Risk Paper

EPIDEMIOLOGY IN HAZARD
AND RISK ASSESSMENT

A Partial Review of the "London Principles",
With Recommendations for Revisions,
By an Expert Panel

Based on a Review of Studies
of Possible Relationship Between
Induced Abortion and Breast Cancer

November 1999

Federal Focus, Inc. is a non-profit 501(c)(3) private foundation established in 1986 to engage principally in research, analytical, and educational activities pertaining to Federal government policy and technical issues, particularly ones of inter-agency concern. For the last eight years, environmental health issues, especially the development of improved Federal government risk assessment principles and guidance, have been a principal focus of the foundation.

Library of Congress Catalog Card No. 99-76949

International Standard Book No.

Copyright Federal Focus, Inc. 1999. All rights reserved, except that any person may copy the "Recommended Revisions to the Principles" and "Additional Principles for Hazard and Risk Assessment" portions of the publication which are printed on light blue pages.

Additional copies of this publication can be ordered from Federal Focus, Inc. at the address shown above.

Printed in the United States of America.


TABLE OF CONTENTS
INTRODUCTION 1
PANEL MEMBERS 6
DESCRIPTION OF THE REVIEW PROCESS 8
DISCUSSION SUMMARY 10
RECOMMENDED REVISIONS TO THE LONDON PRINCIPLES 36
I. PRINCIPLES AND QUESTIONS FOR
EVALUATING AN EPIDEMIOLOGIC
REPORT FOR CAUSE-EFFECT
RELATIONSHIP
36
II. PRINCIPLES FOR EVALUATING
A BODY OF STUDIES ("META-ANALYSIS")
46
III. A. ADDITIONAL PRINCIPLES FOR
THE USE OF EPIEMIOLOGICAL DATA
IN HEALTH HAZARD IDENTIFICATON
AND RISK ASSESSMENT
51
III. B. QUESTIONS FOR SUPERVISING
RISK ASSESSORS AND RISK
MANAGEMENT OFFICIALS TO
ASK THE EPIDEMIOLOGISTS AND
OTHER RISK ASSESSMENT TEAM
MEMBERS
53
PUBLICATIONS USED IN THE REVIEW 59


INTRODUCTION

This report is the latest step in a multi-phase project to develop generally agreed upon principles or other methods for evaluating epidemiologic data for use in human health hazard identification and risk assessment, particularly when such hazard identification or risk assessment is intended to form a basis for regulatory action or government dissemination of health information.

The first two phases resulted in the August 1996 publication by Federal Focus of PRINCIPLES FOR EVALUATING EPIDEMIOLOGIC DATA IN REGULATORY RISK ASSESSMENT. 1 Those principles, which we refer to as the "London Principles", were drafted by a panel of eighteen experts in public health and medicine in London, England in October 1995. The London effort followed a preliminary conference by a similar panel in 1994.2 The London Principles were divided into basically two parts: A set of principles for evaluating an epidemiologic report for cause-effect relationship, and a set of principles for using human and animal data in dose-response evaluation. Additionally, at the request of the London panel, Federal Focus developed and published separately at the same time a companion set of recommendations for implementing the London Principles in governmental hazard and risk assessment guidance.3


p. 1

Subsequently, Federal Focus convened a number of workshops with Federal agency and private sector scientists and risk assessors to obtain their views on the London Principles. While the principles were generally regarded as sound, many participants in those workshops, as well as members of the original London panel, urged Federal Focus to organize one or more "test runs" of the principles, using them to evaluate specific databases and thereby acquire more insight into their practical utility and any need for modifications or enhancements.

Federal Focus followed these recommendations, and began reviewing the general composition of databases for a number of potential hazards likely to be of general interest. After spending some time in this review process, and consulting with a number of interested parties, we came to two basic conclusions: First, the "test run" would be more manageable if we concentrated on only one of the two sets of principles -- the cause-effect relationship set.4 Second, the exercise would be more likely to be perceived as completely objective if we could avoid review of an exposure which was of current regulatory interest, or might be in the near future. We also had to take into consideration the need for a database including a significant number and variety of epidemiologic studies, the desirability of including in the test run one or more recent meta-analyses, and the desirability of having issues of biological plausibility and dose-response related to interpreting the epidemiologic findings. Such considerations led us to select for the test run the body of studies on the possible relationship between induced abortion and breast cancer. That database also had the following characteristics that seem to appear often in the databases for many currently controversial risk issues:


2

  • apparently inconsistent study results
  • relatively low reported "positive" risk ratios
  • potential for recall/response bias
  • possible confounding by other risk factors
  • a recent and reasonably comprehensive meta-analysis combining statistical analysis and qualitative discussion
  • possible issues regarding coherence between epidemiologic study results and biological findings

An effort was made to assemble a body of epidemiologic studies on the subject that was as comprehensive as feasible. A selection of pertinent animal and human biological studies was added for consideration of factors such as biological plausibility and dose-response. Those materials, as distributed to the panel members in full text (with the exception of translation of only pertinent portions of some foreign-language articles), are listed at the end of this publication. There are 31 epidemiologic study and review articles, and 7 articles or book chapters on animal experiments and human biology. As explained below in the "Description of the Review Process", only a subset of the epidemiologic studies were assigned for detailed review and discussion, while the others were optional.

Consistent with the recommendations of the London panel, Federal Focus recruited a review panel that was multidisciplinary. Four panel members were expert cancer epidemiologists; one was an expert in pharmacology and toxicology5 ; and one was a clinical practitioner in reproductive endocrinology, ob/gyn, and psychiatry. Only one of the epidemiologists had previously worked on the specific association being


3

reviewed. The panel members and their affiliations are listed following this Introduction.6

As described in the following section on the review process, this review panel met for a full day in Denver, Colorado in July 1998, and engaged in extensive pre-meeting and post-meeting work.

The fundamental purpose of this review exercise was to critique the London Principles. The specific review materials were considered a device to test practical application of the Principles and coax focused thinking on their accuracy, completeness, and practicality. Thus, the panel did not arrive at any overall conclusions on causal relationship for the particular database. Moreover, since only a subset of the literature was reviewed in any detail, this would not have been appropriate. Nevertheless, the discussion summary does reflect opinions by individual panel members on some of the more notable studies which individual readers might wish to consider in conducting their own review and drawing conclusions.7

This report contains numerous recommendations for revisions in the original hazard identification principles. It also recommends an almost complete revision of the principle covering meta-analysis and pooling, which was initially placed in the dose-response set of the London Principles (B-6). Finally, the principles and questions for hazard and risk assessment are completely new.


4

It may be necessary to read the Discussion Summary section to discern the rationales for the specific recommended revisions and additions.8

All involved in this phase of the project had to work within time and budget constraints, and the work presented in this report, while a best effort within those constraints, surely cannot purport to be definitive. Rather, this report, together with the previous reports, is, in our opinion, best regarded as a work in progress, on which others will doubtless wish to comment. We look forward to this. Federal Focus has previously posted the London Principles on its website (www.fedfocus.org) and will also post the recommended revisions and additional principles from this report. We will also provide a means for interested parties to comment interactively on that site.9


5

PANEL MEMBERS

Sarah L. Berga, M.D.
Dept. of Obst., Gyn. and Repro. Sciences
Magee-Women's Hospital
The University of Pittsburgh
School of Medicine
Pittsburgh, Pennsylvania
USA

Janet R. Daling, Ph.D.
Professor of Epidemiology
University of Washington, and
Program in Epidemiology
Fred Hutchinson Cancer Research Center
Seattle, Washington
USA

Christine Friedenreich, Ph.D.
National Health Research Scholar
Division of Epidemiology, Prevention
and Screening
Alberta Cancer Board, and
Adjunct Assistant Professor
Dept. of Community Health Sciences
University of Calgary
Calgary, Alberta
Canada

Sander Greenland, Dr.P.H.
Professor
Dept. of Epidemiology
UCLA School of Public Health
Topanga, California
USA


6

David A. Savitz, Ph.D.
Professor and Chair
Department of Epidemiology
School of Public Health
The University of North Carolina
Chapel Hill, North Carolina
USA

William J. Waddell, M.D.
Professor and Chair Emeritus
Department of Pharmacology & Toxicology
School of Medicine
University of Louisville
Louisville, Kentucky
USA


7

DESCRIPTION OF THE REVIEW PROCESS

Twelve published epidemiologic study reports were selected for more detailed review and discussion for the purpose of considering the utility of the London Principles and any recommendations for revisions. Those articles are referenced in the Publications section of this report. Since the goal of the exercise was not to arrive at conclusions regarding the overall database, there were not rigorous criteria for selection of these twelve studies. The studies were selected largely to provide a range of both positive and negative (or null) findings, diversity in study population origins and study design, and apparent influence on the statistical portion of the meta-analysis presented in the Brind et al. meta-analysis article.

Each of the four epidemiologists on the panel was assigned six of the twelve studies to review and comment on; and each study had two assigned reviewers/lead discussants. Assignments were made by random drawing (with the exception that no one was assigned to review a study to which they had contributed). The epidemiologists reviewed their assigned studies prior to the meeting using a copy of the London Principles and actually checking boxes where called for and noting comments and questions on the forms.

The panel met for a full day in Denver, Colorado during July 1998. The meeting was moderated by William G. Kelly, Jr. of Federal Focus. The two non-epidemiologists began the meeting with a discussion of the relevance of the experimental animal and human biological data and the biological plausibility of the hypothesis. The four epidemiologists then took the lead in the discussions. The two assigned reviewers for each study presented their comments and their experience in applying the London Principles to the study, and then there was plenary discussion, with particular emphasis on what was learned about the accuracy and utility of the London Principles and what, if anything, should be done to revise or augment the Principles.

After all of the review presentations and related discussion, there was further discussion and agreement on the intended work product and post-meeting work assignments. The four epidemiologists divided the London Principles into three sets (A-1 to A-3, A-4 to A-6, and B-6), and assigned lead responsibility for developing specific recommendations for revisions or


8

additions. The two non-epidemiologists, along with the moderator, took lead responsibility for developing the recommended additional principles for hazard and risk assessment. The moderator also prepared the discussion summary from the transcript which appears in this report and circulated it to the panel for use in the post-meeting work. All recommendations for revisions and additions were then circulated to the full panel for comment and further revision as necessary.


9


1 ISBN 0-9654148-0-9. A limited number of copies are available from Federal Focus. The Principles and Recommendations portion of the London report is also posted on the Federal Focus website, www.fedfocus.org, and Federal Focus has waived copyright to that portion of the report, as well as those portions of the current report containing the specific recommended revisions and additional principles.

2 THE ROLE OF EPIDEMIOLOGY IN REGULATORY RISK ASSESSMENT (J. Graham ed., Elsevier Science, B.V 1995, ISBN 0-444-82201-1).

3"Epidemiologic Data in Regulatory Risk Assessments: Recommendations for Implementing the London Principles and for Risk Assessment Guidance" (Federal Focus, Inc. 1996). Copies are available from Federal Focus, Inc.

4 This report is titled a "partial review" of the London Principles because it addresses mainly the cause-effect principles and questions (including Principle B-6 on meta-analysis and pooling, which was originally placed in the dose-response set), and does not address the dose-response evaluation principles. Although the title of this report refers to hazard and risk "assessment" for simplicity, there is a more technically accurate distinction between hazard "identification" and risk "assessment", since "assessment" indicates an attempt to quantify risk taking the level of exposure or dose into account, whereas hazard "identification" usually indicates a simple Yes or No determination of causal relationship without regard to level of exposure or dose.

5Dr. Waddell was unable to attend the meeting in Denver, and participated via teleconference in the initial portion of that meeting concerning biological issues. He subsequently reviewed the full transcript of the meeting and submitted additional comments to the rest of the panel. His post-meeting comments are incorporated into the discussion summary as if he had been present. He also contributed to drafting of the "Additional Principles for Hazard and Risk Assessment".

6One panel member, Dr. Friedenreich, had participated in the Federal Focus 1994 Lansdowne conference, which laid the groundwork for the London conference. Dr. Friedenreich also contributed a paper to the publication resulting from that conference, THE ROLE OF EPIDEMIOLOGY IN REGULATORY RISK ASSESSMENT, supra: Friedenreich C, "An evaluation of biases introduced by confounding and imperfect retrospective and prospective exposure assessments", at pp. 43-51. The other three epidemiologists on the Denver panel had been quoted or referenced in the Science article by G. Taubes which was reprinted as Appendix G of the London report.

7To facilitate candor, it was agreed beforehand that there would be no attribution of views.

8The Discussion Summary was prepared by the moderator from a full transcript of the meeting, and underwent limited review and editing by the panel members.

9This project was financed solely by Federal Focus, Inc., and was not supported by any grants or specific contributions.


CONTINUE