Experimental Studies

In scientific experiments, investigators deliberately set one or more factors to a specific level. The word ‘experiment’ has a different meaning in statistics and science from that in the everyday life where the word is used to mean “try something to see what happens”. In science, an experiment is carefully controlled in such a way to test a priori scientific hypotheses, as explained in the first module.

As rigorous inductive logic-based statistical hypothesis testing is involved, experimental studies always lead to stronger scientific inferences than do observational studies. In contrast to observational studies, experimental studies involve deliberate intervention by the investigator and therefore ethics plays a major role.

For example, to study whether heavy alcohol consumption is associated with lung diseases, it is unethical to ask someone to heavily consume alcohol (in such situations only ethical option is to resort observational study as explained previously). In studies where human or other animal subjects are not involved, experimental manipulations usually do not infringe ethical interpretations.

Comparative experimental study

A comparative experimental study involves experiment that compares two or more techniques, treatments, or levels of a variable.

For example, subjects with a specific treatment are compared with subjects with placebos (without treatment, negative control). Subjects on a specific drug are compared with that on another drug. The growth of laboratory animals on various diet regimes are compared.

Microalgal cultures were grown under various light regimes to infer the optimum luminous intensity for the growth. Case-control study as previously explained is a type of comparative study, but it is not an experimental study as no deliberate intervention is involved.

Randomized Control Trials

In Randomized Control Trials (RCT), cohorts are randomly assigned to either of the two groups (cases or control) and are followed in time towards the future. RCT is a type of longitudinal and prospective experimental study.

RCT is similar to case-control study, but rather than merely observing cases and controls, in RCT cases are deliberately intervened (with therapeutics, diet, exercises etc).

For example, a new generation chemotherapeutic drug to see whether the drug is effective. Oftentimes controls are patients undergoing treatments with standard drugs so that the efficiency of a new drug can be compared with the standard drugs. While case-control studies are retrospective, RCT is prospective.

RCT is oftentimes executed as a cross-over study where the same cohort receives more than one treatment sequentially in non-overlapping time periods. This is because problems with confounding (discussed in next section) can be minimized as the same cohort in one experiment also serves as his or her own control in the subsequent experiment.

In RCT, assignment of subjects into groups is done randomly (statistical randomization using tables, software etc.). For example, to study whether dental flossing is associated with good heart health, subjects are randomly assigned into two groups with one group instructed to floss every day while the other not to floss.

Subjects are then followed into future to see the development of cardiovascular diseases, the endpoint. While flossing is an innocuous intervention, many other interventions are unethical; for example asking the subjects to abuse alcohol, smoke and so on. In such cases, observational studies (prospective cohort study or retrospective case-control study) can only be done.

RCT is considered as a gold standard in pharmaceutical experimentation and exposure assessment and it leads to strong inferences about the efficacy of the drug candidates, although this type of studies tends to be very expensive.

Superiority Trials

Trials are experiments done on human subjects; trials are often employed in epidemiological, clinical, pharmaceutical, psychological, behavioural research etc.

In superiority trials, efficacy of an intervention (exercise, diet, devices, new drug candidate) is compared against the control which is usually patients undergoing a standard course of treatment (standard exercise regime, diet, devices, drugs etc.). These trials are aimed at testing the superiority of new clinical intervention compared with the existing treatment scenario.

Noninferiority Trials

In noninferiority trials, a therapeutic intervention is compared with the standard treatment scenario to test whether the new intervention is performing ‘as good as’ (or ‘not inferior to’) the existing scenario.

This is especially useful when testing a new intervention that is as effective as the existing drug but is significantly cheaper (generic drugs for example), or has far lesser side effects in comparison with the existing drug. Like in superiority trials, noninferiority trials also have control groups with standard treatment scenario.

Single Blind vs. Double Blind

A study is single blind if subjects being treated are unaware of which treatment (including any control or placebos) they are receiving. For example, doctors injecting the patient with a drug, exact constituents of which deliberately withheld from informing the patients.

Patients are then followed towards the future and are evaluated. Here, the person who is evaluating the outcome (the doctor) is not blind; he is aware which drug he administered to the patient.

Another example is single-blind review system adopted by many journals; when an author submits a paper to a journal, the journal forwards the paper to an external reviewer who remains anonymous to the author. However, the author is not anonymous to the reviewer.

A study is double-blind if it is single blind and the people who are evaluating the outcome variables are also unaware of which treatment the subjects are receiving. A simple example is the standard practice of modern diagnostic centres; when a patient gives her blood sample, technician affixes a barcode on the sample vial and transfers the vial to the laboratory.

The biochemist performs the test and writes the report. Neither the patient knows the identity of the biochemist who is writing the report, nor the biochemist knows the identity of the patient.

Most of the modern scientific journals adopt double-blind peer review system in which in addition to the fact that reviewers remain anonymous to the authors (as in single-blind peer review), authors too remain anonymous to the reviewers.

Quasi Experimental Designs

These types of studies fall somewhere in between observational studies and experimental studies. In quasi-experiments, there is an intervention although the execution of such an intervention is not completely planned.

For example in a number of places around the world food containing trans fats (for example desi ghee foods, samosa, ladoo, burgers etc.) have banned, as the link between trans fats and cardiovascular diseases has been revealed by many recent studies.

An investigator can study effects of this legislative policy decision by studying the prevalence of cardiovascular diseases prior to the enforcement of this policy (by looking at registers and hospital records etc.) and post-implementation.

Another example would be studying the effects of nuclear fall out on the populations of Chernobyl, Hiroshima and Nagasaki before and after nuclear fallout. Here, intervention (exposure to nuclear radiation) is not deliberately planned and executed.

Natural Experiments

Natural experiments too, like quasi-experiments, falls in between observational study and experimental study. Like quasi-experiments, there is an intervention and this intervention is not deliberately planned or executed. Intervention, in this case, is a natural phenomenon. For example, a disease outbreak or natural calamity.

Does religiosity be related to natural calamities? To study such an association, investigators can assess religiosity of a number of communities that are living in a high-risk zone for natural calamities (earthquakes, tsunami, cyclone etc.). Investigators can re-measure their religiosity after the incidence of natural calamity; if at all the natural calamity strikes.

Summary

• In the experimental study, investigator first formulates a hypothesis and plan the experimental design to test the prior hypothesis. Such studies involve deliberate intervention by the investigator. Such an intervention has ethical constraints in studies involving humans.
• A comparative experimental study involves experiment that compares two or more techniques, treatments, or levels of a variable.
• In Randomized Control Trials-the current gold standard in clinical research, cohorts are randomly assigned to either of the two groups (cases or control) and are followed in time towards the future. Cases are intervened with certain drugs or procedures.
• In superiority trials, the efficacy of a certain drug candidate or a new procedure/device is tested with the standard treatment scenario to see whether the new intervention is superior or not. In noninferiority trials, new treatment scenario is tested against standard treatment scenario to infer whether the new technique is as good as the existing one.
• A study is single blind if subjects being treated are unaware of which treatment (including any control or placebos) they are receiving. A study is double-blind if it is single blind and the people who are evaluating the outcome variables are also unaware of which treatment the subjects are receiving.
• In quasi-experiments, there is an intervention (a governmental policy decision, an accident etc.) although the execution of such an intervention is not completely planned. In the natural experiment, intervention is a natural phenomenon (spring tide, solar eclipse etc.) or a natural disaster (earthquake etc.)