Cross-sectional and ecological studies and agent-based modeling are three other types of observational research. Cross-sectional stud- ies examine exposure prevalence in relation to disease prevalence in a defined population at a single point in time. Ecological studies examine disease rates in relation to a population-level factor. Both types of designs have important limitations absent from the other observational studies.
An unclear temporal relationship between exposure and disease arises in cross-sectional studies of changeable exposures. Problems making cross-level inferences from the group to the individual (known as the ecological fallacy) occur in ecological studies. Agent-based modeling is not a study design but rather a method of analysis that uses computer simulations to study complex interactions between individuals, their environment, and time.
References
1. MacMahon B, Trichopoulos D. Epidemiology Princi
ples and Methods. 2nd ed. Boston, MA: Little, Brown and Company; 1996.
2. Porta, M. A Dictionary of Epidemiology. 6th ed. New York, NY: Oxford University Press; 2014.
3. Heymann, DL. Control of Communicable Diseases Manual. 20th ed. Washington, DC: American Public Health Association; 2015.
4. Pickett, JP (exec. ed.). The American Heritage Dictio
nary of the English Language. 4th ed. Boston, MA:
Houghton Mifflin; 2000.
5. Colton, T. Statistics in Medicine. Boston, MA: Little, Brown and Company; 1974.
6. Meinert, CL. Clinical Trials Dictionary: Terminology and Usage Recommendations. Baltimore, MD: The Johns Hopkins Center for Clinical Trials; 1996.
7. Friedman LM, Furberg CD, Demets DL. Funda
mentals of Clinical Trials. 2nd ed. Littleton, MA: PSG Publishing Co.; 1985.
8. Newell, DJ. Intention-to-treat analysis: implications for quantitative and qualitative research. Int J Epide
miol. 1992;21:837-841.
9. Hunt JR, White E. Retaining and tracking cohort study members. Epidemiol Rev. 1998;20:57-70.
10. Miettinen OS. The “case-control” study: valid selection of study subjects. J Chron Dis. 1985;38:543-548.
11. McLaughlin JK, Blot WJ, Mehl ES, Mandel JS. Prob- lems in the use of dead controls in case–control studies.
I. general results. Am J Epidemiol. 1985;121:131-139.
12. Correa A, Stewart WF, Yeh HC, Santos-Burgoa C.
Exposure measurement in case-control studies:
reported methods and recommendations. Epidemiol Rev. 1994;16:18-32.
13. Maclure, M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol. 1991;133:144-153.
14. Mittleman MA, Maclure M, Robins JM. Control sampling strategies for case-crossover studies: an assessment of relative efficiency. Am J Epidemiol.
1995;142:91-98.
15. Susser M. Editorial: the tribulations of trials—
interventions in communities. Am J Public Health.
1995;85:156-158.
16. Rothman KJ. Induction and latent periods. Am J Epidemiol. 1981;114:253-259.
17. Austin H, Hill HA, Flanders WD, Greenberg RS.
Limitations in the application of case-control methodology. Epidemiol Rev. 1994;16:65-76.
18. Key statistics from the National Survey of Family Growth. cdc.gov. https://www.cdc.gov/nchs/nsfg /key_statistics.htm. Accessed October 2016.
19. Monson RR. Occupational Epidemiology. 2nd ed.
Boca Raton, FL: CRC Press; 1990.
20. Burdorf A, Naaktgeboren B, De Groot H. Occupa- tional risk factors for low back pain among sedentary workers. J Occup Med. 1993;35:1213-1220.
21. Hu H, Shih R, Rothenberg S, Swartz BS. The epidemiology of lead toxicity in adults: measuring dose and consideration of other methodologic issues. Environ Health Perspect. 2007;115:455-462.
22. Morgenstern H. Ecologic studies. In: Rothman KJ, Greenland S, Lash TL (eds.). Modern Epidemiology.
3rd ed. Philadelphia, PA: Wolters Kluwer Health/
Lippincott Williams and Wilkins; 2008.
23. Zhang J, Zhao Z, Berkel HJ. Egg consumption and mortality from colon and rectal cancers: an ecological study. Nutr Cancer. 2003;46:158-165.
24. Hurley SF, Jolley DJ, Kaldor JM. Effectiveness of needle-exchange programmes for prevention of HIV infection. Lancet. 1997;349:1797-1800.
25. Shridhar K, Rajaraman P, Koyande S, et al. Trends in mouth cancer incidence in Mumbai, India (1995-2009):
an age-period-cohort analysis. Cancer Epidemiol. 2016;
42:66-71.
26. Magnus P, Jaakkola JJK, Skrondal A, et al. Water chlorination and birth defects. Epidemiology.
1999;10:513-517.
27. Susser M. The logic in ecological: I. the logic of analysis. Am J Public Health. 1994;84:825-829.
28. Koopman JS, Longini IM. The ecological effects of individual exposures and nonlinear disease dynamics in populations. Am J Public Health. 1994;84:836-842.
29. Cerda M, Tracy M, Keyes KM, Galea S. To treat or to prevent? Reducing the population burden of violence-related post-traumatic stress disorder.
Epidemiology. 2015;26:681-689.
Chapter Questions
1. State the main difference between the following study designs:
a. Observational and experimental studies
b. Retrospective cohort and prospective cohort studies c. Cohort and case–control studies
2. Briefly describe a cross-sectional study and indicate its main limitation.
3. Briefly describe the situations in which the hallmark limitation of a cross-sectional study is avoided.
4. Briefly describe an ecological study and indicate its main limitation.
5. Briefly describe the situations in which an ecological study may be preferred over other observational studies.
6. State which observational study design is best (i.e., most efficient and logical) in each of the following scenarios:
a. Identifying the causes of a rare disease
b. Identifying the long-term effects of a rare exposure
c. Studying the health effects of an exposure for which information is difficult and expen- sive to obtain
d. Identifying the causes of a new disease about which little is known
e. Identifying the short-term health effects of a new exposure about which little is known
f. Identifying the causes of a disease with a long latent period
7. Which type of study is being described in each of the following scenarios?
a. A study that examines the death rates from colon cancer in each of the 50 U.S. states in relation to the average percentage of residents in each state undergoing colonoscopy screening
b. A study that compares the prevalence of back pain among current members of the automobile manufacturing union with that of current members of the bakers and con- fectionary union
c. A study that evaluates the relationship between breast cancer and a woman’s history of breastfeeding. The investigator selects women with breast cancer and an age-matched sample of women who live in the same neighborhoods as the women with breast can- cer. Study subjects are interviewed to determine whether they breastfed any of their children.
d. A study that evaluates two treatments for breast cancer. Women with stage 1 breast can- cer are randomized to receive either cryotherapy (a new treatment involving extreme cold to kill cancer cells) or traditional lumpectomy. Women are followed for 5 years to determine whether there are any differences in breast cancer recurrence and survival.
179 Chapter Questions
e. A study that began in 2010 of the relationship between exposure to chest irradiation and subsequent risk of breast cancer. In this study, women who received radiation ther- apy for postpartum mastitis (an inflammation of the breast that occurs after giving birth) in the 1950s were compared with women who received a nonradiation therapy for postpartum mastitis in the 1950s. The women were followed for 60 years to deter- mine the incidence rates of breast cancer in each group.
8. Indicate whether the following statements are true or false:
a. Observational studies of preventions and treatments are often conducted when exper- imental studies are unethical or infeasible.
b. The main limitation of observational studies is the investigator’s inability to have com- plete control of extraneous factors called confounders.
c. A cross-sectional study of the relationship between blood type and the risk of cataracts will produce misleading results because you cannot tell the correct temporal relation- ship between the exposure (blood type) and disease (cataracts).
d. An ecological study was done to determine the relationship between per capita soft drink consumption and mortality rates from diabetes in 10 U.S. states. The investiga- tors found a strong association between soft drink consumption and diabetes mortal- ity. Based on this study, we can conclude that the individuals who consumed the soft drinks were the ones who died from diabetes.
e. Case–control studies are inherently inferior to cohort studies.
f. Experimental studies are inherently superior to observational studies.
g. Prospective cohort studies are inherently superior to retrospective cohort studies.
© Smartboy10/DigitalVision Vectors/Getty Images
▸ Introduction
An experimental study, commonly known as a trial, involves the use of designed experiments to investigate the role of some agent in the prevention or treatment of a disease. In this type of study, the investiga- tor assigns individuals to two or more groups that either receive or do not receive a preventive or therapeutic treatment. The active manipulation of the treatment by the investigator is the hallmark that distinguishes exper- imental from observational studies (see FIGURE 7-1).
Because experimental studies more closely resemble controlled laboratory investigations, most epidemiologists believe that, if well con- ducted, they produce more scientifically rigorous results than do obser- vational studies. In a laboratory experiment, the investigator regulates all important aspects of the experimental conditions, permitting the exper- imental subjects to differ only for the purpose of testing the hypothesis.
For example, a laboratory experiment testing the toxicity of a chemical is conducted on genetically similar animals (such as mice or rats).1 Ani- mals are assigned (usually by chance) to either the test or control group.
CHAPTER 7
Experimental Studies
LEARNING OBJECTIVES
By the end of this chapter the reader will be able to:
■ Distinguish between the types of experimental studies, including individual versus community trials, preventive versus therapeutic trials, parallel versus crossover trials, and simple versus factorial trials.
■ State the established sequence for conducting trials of new drugs.
■ Describe the key features of conducting experimental studies, including the enrollment and consent process, randomization, use of placebos and masking, maintenance and assessment of compliance, follow-up and ascertaining the outcomes, and data analysis.
■ Discuss the special ethical issues of experimental studies, including equipoise and use of placebo controls.
181
Using identical routes of administration (e.g., by mouth or injection), the chemical under investigation is given to the test group, whereas an inert chemical such as a salt solution is dispensed to the control group. All experimental animals are kept in the same physical environment and fed the same diet and follow the same daily schedule. Thus, the only differ- ence between the two groups is the dissimilar chemical (test versus inert) deliberately introduced by the investigator.
Although experimental studies conducted among free-living humans can never achieve the same degree of control as laboratory animal experiments, many aspects of human experimental research emulate the principles of laboratory research. This chapter describes the design, conduct, and analysis of experimental studies among humans. First, an overview of experimental studies is presented, followed by a detailed discussion of each aspect of an experimental study. The chapter concludes with a discussion of the numerous special considerations in experimental studies that make them scientifically rigorous but difficult to perform.
▸ Overview of Experimental Studies
Investigators must formulate a hypothesis before launching an epide- miological study. Hypothesis generation is a creative endeavor in which an investigator proposes a specific idea to explain a set of observations.
Next, the epidemiologist must decide which type of study design will FIGURE 7-1 Schematic representation of experimental study implementation.
Investigator
assigns Time passes
Treatment group
Source population
Comparison group
Compares incidence of
outcomes
efficiently provide valid information to either support or refute the hypothesis. The appropriate study design is determined by considering the state of knowledge, the frequency of the exposure and disease, and the expected strength of the association between the two. In addition, several practical and ethical problems must be solved when experimental studies are conducted.
Once the study has been designed, a research proposal is written and funding is solicited. At this time, the investigators usually request approval for the study from the ethics committee of the participating institutions. The ethics committee, also called the institutional review board, reviews all studies to ensure that the research is ethical and legit- imate.2(p86) The review is particularly important in experimental stud- ies because a state of equipoise must exist for the study to be ethical.
Equipoise is characterized by genuine uncertainty about the risks and benefits of the test treatment. That is, there must be sufficient certainty that the test treatment might be beneficial to administer it to some individuals while sufficient doubt exists about its benefits to withhold it from others. Next, during the recruitment phase, investigators enroll individuals in the study on the basis of specific eligibility criteria. The eligibility criteria consist of inclusion and exclusion criteria that stem from both scientific and practical concerns. For example, the clopi- dogrel versus aspirin in patients at risk of ischemic events (CAPRIE) experimental study compared the effectiveness of clopidogrel with that of asprin in reducing the risk of heart attacks and strokes among patients with atherosclerotic vascular disease.3 Inclusion criteria consisted of an established diagnosis of recent stroke, heart attack, or peripheral arte- rial disease. Exclusion criteria included contraindications to the study drugs because of potential side effects and geographic factors that made participation unrealistic. The eligibility criteria influence the general- izability of the study, which is the larger population to whom the study results are applicable.
Subsequently, eligible individuals must give formal consent to par- ticipate, usually by signing a consent form that has been approved by the ethics committee. Investigators then use either random or nonrandom methods to assign individuals to receive one of the two or more treat- ments being compared. Random assignment, usually termed random- ization, is preferred because its unpredictable nature makes it less prone to bias. The group that is allocated to the agent under study is generally called the treatment group, and the other group is called the compari- son or control group. The comparison group may receive no treatment at all, an inactive treatment such as a placebo, or another active treatment.
Often, the active treatment given to the comparison group is the current standard of care. For example, the CAPRIE study compared a new drug (clopidogrel) with the standard treatment (aspirin). Because the aspirin treatment, while effective, had serious potential side effects, such as gas- trointestinal bleeding, there was great interest in finding another effec- tive treatment with fewer adverse effects.3
Overview of Experimental Studies 183
In the next phase of a trial, the treatments are administered accord- ing to a specific protocol. For example, the CAPRIE study patients were asked to take tablets of clopidogrel or aspirin every day with their morn- ing meal for at least 1 year.3
During the follow-up stage, the treatment and comparison groups are monitored for the outcomes under study. If the goal of the study is to prevent the occurrence of disease, the outcome may include the pre- cursors of disease or the first occurrence of disease (i.e., incidence). On the other hand, if the study is testing a new treatment among individ- uals who already have a disease, outcomes may include disease recur- rence, symptom improvement, length of survival, or side effects. In the CAPRIE therapeutic study, outcome measures included the recur- rence of stroke and heart attacks, primary brain hemorrhage, and leg amputation.3
The length of time that subjects are followed depends on the out- comes being studied. Only a few months may be needed in a short-term study of drug side effects, but a decade may be necessary for examining slowly developing diseases such as cancer. During follow-up, investiga- tors maintain contact with participants through periodic visits, phone calls, or letters. Losses to follow-up and dropouts must be kept to a min- imum during this phase to ensure a successful trial. For example, in the CAPRIE study, follow-up visits with physicians took place several times a year for up to 3 years, and therefore the number of losses to follow-up and dropouts was quite low.3
All reported outcomes of interest are confirmed to guarantee their accuracy. Confirmation is usually done by masked investigators who gather corroborating information from objective sources such as med- ical records. Masking, which means that the investigator is unaware of the participant’s treatment assignment, reduces the chance of making a biased assessment of the outcome and therefore improves the rigor of the study.
During the follow-up stage, investigators assess participants’ level of compliance with the treatment regimen. That is, they determine whether participants are following exactly the study protocol. In the CAPRIE study, follow-up visits included questions about the use of study medications and any other drugs.3 Fortunately, less than 1% of the patients were completely noncompliant (i.e., they never took the study drugs as directed). Noncompliance is problematic because it makes it more difficult for investigators to determine whether an experimental treatment is effective.
The classical approach to the analysis of an experimental study is known as an intent-to-treat analysis. Here, all individuals allocated to a treatment are analyzed as representing that treatment regardless of whether they completed or even received the treatment. Investigators take this approach to preserve the baseline comparability of the groups and to provide information on the effectiveness of a treatment under real-life conditions.
▸ Types of Experimental Studies
Experimental studies can be classified in several ways depending on their design and purpose. The major types of categorization are described in the following sections and are summarized in TABLE 7-1.