«EPIDEMIOLOGY AND BIOSTASTICS: AN INTRODUCTION TO CLINICAL RESEARCH Bryan Kestenbaum, MD MS University of Washington, Seattle TABLE OF CONTENTS. ...»
Despite limitations of case series data, they may be highly suggestive of an important new association, disease process, or unintended side effect of a medication or treatment.
In 2007, a case series described three cases of male prepubertal gynecomastia.13 The report included detailed information on each subjects’ age, body size, serum levels of endogenous steroids, and known exposures to exogenous hormones. It was discovered that all three otherwise healthy boys had been exposed to some product containing lavender oil (lotion, shampoo, soap), and that in each case the gynecomastia resolved upon discontinuation of the product. Subsequent in vitro studies demonstrated endocrine-disrupting activity of lavender oil.
This novel case series data may lead to further investigations to determine whether lavender oil, a common ingredient in commercially-available products, may be a cause of gynecomastia.
A vaccine designed to prevent rotavirus infection was found to cause weakening of the intestinal muscle layers in animals. Following release of the vaccine, a number of cases of intussusception (when one portion of the bowel slides into the next) were reported in children who received the vaccine, with some fatal cases.14 The strong biologic plausibility underlying this initial association, and knowledge that intussusception is otherwise rare in infants, was highly suggestive of a causal relationship and the vaccine was removed from the market.
1. Cross-sectional studies measure the exposure and the outcome at the same time.
2. Cross-sectional studies estimate the prevalence of a disease or condition.
3. Cross-sectional studies cannot establish a temporal relationship between the exposure and the outcome.
A cross-sectional study refers to a study design in which ascertainment of the exposure and the outcome occurs simultaneously. Measuring the exposure and outcome at the same time implies that there is no follow-up time in a cross-sectional study.
Homocysteine, an amino acid formed during the conversion of methionine to cysteine, possesses pro-inflammatory and pro-thrombotic properties that might contribute to atherosclerosis. Researchers investigated whether higher serum homocysteine levels were associated with peripheral arterial disease among 6600 men and women. Findings are presented in Table 4.1.
Association of homocysteine level and peripheral arterial disease.
How can we interpret these data? Among the 6600 study participants, 1000 had homocysteine levels that were classified as high and 5600 had levels that were classified as normal. We can calculate the prevalence of peripheral arterial disease among study participants who had high
Prevalence = 100 people with disease / 1000 people with high homocysteine *100% = 10%.
Similarly, we can calculate the prevalence of peripheral arterial disease among study participants
who had normal homocysteine levels:
Prevalence = 175 people with disease / 5600 people with normal homocysteine *100% = 3%.
The cross-sectional study design results in knowing the amount of peripheral arterial disease that is present at the time of homocysteine measurement. As defined in chapter 1, prevalence is used to describe the amount of disease in a population at any given point in time. This contrasts with incidence, which is used to describe the number of new cases of disease that develop over time.
The incidence of peripheral arterial disease cannot be determined in this cross-sectional study.
An important disadvantage of cross-sectional studies is the inability to discern a temporal relationship between the exposure and the outcome. Two alternate explanations for the homocysteine data are possible. On one hand, it is possible that higher homocysteine levels precede and participate in the development of peripheral arterial disease. On the other hand, it is also possible that peripheral arterial disease leads to metabolic changes that include a subsequent rise in homocysteine levels. The cross-sectional study design cannot distinguish between these two possibilities, because homocysteine levels and peripheral arterial disease were measured simultaneously. As a result, we are left uncertain as to whether higher homocysteine levels might cause peripheral arterial disease, or whether peripheral arterial disease might cause high homocysteine levels.
In some instances, the inability to discern temporality in a cross-sectional study is not of concern because only one direction of causality is biologically plausible. For example, consider a second study that simultaneously measures peripheral arterial disease status and polymorphisms (different sequences of nucleic acids) within the methylenetetrahydrofolate reductase (MTHFR) gene, which codes for an enzyme involved in homocysteine metabolism. The genetic study finds that a particular polymorphism within the MTHFR gene is associated with a two-fold greater risk of peripheral arterial disease. In this example, one direction of causality, that peripheral arterial disease causes the MTHFR gene polymorphisms, is biologically impossible. We are left to conclude that differences in the MTHFR gene precede, and may cause peripheral arterial disease.
Cross-sectional studies are frequently performed while researchers are waiting for follow-up data to become available. For example, after conflicting interpretations of animal studies concerning the safety of the estrogenic chemical BPA in food packaging materials, the first large-scale epidemiologic study of BPA in humans was published.15 The investigators used data from the National Health and Nutrition Examination Study (NHANES), a cross-sectional study that randomly sampled adults from the U.S. population for interviews and laboratory measures.
Participants provided a urine sample, which was analyzed for BPA concentration, and at the same time reported their history of a number of physician-diagnosed diseases. Investigators found that participants with a self-reported history of cardiovascular disease or diabetes had higher mean urinary BPA levels compared to participants without these conditions.
These cross-sectional data indicate a greater prevalence of cardiovascular disease and diabetes among people with greater BPA exposure, as indicated by their urinary levels of this compound.
It is tempting to conclude from these data that BPA might be a cause of these diseases. However, we should remain alert to the alternate possibility that cardiovascular disease and diabetes might influence dietary habits, which subsequently alter the amount of BPA exposure. This alternative explanation weakens the case for BPA exposure as a novel cause of disease. Ideally, this provocative cross-sectional study will be followed by more formal studies that measure BPA levels in healthy people without diabetes or cardiovascular disease, and then follow them for the development of incident diabetes and cardiovascular disease during long-term follow-up.
However, it may take years or decades to collect these follow-up data.
1. The fundamentals of a cohort study design are:
a. Identify people who are free of disease at the beginning of the study b. Assemble cohorts of exposed and unexposed individuals c. Follow cohorts for the development of incident outcomes d. Compare the risks of incident outcomes in each cohort
2. Cohort studies have certain advantages:
a. Can discern temporal relationships between the exposure and outcome b. Can be used to evaluate multiple outcomes
3. Cohort studies have certain disadvantages:
a. Observational design: other factors may be responsible for observed association b. May be inefficient for studying rare diseases or those with long latency periods
4. Cohort studies can be used to evaluate the risks and benefits of medication use.
5. Cohort studies can be used to calculate relative and attributable risks of disease.
I. OVERVIEW OF COHORT STUDY DESIGNCohort studies are a particular type of observational study design that improve upon case series / case reports and cross-sectional studies. Cohort studies provide an estimate of the incidence of disease or outcome, typically include a formal comparison group, and temporally dissociate the exposure from the outcome as a means to strengthen the evidence for causation.
Cohort studies are conducted in three fundamental steps:
1. Assemble or identify cohorts of exposed and unexposed individuals who are free of the disease/outcome of interest at the beginning of the study
2. Observe each cohort over time for the development of the outcome(s) of interest
3. Compare the risks of outcomes between the cohorts Recall from chapter 2 that the term ‘exposure’ is used to describe any factor or characteristic that may explain or predict the presence of an outcome. Examples of exposures include serum levels of cholesterol, the use of a diuretic medication, smoking, and kidney function.
The first step in a cohort study is to identify cohorts of exposed and unexposed individuals. A cohort is a group of people, derived from the study population, who share a common experience or condition, and whose outcome is unknown at the beginning of the study. In cohort studies, the investigators observe exposure status, which occurs “naturally.” In contrast, in randomized trials investigators assign the exposure to the study participants.
Investigators wish to evaluate whether smoking causes premature failure of a kidney transplant. They recruit 200 patients with a functioning kidney transplant who are receiving care at a single transplant clinic. The investigators use a questionnaire to ascertain smoking status for all study subjects, and then divide the study population into a cohort of smokers, and another cohort of non-smokers. They observe each cohort over time and compare the risks of incident kidney transplant failure between the cohorts.
Investigators wish to ascertain whether a new antibiotic, supramycin, is associated with the development of a skin rash. They identify a group of patients who are prescribed antibiotic therapy for community-acquired pneumonia. Some patients are prescribed the new antibiotic supramycin, whereas others receive a different antibiotic. The investigators divide the study population into a cohort of supramycin users and a cohort of supramycin non-users, and then compare the incidence of skin rash between the cohorts.
Cohort studies typically focus on incident, or new cases of disease that occur during follow-up.
To accomplish this goal, investigators typically exclude individuals who have prevalent disease at the beginning of a cohort study. The evaluation of incident disease outcomes helps to establish that the exposure of interest preceded the outcome, and therefore might represent a cause of the disease. For the smoking and kidney transplant example, investigators would likely measure kidney function in all subjects at the beginning of the study, and then exclude those who have evidence of transplant failure. For the supramycin and rash example, investigators may require a baseline skin examination to exclude subjects who have a rash at the beginning of the study.
In cohort studies, researchers observe the exposure of interest. As a result, exposed and unexposed individuals may differ by characteristics other than the exposure. For example, exposed subjects in the supramycin study (supramycin users) may differ from unexposed subjects (non-users) by characteristics that influenced the decision to prescribe supramycin, such as the severity of pneumonia, practice patterns of the prescribing physician, and health insurance status.
If these potential differences also influenced the risk of developing a rash, they could distort the study findings. In other words, a greater risk of drug rash among supramycin users might be caused by supramycin itself, or might be due to other characteristics that are linked with supramycin use, obscuring causal inference. The concept that factors other than the exposure may influence the study results is called confounding. Confounding is a major limitation of observational studies and is discussed in detail in chapters 9-10.
II. ASCERTAINMENT OF STUDY DATAOnce investigators decide on the specific exposures and outcomes to study, they should attempt to measure these characteristics using the most accurate methods available within their resources.
For the smoking and kidney transplant failure study, possible choices to ascertain kidney transplant failure include the use of medical records from the transplant clinic, serum and urine markers of kidney function, a kidney transplant biopsy, and/or data from a national registry that tracks kidney transplant failures. Investigators must consider which of these sources provide the most valid measurements, and whether these data are uniformly available for all study participants. Kidney biopsy data may be the most accurate method for detecting kidney transplant failure; however, kidney biopsy data may be available for only a small number of study participants. For the supramycin and drug rash example, possible methods to ascertain supramycin use include the use of questionnaires, review of medical charts, and/or query of an electronic pharmacy database, if such data are readily available.
Important considerations in measuring study data are the validity of the measurements, timing of the measurements, and availability of uniform measurements among the study population.
A. Validity of measurements The validity of a measurement refers to how closely the measured data represent the true data.