differences in pericardial thickness: the goal of the Vieques Heart Study was to determine whether the pericardia among Vieques fishermen was more than 1 mm thicker than those of a comparison population (see Section 4.3).

2.2 Summary of Presentation by Dr. Ríos, Ponce School of Medicine

Dr. Ríos summarized various aspects of the Fishermen Cardiovascular Study (referred to in this report as the Vieques Heart Study), focusing primarily on the study methods and results. Dr. Ríos first identified the investigators who contributed to the study, then indicated the study objective: "to determine the occurrence of an association between place of residence and the presence of morphological cardiovascular changes among fishermen." To meet this objective, the team of investigators designed and implemented a cross-sectional epidemiological study, the details of which are reviewed in the following sections.

2.2.1 Methods Used for the Vieques Heart Study

Dr. Ríos described the methodology used to implement the Vieques Heart Study. First, the researchers attempted to recruit 80 commercial fishermen from Vieques and another 80 from Ponce de Playa (Ponce) to participate in the study. The investigators' power calculations indicated that this total number of subjects (160) was necessary to detect an average pericardium size difference of 0.5 mm. Details on recruiting subjects follow:

• Vieques fishermen. The registry of licensed fishermen in Vieques was the sampling frame. The investigators randomly selected and subsequently contacted 80 individuals from this registry. Of these 80 individuals, 2 were deceased, 10 no longer lived in Vieques, and 16 refused to participate in the study, leaving a sample size of 53 Vieques fishermen (because of overlap among excluded groups). No efforts were made to characterize non-participants.
• Ponce fishermen. Because the registry of commercial fishermen in Ponce included 52 individuals, the PSM investigators attempted to recruit all fishermen in the registry, not a subset of the individuals. Of the individuals in the registry, the PSM investigators recruited 43 to participate in the study.

All 96 subjects were then asked to complete a questionnaire. The questionnaire addressed basic demographic information (e.g., age and place of residence) and numerous questions concerning potential confounding factors. These confounding factors included previous diagnosis of conditions or diseases that might be associated with pericardial thickening, as identified by cardiologists on the research team. Such diseases and conditions include tuberculosis, high blood pressure, scleroderma, lupus, diabetes, and others. All responses on these health status variables were self-reported. One panelist asked Dr. Ríos about potential biases associated with using self-reported morbidity data in communities known to have very limited access to health care, such as Vieques (see Section 1.3). This panelist was particularly concerned that Vieques residents might not be aware of confounding diseases or health conditions, because they might never have been diagnosed. Dr. Ríos acknowledged that such biases, if any, are impossible to characterize, although he expressed confidence that subjects provided accurate responses regarding their health histories.

Dr. Ríos then reviewed details on how echocardiographic images were collected and analyzed. Echocardiographic examinations were conducted on all 96 subjects, and both groups of subjects were examined in the communities where they reside. One certified technician administered the examinations for all subjects. A group of cardiologists then met at Ponce to read the echocardiographic images. The readers were blinded to information on the subjects (i.e., they did not know if a given image was for a subject from Ponce or from Vieques). All readings were done by consensus among the observers; no repeated measures were obtained to characterize reproducibility.

When responding to panelists' questions, Dr. Ríos and other PSM investigators provided additional information on the study methods. For example, the study did not include any measures of exposure to noise, because the study objective was simply to identify differences between the groups in cardiac structure, without attributing them to any factor other than location of residence. Dr. Martínez Maldonado clarified that the Vieques Heart Study considered fishermen because they are believed to be the subset of Vieques residents most exposed to noise during the Navy exercises. Fishermen, he said, are consistently closer to the Navy ships than any other subset of Vieques residents.

2.2.2 Preliminary Results of the Vieques Heart Study

Dr. Ríos presented findings from the questionnaires and the echocardiographic readings. First, he noted that the average age of the Vieques subjects (N = 53) was 45, while the average age of the Ponce subjects (N = 43) was 55, and that this age difference was statistically significant. He added that both study groups were very similar in terms of the health conditions reported on the questionnaires: some slight differences were observed between the groups, but none was statistically significant. The groups also were similar in terms of highest level of education received. Dr. Ríos argued that this parameter is probably the best reflection of socioeconomic status for the study groups. Dr. Ríos added that no statistically significant differences existed between the populations in several other parameters, including smoking status, height, and weight. He then emphasized that the only statistically significant difference between the populations was average age.

Next, Dr. Ríos presented the main finding from PSM's reading of the echocardiographic images: the pericardia of Vieques fishermen were, on average, thicker than those of Ponce fishermen; and the "aortic valve thickness" among Vieques fishermen was greater than that among the control group. Both differences in average thickness were statistically significant. Dr. Ríos indicated that these statistically significant differences remained even after the investigators adjusted for age. Statistical analyses were not performed to test whether other variables (e.g., smoking history) are confounding factors. Table 2-1 presents the preliminary data, and compares the results generated by the PSM investigators with those generated by the Mayo Clinic investigators. Section 2.2.3 discusses the measurements of pericardial thickness in greater detail.

Finally, Dr. Martínez Maldonado presented data on measurements other than pericardial thickness made from the echocardiographic images. He first displayed average measurements for the following parameters: left ventricle outflow tract velocity, height of the E-wave, height of the A-wave, left ventricle outflow tract diameter, end-diastolic volume of left ventricle, and others. For all of these parameters, Dr. Martínez Maldonado stated that the average readings generated by the Mayo Clinic were not considerably different from those generated by the PSM investigators, and no differences were statistically significant. Moreover, for this list of parameters, his data indicated that the average readings for the Vieques subjects (as determined by either team of investigators) were not significantly different from those for the Ponce subjects. For end-systolic volume of left ventricle, however, the measurements by the two teams of investigators differed, on average, by approximately 45%. When discussing this difference further, PSM investigators indicated that they used the Simpson method to measure the end-systolic volume of the left ventricle, while the Mayo Clinic investigators used the bullet method. Several participants opined that the difference in the measurement of end-systolic volume was consistent with the fact that the two groups used different methods to measure this variable.

2.2.3 Questions and Answers

After reviewing the methods and results of the Vieques Heart Study, Dr. Ríos and the other PSM investigators responded to the panelists' and participants' questions, as summarized below:

• Comments on sample sizes for the Mayo Clinic and PSM data. After Dr. Ríos presented the data shown in Table 2-1, panelists asked Dr. Ríos and the PSM investigators whether the difference in the results (that is, a statistically significant difference between groups in pericardial thickness using PSM measurements but not Mayo measurements) might arise from the two research teams having different sets of images. (The Mayo investigators had rejected some images for pericardial measurement. These were not precisely the same images as had been rejected by the PSM investigators.) Dr. Ríos noted that sample sizes were different because the Mayo Clinic investigators rejected more echocardiographic images than did the PSM investigators. The panelists asked the PSM investigators to present paired comparisons of the pericardial thickness measurements for those subjects for whom a pericardial measurement was made by both teams of investigators. This comparison was displayed later in the meeting (see Section 3).
• Multivariate statistical analyses. Two panelists asked whether PSM investigators conducted multivariate statistical analyses to investigate potential confounders. Dr. Ríos responded that their statistical methodology was to consider only those factors with a statistically significant difference (p < 0.05) between the study and control populations as potential confounders. Because age was the only parameter found to have statistically significant differences between the populations, Dr. Ríos indicated that PSM investigators controlled only for age when analyzing data on pericardial thickness, but they did not do so for any other variable. Later in the meeting, Dr. Ríos presented summary statistics for various characteristics of the study populations, such as age, weight, smoking history, and a recollection of being diagnosed with certain diseases.
• Measurements of pericardial thickness from unmagnified views. After viewing the data on pericardial thickness measurements made from magnified views (see Table 2-1), one participant asked if a similar comparison could be made between pericardial thickness measurements made from unmagnified views. Dr. Martínez Maldonado responded that such a comparison is not possible, because the PSM investigators measured pericardial thickness using only magnified views.
• Clinical significance of findings. Some panelists began to discuss the clinical significance of pericardial thickening during the question and answer session - an issue the panelists discussed in greater detail when responding to Question Area 2 (see Section 4). During this discussion, Dr. Ríos clarified that the purpose of the Vieques Heart Study was to determine whether a group of Vieques fishermen had thicker pericardia than a comparison group from Ponce. He added that addressing clinical significance of thicker pericardia would be a logical follow-up research project, if such an outcome were observed.
• Measurements other than pericardial thickness. Although the primary objective of the Vieques Health Study was to characterize pericardial thickness among the two study populations, one participant thought the available data allow for more detailed analyses of ventricular function (including its age-dependence) and the dimensions of other heart structures. He advocated conducting these additional analyses both because they are conventionally measured using echocardiography and because they have much greater clinical relevance than sub-millimeter increases in pericardial thickness. In short, this participant advocated expanding the data analysis beyond the original objectives of the Vieques Heart Study. The PSM investigators commented on this issue later in the meeting (see Sections 4.5 and 5).
• Other comments. One panelist asked the panel to comment on what, if anything, is known about the age-dependence of pericardial thickness in the overall population, given that age was apparently the main difference between the study and control populations. No panelists or participants responded. Finally, another panelist recommended, and other panelists agreed, that the two teams of investigators collaborate on a publication on inter-institutional variability in echocardiographic measurements of pericardial thickness.

Table 2-1 - Average Pericardial Thicknesses Based on Echocardiographic Readings
Conducted by PSM and by Mayo Clinic

3.0 COMMENTS ON STUDY DESIGN AND DATA ASCERTAINMENT

Question Area 1 in the charge to the reviewers (see Appendix C) addressed the design of the Vieques Heart Study and data ascertainment. The charge question asked: "How effectively does the study design minimize the possibility that bias or confounding explain any of the observed associations? Please consider the following issues in your answer:

- Identification and comparability of sampling frames
- Sampling (of exposed and control persons)
- Non-response and measures taken to deal with it
- Ascertainment of non-echocardiographic information related to exposed persons and controls
- Availability of data on potential confounders"

Discussions on this question area opened with the PSM investigators presenting some information on study design not mentioned in their opening presentation (see Section 2.2). A PSM investigator explained that the sampling frame was intended to be 80 fishermen from Vieques and 80 fishermen from Ponce. Subjects were selected from the fishing registries for the two locations. Because the fishing registry at Vieques included more than 80 individuals, the PSM investigators randomly selected a subset of 80 as the sampling frame. The fishing registry at Ponce included 52 individuals, all of whom were considered as the sampling frame. The PSM investigators noted that the two communities have licensing requirements for all commercial fishermen, thus ensuring that the registries themselves do not include a non-random subset of the overall fishing population.

Regarding comparability of the populations, the PSM investigators noted that the only statistically significant difference between the study and control groups was age (the Ponce population being 10 years older, on average, than the Vieques population). The information collected on medical status and other demographic data revealed no statistically significant differences between the two groups. Medical status information included blood pressure, alcohol usage, and past diagnoses of diseases known or suspected to be associated with pericardial thickening (e.g., diabetes, lupus, scleroderma, tuberculosis).

The subsequent discussions focused on various aspects of Question Area 1, and are summarized below in five main categories:

• Response rates. The panelists asked the PSM investigators to present additional detailed data on sampling frames and response rates. A PSM investigator cited a series of relevant data on this matter, summarized in Table 3-1. The data indicated that, on average, nearly 80% of the individuals in the selected sampling frame agreed to participate in the Vieques Heart Study.

• Referring back to the summary statistics presented earlier (see Table 2-1), a panelist asked why the number of individuals who agreed to participate in the study (96; see Table 3-1) exceeded the number of recordings considered when calculating pericardial thickness (84; see Table 2-1). The PSM investigators explained that some echocardiographic images were rejected from the analysis because they did not have corresponding EKG data, which are needed to ensure that all readings take place at the same point in the cardiac cycle. Some discussion followed on whether response rates should be calculated as the number of valid images obtained divided by the sampling frame or as the number of individuals who agreed to participate divided by the number of eligible participants. No resolution was reached on this matter.

  The data on response rate triggered two additional concerns. First, two panelists wondered why a considerably larger proportion of echocardiographic images collected on Vieques were rejected for measurement of pericardial thickness than those collected on Ponce (see Table 3-1). Given the unlikely possibility that this difference resulted merely from chance, the two panelists thought the different proportion of rejected images suggested that the technician who administered the examinations might not have done so in an identical fashion in the two study populations. The panelists revisited this issue when discussing Question Area 2 (see Section 4). Second, a participant wondered if the Vieques Heart Study might lack statistical power, given that the power calculations supported the use of 160 total subjects and only 85 valid echocardiographic images were collected. The PSM investigators noted that their power calculations were very conservative and their sample size was adequate to detect the differences in pericardial thickness (i.e., greater than 3 mm) reported in the pilot study.3

• Issues of non-response. The panelists discussed three issues pertaining to non-response and the PSM investigators' attempts to deal with it. First, several panelists and participants asked the PSM investigators if they made attempts to characterize the non-respondents, possibly through a questionnaire. The PSM investigators indicated that their study, as approved by the Institutional Review Board (IRB) did not include any provision for such follow-up activities. As a result, they were not authorized to ask any questions of individuals who did not sign informed consent forms. Given the importance of ensuring that non-respondents account for a random subset of the sampling frame, several panelists and participants recommended that the PSM investigators request permission from the IRB to conduct a brief questionnaire study of non-participants.

• Second, the panelists had some concerns regarding the use of compensation to recruit subjects. When asked about measures to increase response, the PSM investigators indicated that they paid a subset of the Ponce fishermen $15 each to increase response rates among the control group, while no such efforts were made to increase response rates among the Vieques fishermen. One panelist commented that, with this approach, the subjects in the study had different motivations to participate - a fact that could lead to bias in the results. Others agreed, noting that compensating study participants is standard practice in many epidemiological studies and clinical trials, but compensating only a subset of the study participants is not. These panelists indicated, however, that the PSM investigators can assess the potential bias associated with the compensation practices by comparing the average pericardial thickness among the Ponce fishermen who were compensated to that among the Ponce fishermen who were not.

  Finally, noting that the final number of subjects with valid echocardiographic images (85) was roughly half the number of subjects identified in the power calculations (160), a participant wondered why the PSM investigators did not attempt to identify and recruit additional subjects, perhaps including those not in the fishing registries. A panelist responded that the approach of limiting the sampling frame to fishermen is appropriate, given that an underlying hypothesis in the study is that exposure to noise from Navy bombing exercises causes pericardial thickening and that Vieques fishermen (as opposed to other island residents) are believed to be the population most exposed to this noise.

• Data on potential confounders. Before the panelists discussed potential confounders, the PSM investigators explained their approach to identify and characterize such variables. After consulting with several physicians, the investigators developed a list of potential confounders, including various illnesses (e.g., lupus, scleroderma, kidney disease, diabetes, tuberculosis) and other indicators of health status (e.g., smoking history, alcohol usage, blood pressure, frequency of chest pain). The PSM investigators commented that the two study populations had no statistically significant differences in any of the potential confounding factors, except for a significant difference observed in the average age of the study populations. The PSM investigators added that all participants were given physical examinations, during which certain potential confounding factors were measured (e.g., blood pressure) and others related to medical history and co-morbid factors were discussed.

• The panelists had two general concerns about the potential confounding factors. First, regarding the potential confounders identified by PSM investigators, a panelist wondered how accurate self-reported morbidity data are for the study populations, which reportedly have limited access to medical care. The PSM investigators shared this concern, but doubted that the morbidities considered as potential confounders (e.g., lupus, scleroderma) would be undetected - both historically for a given subject and during the provided physical examinations. A participant agreed, noting that Puerto Rico has active surveillance systems for tuberculosis and that the likely study populations do access medical care for more serious morbidities, such as lupus.

  Second, the panelists wondered if confounders other than those identified by the PSM investigators might bias results. For example, two panelists and a participant commented that pericarditis is associated with conditions that are not readily diagnosed, particularly viral infections. One panelist referred to studies from the 1960s to more recently in which electrocardiographic and clinical evidence of pericarditis were found in individuals with a variety of viral infections. The mild myopericarditis usually resolved over time.,,,,, Referring to the findings of one of his own studies, a participant indicated that the second most common cause of constrictive pericarditis among a sample population was believed to be viral infection, and that many subjects in the study were completely unaware of these infections. Commenting on these concerns, an ATSDR scientist said that some infections that might cause pericardial thickening (e.g., coxsackievirus infections) are difficult to detect, especially among populations with limited access to medical care. In short, these comments suggested that the Vieques Heart Study might not have considered all possible confounding factors for pericardial thickening, although in any event, confounding by viral infection would have been difficult to ascertain.

  In response, a PSM investigator doubted that a widespread pericarditis-causing viral infection (e.g., coxsackievirus) could have gone undetected among either study group. A panelist added that gathering information on past viral infections is extremely difficult and he doubted whether an alternate study design would have been more effective at identifying past exposures to pericarditis-causing viral infections.

• Comparability of study and control groups. When comparing Vieques and Ponce fishermen, the panelists and participants first recounted comments made during the opening presentations (see Section 1.3). For instance, a panelist recalled that the Vieques population has to travel to the main island of Puerto Rico for most types of health care beyond general practice; and one of the co-chairs noted that the demographic data suggest that the population of Vieques is poorer than that of Puerto Rico as a whole. Following these observations, a panelist recounted several ways in which the two study populations differ:
• • The Vieques population was not compensated for participating in the study.
  • The Vieques population, on average, was younger than the Ponce population.
  • The Vieques population does not have as immediate access to health care as the Ponce population and might therefore be less likely to have underlying pathologies diagnosed.
  • The quality of the echocardiographic readings, as determined by the proportion of rejected images, differs between the two populations.
    
  
  Given these differences, the panelist wondered if the relatively small difference observed in pericardial thickness, as determined by the PSM investigators, might simply reflect the populations' dissimilarities. The panelists discussed in detail issues of measurement when they addressed Question Area 2 (see Section 4).
  

  Other issues. The panelists and participants raised the following additional questions about study design, but these were not discussed in great detail. First, a CDC scientist was concerned that individuals who were identified in the pilot study as having thickened pericardia might be more likely to participate in the Vieques Heart Study. The PSM investigators indicated that they did not ask about participation in past studies when recruiting subjects for the Vieques Heart Study. Second, a participant asked if the PSM investigators considered infection with the human immunodeficiency virus as a potential confounder. The PSM investigators indicated that they did not. Third, a panelist asked if the PSM investigators collected any data characterizing subjects' exposures to noise. The PSM investigators acknowledged the utility of such data, but added that the purpose of the study was merely to examine statistical differences in pericardial thickness without attributing observed differences to any specific cause. Finally, after hearing that a considerable portion of the Vieques fishermen might actually come from just two large families, an ATSDR scientist asked the panelists to comment on the possibility of pericardial thickening having genetic causes. Other than one participant's anecdotal account of a family with several members having constrictive pericarditis, no panelists or participants commented on this issue.

Table 3-1 Data on Response Rates and Collection of Echocardiographic Images

Notes: Data presented by Dr. Ríos:
(*) The number of eligible participants in Vieques is lower than the sampling frame because several individuals in the Vieques fishing registry had either died or moved off the island.
(†) Response rate here is defined as the fraction of eligible participants who agreed to participate in the Vieques Heart Study.
(‡) For this display, a "valid image" is considered an echocardiographic image with a corresponding electrocardiogram. Both are needed to ensure that readers measure pericardial thickness at the same point in the cardiac cycle. Note, however, that readers from both PSM and the Mayo Clinic rejected some of these "valid images" for other reasons.

4.0 COMMENTS ON ECHOCARDIOGRAPHIC MEASUREMENTS

The panelists and participants discussed Question Area 2 - echocardiographic measurements - at length, not only during the meeting time designated for Question Area 2, but also during discussions on the other question areas. This section reviews all of these discussions, regardless of when they took place. For reference, the exact text of Question Area 2 reads:

"Please comment on the echocardiographic measurements made, addressing (if relevant) the following issues:

- Sensitivity, specificity, and reproducibility
- Mechanical (machine-based) and human variability
- Blinding
- Interobserver and intraobserver variability of measurements
- Clinical significance of abnormalities noted"

The panelists and participants discussed these issues extensively. The following five subsections summarize the discussions on four specific categories: collection of echocardiographic images (Section 4.1), reading the images (Section 4.2), limitations of echocardiography (Section 4.3), differences between the PSM and Mayo Clinic readings (Section 4.4), and clinical significance of the reported findings (Section 4.5).

4.1 Comments on How Images Were Collected

The PSM investigators provided background information on the procedures followed to collect the echocardiographic images. One trained, experienced technician collected all images, using one echocardiogram machine, and never adjusting the transducer frequency or the compression settings. The technician did, however, adjust the TGC differently for the individual subjects, and these variable settings were not recorded. (Panelists noted that it was also necessary to change the overall gain between examinations because of between-subject differences in chest wall characteristics.) The images were collected at two locations - one on Vieques and one in Ponce - but in examination rooms that had similar set-ups. No one supervised the performance of the technician during image acquisition.

The panelists' comments on the echocardiographic examinations addressed the following two principal issues:

• Potential biases associated with performing echocardiographic examinations. The panelists discussed several potential biases that might be associated with the procedures the PSM investigators followed to collect echocardiographic images. The fact that the technician who collected the images was not blinded to the study and control groups was of particular concern. Specifically, a panelist indicated that the technician might have been more likely to follow slightly different procedures when conducting examinations at Vieques as opposed to at Ponce, especially considering the publicity surrounding the Navy bombing and the preliminary data suggesting pericardial thickening among Vieques residents. In short, the panelist was concerned that a slight, systematic bias might have affected image acquisition because the technician was not blinded to the two study populations.

• A PSM investigator doubted that such a bias could have occurred given the tight schedule followed for acquiring images. The technician, the investigator suspected, would not have had enough time to adjust settings selectively and complete all echocardiographic examinations within the allotted time. Moreover, he felt that that the study would likely have found a much greater magnitude of pericardial thickening if such a systematic bias truly occurred.

  A panelist disagreed with this argument, citing examples of how well-intentioned field workers have unintentionally introduced systematic biases into past studies. Because the differences in pericardial thickness observed by the PSM investigators was relatively small, this panelist wondered if a slight, unintentional bias might account for the main result. A cardiologist on the panel agreed that use of inconsistent practices when conducting echocardiographic examinations can result in slight changes in results, such as the 0.15 mm difference in pericardial thickness reported in the Vieques Heart Study. These panelists acknowledged, however, that systematic biases introduced by the technician, if any, are impossible to quantify from the existing data.

  To support his concern about the practices followed when collecting images, a panelist referred to the data on the percentage of echocardiographic images rejected because they did not include a corresponding EKG (see Table 3-1). The data indicate that 19% of the echocardiographic images taken at Vieques were rejected for this reason, as compared to only 2% of the readings from Ponce. In other words, these rejection rate data do not appear to be randomly split between the groups, as one would expect if the technician consistently applied uniform techniques when acquiring images.

• Concerns about inconsistent use of TGC settings. A specific concern about the technician's practices was the apparent use of different TGC settings from one subject to the next. A panelist indicated that TGC settings affect the intensity of echocardiographic signals, and increasing the setting generally makes cardiac structures appear thicker than they actually are. Another panelist and a participant agreed, citing their experiences of how altering TGC settings can change the apparent dimensions of materials, whether heart structures or test films over lucite blocks. A participant noted that an article in the scientific literature also documents how gain settings on echocardiograms can blur output signals and therefore artificially increase the apparent thickness of heart structures.6 On the other hand, a consultant to the PSM investigators commented that TGC settings actually have little impact on measuring pericardial thickness, because slight changes to the settings cause the pericardium either to vanish from an image altogether or to expand to unrealistically large dimensions - both outcomes, he argued, would be easily identifiable as artifacts by those reading the images.

Following this discussion, some panelists raised concerns that subtle changes in the technician's procedures for acquiring echocardiographic images might account for the very small between-group differences in pericardial thicknesses reported in the Vieques Heart Study. A participant indicated that the only way to verify this concern is to have another research team collect echocardiographic images from the same individuals considered in the Vieques Heart Study and compare the results to those generated by the PSM investigators. As Section 6 notes, however, most panelists eventually agreed that follow-up studies of pericardial thickening are not needed.

4.2 Comments on How Images Were Read

As Section 2 notes, the PSM investigators read all echocardiographic images collected during the Vieques Heart Study. At the request of ATSDR and PSM, investigators from the Mayo Clinic then performed independent readings of the same data. Both teams of investigators measured pericardial thickness at end-diastole, and both teams measured several additional parameters (e.g., left ventricular volume) from the echocardiographic images. Despite these similarities, the main finding from the PSM investigators - that pericardia among Vieques fishermen, on average, were thicker than those among Ponce fishermen - was not reproduced when the measurements made by the Mayo Clinic investigators were used in the analysis.

The panelists considered how practices used to read the echocardiographic images might have led to a measurement bias, and hence the differing conclusions. They highlighted three key points:

• Inherent difficulties reading pericardial thickness from M-mode images. A participant and two cardiologists on the panel listed several difficulties associated with using M-mode echocardiographic images to measure heart structures on sub-millimeter scales. The participant, for example, displayed a typical M-mode image from the Vieques Heart Study and indicated how the signal for the pericardial thickness varied, even over the range of the cardiac cycle that can be considered end diastole.This variation, he argued, can lead to two sonographers reporting different pericardial thicknesses for the same echocardiographic image. A panelist agreed, adding that even the orientation of the viewer with respect to the screen can cause notable differences in echocardiographic data. Specifically, an individual looking down on a computer screen would probably read a greater pericardial thickness than one sitting in front of the computer screen. Finally, a panelist stated that echocardiographic images often have discontinuous lines and other artifacts in the M-mode signal, thus complicating efforts to discern "noise" from where a heart structure actually begins or ends. In fact, for this very reason inexperienced sonographers commonly overestimate the thickness of heart structures (e.g., the left ventricular wall). These panelists wondered if such subtle differences in techniques for reading the echocardiographic images could account for the differences observed between the PSM readings and the Mayo Clinic readings, as Section 4.4 discusses in greater detail.
• Use of different approaches to make the measurements. During this discussion, the panelists commented on the slightly different approaches used by the PSM and Mayo Clinic investigators to read the echocardiographic images. At PSM, a group of physicians came to a consensus reading on every image. At the Mayo Clinic, on the other hand, the images were assigned to two experienced sonographers and each image had a designated primary sonographer. The reading from the primary sonographer was considered "final," except in cases where the two readings had noteable discrepancies. In these cases, which were few in number, the reading from an independent observer (Dr. Oh) was used.

• Two panelists weighed the advantages and disadvantages of these approaches, but eventually agreed that both approaches are reasonably valid. One of these panelists did not think the differences between the PSM and Mayo Clinic readings could be explained by the approaches used by the two teams; a more likely explanation of the results is that the differences observed between the PSM and Mayo Clinic measurements fall within the commonly accepted resolution of echocardiography. Section 4.3 revisits this issue.

• Other concerns. The panelists raised two additional concerns when discussing how the two teams of investigators read the echocardiographic images. First, referring to the data in Table 2-1, one panelist noted that the PSM investigators successfully read 84 images, while the Mayo Clinic investigators successfully read only 69. The Mayo Clinic principal investigator explained that his sonographers rejected numerous images due to poor quality. The panelist wondered if the fact that the two teams of investigators considered different sample subsets might explain the discrepancy between their findings. However, data presented on paired comparisons (see Section 4.4) indicated that the difference between the PSM and Mayo Clinic findings remains and is of approximately the same magnitude, even if one considers only the subset of echocardiographic images successfully read by both groups. Second, one panelist noted that the Mayo Clinic sonographers, who reportedly knew that a "normal" pericardial thickness was less than 1.0 mm, might have been biased to read thickness in this range. This panelist acknowledged that this type of measurement bias, if it existed, would have applied to all readings, and not just those of the Vieques or Ponce fishermen. No other panelists or participants commented on this issue.

In summary, the panelists listed several inherent difficulties associated with measuring pericardial thickness from M-mode echocardiographic images. This discussion generated more detailed discussion on inherent limitations of echocardiography (see Section 4.3) as well as on the exact nature of the differences between the PSM and Mayo Clinic measurements (see Section 4.4).

4.3 Comments on Limitations of Echocardiography (Resolution)

Much of the panelists' discussion addressed the inherent limitations of echocardiography and whether this technique is capable of resolving differences in pericardial thickness on the order of 0.15 mm. The cardiologists on the panel unanimously agreed that echocardiography is a powerful and well-established tool for many applications, particularly for characterizing hemodynamics and the dimensions of certain heart structures. Several panelists and participants doubted, however, that echocardiography can reasonably and consistently measure sub-millimeter differences in a relatively thin cardiac structure such as the pericardium. These panelists acknowledged that echocardiography can detect gross differences in pericardial thickness, like those reported in the pilot study3 that led to the Vieques Heart Study. Accordingly, most thought that use of echocardiography in the Vieques Heart Study was appropriate.

The panelists made numerous notable comments on the inherent limitations of echocardiography, which are grouped here into the following three topics:

Resolution of echocardiographic measurements. Many panelists thought the commonly accepted resolution of echocardiography is a critical issue when one interprets the reported 0.15 mm difference, on average, between pericardia among Vieques fishermen and those among Ponce fishermen. A CDC scientist explained that resolution is an intrinsic property of an instrument; echocardiography has a widely accepted resolution of 1 millimeter, so it cannot distinguish two points separated by this distance. A panelist added that echocardiography, in theory, cannot distinguish two films or layers less than the instrument's resolution. A PSM investigator questioned, however, if cardiologists agree that the minimum resolution of echocardiography is truly one millimeter; that is, most cardiologists could use echocardiography to differentiate pericardia 2 mm thick from those 3 mm thick. The panelists did not respond to this comment. Regardless of the actual resolution of transthoracic echocardiography, however, several cardiologists on the panel agreed that this type of echocardiography is not capable of detecting a 0.15 mm difference in the thickness of the pericardium.

Another panelist added that devices known to have a resolution of approximately 1 millimeter do not necessarily yield measurements reproducible at that level. In the case of echocardiography, for instance, the resolution is strictly a property of the instrument. Numerous human factors, however, can cause echocardiographic measurements to have poor reproducibility, even when measuring structures thicker than 1.0 mm. A participant agreed, referring back to the panelists' previous discussions on the inherent difficulties - and potential biases - associated with both acquiring the echocardiographic images in the field and reading the acquired images in the laboratory.

Relative performance of other imaging techniques. After reviewing the inherent limitations of echocardiography for measuring pericardial thickness, the panelists briefly discussed the utility of other imaging techniques, particularly computed tomography (CT) and magnetic resonance (MR) imaging. A panelist noted that studies have shown that echocardiographic measures routinely overstate dimensions of heart structures, in comparison to actual anatomic dimensions measured during autopsy. A participant agreed and added that the variability associated with measuring other heart structures, such as left ventricular volume, is considerably greater for echocardiographic measurements than for either CT or MR imaging measurements. Based on these observations, the participant questioned the validity of using transthoracic echocardiography for deriving valid, reproducible measurements of pericardial thickness. This comment led to a debate on whether a "gold standard" exists for this measurement, as summarized below.

The "gold standard" for measuring pericardial thickness. After debating the benefits of echocardiographic, MR, and CT imaging, the meeting co-chairs asked the panelists to comment on whether a gold standard exists for measuring pericardial thickness. The panelists agreed that no one device has been widely used for measuring pericardial thickness, largely because this type of measurement is not routinely made in echocardiography when one evaluates cardiac function. Rather, echocardiographers measure hemodynamic and other parameters, which they then use to characterize systolic and diastolic function.

So although no real gold standard exists, said several panelists and participants, both CT and MR imaging are likely the closest to a gold standard, given these techniques' superior capability to measure heart structures. A CDC scientist explained that the superior resolution of CT and MR imaging suggests their measurements are more likely to be reproducible if the same population is studied twice, as compared to the relatively poor reproducibility observed in the echocardiographic measurements made during the Vieques Heart Study (see Section 4.4). A panelist agreed and added that echocardiographic readings can be easily biased, even unintentionally, by field collection practices as well as by approaches to reading images. This panelist noted that CT and MR imaging produce images of almost the entire pericardium, leading to much more standardized readings. Finally, noting the advantages of CT and MR imaging, a panelist and a participant said they would recommend use of one of these techniques if the Vieques Heart Study were to be conducted again. As Section 6 indicates, however, most panelists did not think the issue of potential pericardial thickening in Vieques fishermen warranted further study.

Based on these discussions, the panelists and most participants agreed that echocardiography is a "sub-optimal tool" for measuring pericardial thickness at the dimensions of interest in the Vieques Heart Study. The fact that two experienced laboratories read the same set of images but obtained notably different results confirms this hypothesis. The next section addresses in greater detail the reproducibility of measurements.

4.4 Comments on Differences Between the PSM and Mayo Clinic Data

After discussing their general concerns regarding echocardiographic measurements and potential biases associated with how images are collected and read, the panelists commented more specifically on differences between the PSM and Mayo Clinic readings of the Vieques Heart Study echocardiographic images. This discussion opened with the PSM investigators comparing their results to the Mayo Clinic results for the 66 subjects for whom both research groups successfully read echocardiograms (see Table 4-1).

The results (see Table 4-1) continue to show the same basic trend of the Vieques Heart Study results (see Table 2-1), even when considering this subset of subjects. Specifically, the data generated at PSM for these 66 subjects suggest that the pericardia of Vieques fishermen are, on average, 0.12 mm thicker than those of Ponce fishermen - a statistically significant difference (P value = 0.03). On the other hand, the Mayo Clinic data for the same subjects suggest that the pericardia of Ponce fishermen are, on average, thicker than those of Vieques fishermen, but this difference is not statistically significant (P value = 0.20). Noting the similarity between the data for the 66 subjects successfully read by both groups and the data for the overall pool of subjects (see Table 2-1), an ATSDR scientist commented that the difference between the PSM data and the Mayo Clinic data cannot be explained by the fact that the two groups rejected different numbers of echocardiographic images. Several panelists agreed.

After reviewing these data, the panelists debated the true nature of the differences between the PSM and Mayo Clinic measurements. The discussions focused on three topics:

Magnitude of differences between the PSM and Mayo Clinic data. Referring to the data in Table 4-1, a panelist noted that the differences between the PSM and Mayo Clinic data are relatively small. Specifically, for Vieques fishermen, the PSM and Mayo Clinic average readings differ by 0.38 mm; for Ponce fishermen, this difference is 0.21 mm. This panelist and a participant were encouraged by the fact that these differences fall within the generally accepted resolution of transthoracic echocardiography. The panelist added that the relatively high inter- and intra-observer variability observed by the Mayo Clinic suggests that the measurements were conducted in a range lower than can be resolved using echocardiography. Or, stated differently, the dimensions are so small that one would expect to see highly variable data due to random measurement errors (i.e., "noise"). The panelists revisited this issue when discussing the correlations between the PSM and Mayo Clinic data (see below).

When discussing the differences observed in pericardial thickness, a participant and some panelists noted that the PSM and Mayo Clinic measurements for all parameters other than pericardial thickness appeared to be in excellent agreement. They were encouraged by this similarity and suspected that the differences observed in measurements of pericardial thickness probably resulted from attempting to discern differences of magnitudes lower than commonly accepted echocardiography resolutions.

Inconsisten