Understanding epidemiology and its use in drug and medical device litigation.

• Author: Parker, Bruce R.

Defense counsel must be prepared to combat expert testimony that is inconsistent with the sound use of epidemiological data

OVER the last 20 years, there has been an explosion in toxic tort industry-wide litigation. Drugs and medical devices have not been exempt from the epidemic of lawsuits. Notable examples include Bendectin, DES, L-Tryptophan, the Dalkon shield, tampons and silicone breast implants. Among the explanations for the explosion is the deference courts have given to plaintiffs' experts who have been permitted to give causation opinions based on nothing more than personal speculation dressed up to sound like scientific evidence. Few decisions before Daubert v. Merrell Dow Pharmaceuticals Inc.(1) recognized the critical role that epidemiology plays in the scientific method for ascertaining causation.

Daubert has had many effects, not the least of which is to compel trial courts to examine critically the validity of the scientific methodology that purportedly supports an expert's testimony. This means that defense counsel need to understand the role epidemiology plays in the scientific search for the cause of disease and how to differentiate good from weak epidemiological evidence.

The goal of this article is to provide an overview of basic epidemiological principles and to discuss how such evidence is used and misused in trials. It is not an exhaustive treatise on epidemiological study designs and statistical analysis. However, it addresses the major issues raised in defending a drug or medical device manufacturer against claims that the product caused systemic disease. The article contains a general discussion on what epidemiology is and the role it plays in the scientific investigation of the cause of disease, discusses how an epidemiological study is performed and the differences among epidemiological studies. addresses statistical issues raised by epidemiological studies and questions to consider when interpreting the reliability and validity of epidemiological data, discusses how and under what circumstances epidemiological data can be used to reach conclusions regarding causation, considers how courts have used epidemiology, and offers practice pointers on cross examining plaintiffs' epidemiologist and presenting epidemiological testimony in the defense case.

WHAT IS EPIDEMIOLOGY?

Epidemiology can be broadly defined as "the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to control of health problems."(2) The objectives of epidemiology are to: (1) identify what factors increase the risk of disease, (2) determine how many new cases of a disease develop in a defined period (incidence) and how many total cases of the disease exist at specified points in time (prevalence): (3) obtain an understanding of how a disease develops in terms of time and prognosis: (4) evaluate the efficacy of new therapies, such as surgery and drugs;(3) and (5) assist in formulating regulatory decisions regarding environmental issues that affect the rate at which diseases develop and persist.(4)

All epidemiological studies, regardless of whether their goal is to study risk factors of disease or efficacy of treatment modalities, share a common feature. What distinguishes an epidemiological study from a case report or case series is the inclusion of a control group that permits the investigator (assuming the study is done correctly) to draw reasonable conclusions with respect to the effect, if any, of the exposure or treatment being studied. Proper scientific methodology, as required by Daubert, does not permit conclusions to be drawn regarding a causal relationship between exposure and disease or the efficacy of treatment based on case reports or case series.(5)

A number of examples can be used to demonstrate this point. For example, many case series of patients with low back pain who had undergone MRI concluded that the majority of the patients with back pain had disc abnormalities, leading some practitioners to conclude that disc abnormalities were a cause of low back pain. However, later studies using a controlled group of patients who did not have back pain revealed similar rates of disc abnormalities. Such evidence does not support a conclusion that disc abnormalities as revealed by MRI are the cause of low back pain.(6)

Unlike animal studies and in vitro cell testing, an epidemiological study is conducted on human populations and is therefore uniquely capable of quantifying the exposure-disease relationship.(7) Epidemiology is not saddled with the difficulties of extrapolating the data to human beings, as are the other forms of scientific evidence. For example, the best animal experiment is always subject to the question of its applicability to human beings. Similarly, the results of laboratory (in vitro) studies on human cells are not necessarily good predictors of what happens in the body since the laboratory environment is often dissimilar to the in vivo condition.

For these reasons, as discussed later, courts are increasingly viewing epidemiological evidence as the most probative scientific evidence of causation.(8) Although a well-designed, properly performed epidemiological study provides the most probative scientific evidence on causation, there are limitations to the results. As explained later, a study may reveal a statistically significant association between exposure and disease. Even a strong statistical association between exposure and disease, however, does not in and of itself prove that the exposure is capable of causing disease generally. It follows that epidemiological data of a strong association between exposure and disease provides no evidence that in a particular individual the exposure has caused the plaintiffs injury. Finally, although most epidemiological studies generate data that reflect the degree, if any, of the statistical association between a suspected risk factor and disease, epidemiological studies are equally capable of revealing whether a risk factor has the opposite effect of providing a protective mechanism against the occurrence of disease.

HOW ARE EPIDEMIOLOGICAL STUDIES DONE?

1. Observation Versus Experimental Studies

   A controlled epidemiological study can be either experimental or observational. The primary difference between the two is the role of the investigator. In an experimental study, which is also called a clinical study, the investigator randomly and blindly selects which of the subjects will be exposed to the agent being studied. In well-conducted experimental studies, the subjects are not told ("blind") whether they have been exposed (e.g., been given the new drug or placebo). Because it would be unethical for investigators to expose people to substances known to be harmful, experimental epidemiology is limited to evaluating new medical treatments that are thought to be beneficial. Because people are randomly exposed in a clinical study, such a study provides the best scientific data on the association, if any, between exposure and the outcome being studied ("endpoint").

   As explained later, an association revealed in an epidemiological study may be the result of random chance. If study subjects are exposed by random selection, then it is less likely that chance has produced the results.(9)

   In contrast, in an observational study the investigator does not select which subjects are to be exposed. Rather, the investigator selects subjects who have already been exposed and observes which of the exposed and controls develops an illness.

2. Types of Observational Studies

   Observational studies are either cohort, case-controlled or nested case-controlled in their study design. The variables that make one person different from another, such as age. genetics, lifestyle, race, sex and exposure duration cannot be controlled directly by investigators in an observational study. Consequently, the effect, if any, of such variables on the rate of disease are controlled in designing a study and by using appropriate statistical tests to analyze the data.(10) The quality of epidemiological data is directly correlated to the design of the study. Studies that are better designed to control for variables that can affect the outcome of the study produce more reliable data on whether a true association exist between exposure and disease.

   Whether a cohort, case-controlled or nested case-controlled type of study is performed depends on a variety of factors, including cost, availability of exposure data and whether the disease is rare or common. In general, however, all three studies are capable of generating useful epidemiological data on the relationship between exposure and disease.

   1. Cohort Studies

      A cohort a group that is traced over time. A cohort study is one in which groups of exposed and non-exposed subjects are followed over time. Observations are made as to how many subjects in both groups develop the disease of interest.

      1. Prospective/Concurrent Cohort Study

         In a prospective cohort study, investigators begin by identifying a population of people who have been exposed to the agent (such as a chemical) being studied. The control population is comprised of people who, in all important respects, are similar to the exposed population except for the fact that they have not been exposed. The investigators, either by periodic physical examination, questionnaires or other means, will follow both populations over time until the pre-determined period of study has elapsed. At the conclusion of the study, an analysis is conducted of the proportion of exposed subjects who develop one or more diseases of interest. The rate of disease in the exposed population is then compared to the rate of disease in the non-exposed population to determine if the exposure is associated with an increased rate of disease.

         The term "bias," discussed more fully later, for present purposes can be defined as any...