2.10: Controlling Variables

Experimental and control groups should be as similar as possible except for the studied variable (e.g., diet pill). Subjects should be randomly assigned. Allowing them to choose their groups adds bias. Many people with advanced cancer desperately want to take part in studies of new treatments, but some will do so only if they’re assured of being treated. But putting them into the treatment group, and those who “don’t care” into the control group clouds the outcome. Did one group do better because of the treatment or because of uncontrolled differences in the groups?

Sometimes, due to public pressure, patients are allowed access to an unproven treatment without participating in the study. But these same patients can get angry if the study finds that the treatment is ineffective or harmful, especially if the treatment was severe.

Placebos should be used whenever possible. A placebo is a substance or sham procedure used for its psychological or peripheral effect (placebo effect). The placebo for a vitamin pill might be a sugar pill. If a drug dissolved in saline (salt water) is injected, an injection of saline alone would be the placebo; the injection process or the saline might have effects that would otherwise be falsely attributed to the drug.

Placebos should look identical to the test substance (small red pills tend to “work better” than medium-sized green pills). Placebo effects can be powerful. It would be easier for pharmaceutical companies to prove a drug’s effectiveness if they could test it alone, rather than in comparison to a placebo.

Ideally, studies are double-blind—the subjects and investigators both (double) are blind to experimental groupings. If there are injections, neither the person getting the injection nor the person giving the injection knows whether it contains the drug or is a placebo.

Double‑blind studies control for bias and other psychological effects in the subjects and investigators. For example, many parents and teachers believe that sugar causes hyperactivity, and many children have been told this also. In studying the effect of sugar on hyperactivity, you can see why the children and the parents and teachers evaluating them shouldn’t know whether the children have been given sugar. (Double-blind studies show that sugar doesn’t cause hyperactivity.)

In studying a cancer drug, pathologists evaluating cancer in tissue samples may have to decide between two grades of severity when the pathology is intermediate. Investigators and subjects alike hope the drug is effective. If it’s ineffective or only marginally effective, bias can be enough to tilt the outcome.

Double‑blind studies typically have an oversight committee that knows the groupings and monitors the results so the study can be stopped if a clear benefit or hazard appears. In the double-blind study of aspirin in preventing heart attacks, the oversight committee advised ending the study early because far fewer heart attacks occurred in the group taking aspirin. It was deemed unethical to continue the study because this would withhold the benefit of aspirin from the placebo group.*

In a Women’s Health Initiative study, it became shockingly clear that the women given a combination of estrogen and progestin had an increased risk of breast cancer, and this study was stopped early.

In some studies, psychological or peripheral effects can’t be neutralized. One can’t do a double-blind study comparing coronary bypass surgery to drugs for preventing a second heart attack. To control for a placebo effect of surgery itself, the surgeon needs to open the chest of those in the medication group without doing a bypass (sham surgery). This usually isn’t done, but possible peripheral effects should be kept in mind when assessing such studies.**

This point is well illustrated by a study reported in 1959. Earlier studies of patients with heart disease showed that tying off arteries to the mammary (breast) glands relieved angina (pain from inadequate oxygen to the heart muscle), and the procedure was popularized by an enthusiastic article in Reader’s Digest. There was no physiological explanation for why it worked; a placebo effect was suspected.

To test this hypothesis, 17 patients with severe, disabling angina were told only that they were participating in an evaluation of this simple surgical procedure done under local anesthesia. All subjects thought that their mammary arteries were going to be tied off, but were randomly assigned to either the control or experimental group. In both groups, the same incisions were made and the mammary arteries located, but the arteries were tied off only in the experimental group (8 patients). All subjects were followed for 3 to 15 months, and exercise tolerance (minutes on treadmill without angina), number of nitroglycerin pills self-administered for pain relief, etc., were compared with the same measurements before surgery.

Both groups improved, not only in the patients’ own subjective evaluations, but also in exercise tolerance and number of nitroglycerin pills taken per week (average 37 pills pre-surgery; 21 post‑surgery). In fact, the control group improved more than the experimental group. The report says, “...improvement in some cases was extraordinary. One patient [in the control group] who had been unable to work because of his heart disease, was almost immediately rehabilitated and was able to return to his former occupation. He reported a 100% improvement at 6 months and 75% improvement after a year.” Controlled studies are crucial to discovering the scientific truth.

---

*The 2019 American Heart Association Guideline is to avoid daily aspirin unless your physician prescribes it (when benefit deemed to outweigh risk, e.g., bleeding.

**Cost-containment pressures have brought back the controversial use of sham surgery to test new surgical procedures for certain diseases. People with Parkinson’s disease, for example, have good days and bad days in muscular debilitation. A positive outlook can help, perhaps by patients being more active and thereby improving muscle function. Sham surgery for implanting fetal cells into the brains of Parkinson patients would mean at least drilling the requisite holes in the skull. Brain surgery that just “seems to work” isn’t very good evidence for a procedure that in itself is risky and expensive.