Types of Research Studies

CHAPTER 17 Types of Research Studies




Scientific studies pursue truth. In evaluating a study, the initial consideration is whether the study arrived at this truth successfully.


—D. J. Friedland1


Congratulations! You now have the map reading skills necessary to navigate the literature. You will find, as you peruse the journals and other sources of information, that there are several types of studies, each with their own benefits and drawbacks. For instance, the double-blinded, randomized, controlled trial (DBRCT) is considered the gold standard in clinical research. It is purposely designed to minimize potential sources of bias. When it is adequately powered and properly executed, it has both accuracy and precision.


Unfortunately, for various reasons not all trials can be conducted as DBRCTs. Certain criteria have to be met, such as blinding the subjects as to the particular pathway in which they are enrolled. If they are unaware of the treatment to which they are exposed, there will be no subconscious expectations that could potentially influence results. However, certain types of experiments do not lend themselves to blinding. For instance, in studies that compare surgical interventions to medical treatments, neither the subjects nor the researchers can be blinded.


Ethical standards for clinical trials must be met as well. It behooves all researchers to enter into a study with equipoise (discussed in Chapter 10) so that the possibility of either treatment being better, or even potentially harmful, is considered in the analysis. After all, the study is being conducted because we often do not know whether one pathway has an advantage or harbors possible deleterious effects. Participants must be aware of the possibility of harm even though it may be minimal.


Even though the DBRCT has the distinction of producing the most dependable results, many other types of experiments have an important place in clinical literature. Not every trial can be a DBRCT, nor does it need to be. Observational studies, in which people choose their own course and are followed over time, can contribute immensely to the fund of medical knowledge. Most epidemiological studies are done this way, such as the famous Framingham studies that began in 1948.2 This database recorded, among other things, the lifestyle habits of thousands of individuals and then looked at their subsequent rate of heart disease. This invaluable information led to our current understanding of cardiovascular risk factors.



CASE REPORTS


What prompts an individual to pursue a particular research question? The earliest clue that there could be an association among particular variables may be reported in the simple but intriguing case report. When a clinician sees something out of the ordinary, reporting this in a journal alerts other practitioners to a possible connection. For instance, at a dermatology conference in San Francisco in the early 1980s, a cluster of several cases of homosexual men with an unusual type of skin cancer called Kaposi’s sarcoma was reported.3 Was there a connection between gay men and skin cancer? There most certainly was! This observation led to research that resulted in the identification of the virus that caused a major epidemic of immense ramifications, namely, autoimmune deficiency syndrome (AIDS).



OBSERVATIONAL STUDIES


Once a potential connection between variables has been identified through case reports, observational studies are done that attempt to elucidate the presence and strength of this association. Venn diagrams are a rough way to illustrate observed connections between variables, but they have little scientific validation. Correlation is one way to comment on the strength and direction of the association. The correlation coefficient tells us not only whether there is a significant association, but also the degree of impact of one variable upon another and whether increasing the value of one has a positive or negative association with the other. Correlation does not tell us whether one variable caused the effect on the other, only that there is a link. When the association is measured at a single point in time (as opposed to the effect over time), this is known as a cross-sectional study.


Keep in mind that two variables may be linked through a third “hidden” variable. This was the case with gay men and Kaposi’s sarcoma. The link was eventually identified as the human immunodeficiency virus (HIV). If we were comfortable in assuming that there was no other connection between homosexuality and skin cancer, and did not look further, we would miss this fact. These hidden variables are called confounders. They often account for a major portion of the association of other variables through their independent connection with each of them. Do African-American children perform less well in school than Caucasians because of their skin color? Most experts would agree that even though an association between skin color and school performance can be demonstrated, the connection has to do with confounders such as socioeconomic class.



CASE–CONTROL STUDIES


Observational studies can also be done retrospectively. These are known as case– control studies; the outcomes are already known. We take people with the outcome and try to match them in many other respects with healthier counterparts. Then we look back to see if there is a correlation between a particular variable and an adverse outcome. If so, we infer that if we remove that variable we can improve outcome. For instance, a recent retrospective study looked at a sample from a population of people with heart attacks. All other things being equal, did anemia have an adverse effect on the outcome of mortality? The analysis showed that those with severe anemia fared worse than those with normal blood counts. This supports the use of blood transfusion in those patients with heart attacks who also have severe anemia.4


I like to think of case–control studies as people getting off a train at their destination. You greet them at the station named Disease and then look in their suitcases to see what items they have carried with them. Then you go to the Healthy station and see what those individuals have packed. The overall difference in their baggage could account for their destination. For instance, people arriving at the Coronary Artery Disease station will have a greater tendency to pack cigarettes, bacon and eggs, and insulin for their diabetes. They will be more likely to carry the death certificates of parents and siblings who died of heart disease. Those who get off at the Healthy station will have more jogging suits and turkey sandwiches. There will be some overlap of items but you will see a distinct difference in the overall baggage between the two destinations.


Retrospective studies are especially applicable to outcomes that are relatively rare or take a long time to develop. This type of study can be done without waiting for the subjects to develop the disease. One of the limitations of this type of study exists in finding an appropriately matched control group. They need to be similar to the diseased group in all respects (except for the disease and the variable under study). If there is some other difference not accounted for, an unidentified confounding variable could be responsible for the outcome. This could lead to an incorrect conclusion with regard to the other variable under study.


Retrospective studies are also prone to recall bias when they rely on the subjects’ memory of exposures to substances. Those with the disease may have more accurate recall of their exposures. On the other hand, the subjects might also overestimate an exposure they believe might have led to the disease. There are some mathematical limitations as well. Calculation of risk involves the development of disease over time, with a common starting point for both disease and disease-free groups. Since we look backward in time, we have no true starting point. This allows us to calculate odds ratios (a measure of association) rather than risk (which reflects a rate).



COHORT STUDIES


When we have the time to wait for the outcome to develop, a prospective study that goes forward in time can be done. Cohort studies start with a group of subjects and measure the relative incidence of disease in those who are exposed to a variable, in comparison to those not exposed. The groups in a prospective study are not necessarily randomized or blinded; they may choose their exposures. The Framingham Study is a good example of a prospective cohort design. Because the exposure preceded the outcome, there is a stronger argument in favor of the exposure causing the disease, although this in itself is insufficient to prove causation. There is a starting point in time, so risk can be calculated. Recall bias is minimized in this design. However, one drawback is the prolonged time and expense these studies require to carry them to completion while waiting for the observed outcome.



RANDOMIZED CONTROLLED TRIALS


This brings us to the poster child for evidence-based medicine, the double-blinded, randomized, controlled trial (DBRCT). There are many potential opportunities for bias to creep into studies, which can ultimately result in an erroneous conclusion. The carefully contrived DBRCT, although not guaranteed to be completely bias-free, has eliminated many types of bias that are woven into other trial designs. This study resembles a prospective cohort study with a few important distinctions.


When subjects agree to participate, they are allocated, through a random process, to one of two or more pathways. If subjects were allowed to choose their pathway, as in an observational study, this would introduce other factors that could account for the outcome. Consider a cancer trial that allows subjects to choose either the Standard of Care Drug or a New Drug. Within this sample of cancer patients, the healthier ones may decide, even subconsciously, to stick with the familiar Standard of Care Drug while the more desperate ones may be willing to take a chance on the new treatment. The point to be made here is that the Standard of Care Drug pathway in this case is destined to have a better outcome, even if it is not a better drug because the groups going down the two pathways are not the same. Randomization eliminates the possibility that a treatment could show an erroneous advantage through this type of selection bias. If potential confounders are present (even if they cannot be identified), they should be equally distributed among the groups and thus not contribute to a difference in outcomes.

Related posts:

  1. Types of Statistical Tests
  2. Ethics in Medical Research
  3. Measures of Disease
  4. Mathematical Principles

Stay updated, free articles. Join our Telegram channel
Tags:
Jun 18, 2016 | Posted by in BIOCHEMISTRY | Comments Off on Types of Research Studies

Full access? Get Clinical Tree
Get Clinical Tree app for offline access