Why do people view the placebo effect as if it’s “no effect?” Consider the following quote by Steven Novella, MD over at “Science Based Medicine.”

These placebo effects include: Regression to the mean – when symptoms flare, they are likely to return to baseline on their own. If you take any illness that fluctuates in severity, any treatment you take when your symptoms are at their peak is likely by chance alone to be followed by a period of less intense symptoms.

This is incorrect. The placebo effect is not an incorrect statistical analysis, or an incorrect assignment of natural variation to treatment. Placebo means that there is no pharmacological effect, but this does not mean that there is no effect.

A real placebo effect is a psychobiological phenomenon occurring in the patient’s brain after the administration of an inert substance, or of a sham physical treatment such as sham surgery, along with verbal suggestions (or any other cue) of clinical benefit (Price et al, 2008). Therefore, the effect that follows the administration of a placebo cannot be attributable to the inert substance alone, for saline solutions or sugar pills will never acquire therapeutic properties. (How Placebo’s Change the Patient’s Brain)

I think Novella is confusing apparent and actual placebo effects. A placebo group can appear to show results, as can an actual treatment group, due to statistical noise, methodological error, etc. But an apparent change between pre-outcome and post-outcome means is not the placebo effect. Only an actual change in initial and final condition within the plaebo group is an actual placebo effect. This effect is a measurable effect and induces a measurable change in brain activity, as shown in fMRIs. And quite frankly, in some ways it does not matter whether or not an improvement in health is due to the drug itself or the belief that the drug works.

If someone has a blood pressure reading of 180/110 and because they believe a sugar pill works, it goes down to 130/80, that’s a life saving shift. Obviously there is the question of whether or not the effect will continue to work. A placebo may wear off as the person’s belief changes, etc. However, this can also happen for a drug which has its own effect. That effect can be boosted by the placebo effect, which is why we do not say that a treatment has no effect, but rather no effect different from the placebo. In fact, going back to the example of blood pressure, a meta analysis of 23 trials for beta-blockers concluded that “the placebo response accounted for 34% of the drug response for sBP and 47% of the drug response for dBP (Effect of placebo groups on blood pressure in hypertension: a meta-analysis of beta-blocker trials).”

But maybe that effect is simply due to regression to mean or some kind of sampling error. To see if that is the case, a study comparing placebo treatment to no treatment can be performed. This has been done and a difference between placebo treatment and no treatment has been found. (Evaluation of the placebo effect and reproducibility of blood pressure measurement in hypertension) Admittedly, this study is weak, but i is also unnecessary.

The argument that results from the meta-analysis were due to sampling error, regression to mean caused by natural recovery, etc are all unlikely. People with hypertension do not tend to recover on their own. Hypertension is a chronic issue. Also, while there is some variability in blood pressure based on method of reading, time of day, etc, a large sample reduces these statistical anomalies from clouding data. In order for difference of means to be due to regression to mean or other sampling error, individual patients’ blood pressures would have to have varied wildly throughout the studies. Additionally, the meta analysis cited above uses a moderator analysis to compare factors like sample size and trial duration, and found that better quality trials resulted in stronger apparent changes in the placebo group. This is contrary to what one would expect if the apparent change in the placebo group was actually do to factors like sampling, regression to mean, or poor quality control.

Another way to test for the existence of an actual placebo effect is to try to look at cognitive signatures. At least one study has found a relationship between the placebo effect and mu opioid receptors (Chronic mu-opioid receptor stimulation alters cardiovascular regulation in humans: differential effects on muscle sympathetic and heart rate responses to arterial hypotension). Another study created a predictive model using the data from one trial and successfully predicted level of placebo in a second trial (Brain Connectivity Predicts Placebo Response across Chronic Pain Clinical Trials).

All together, we have multiple studies that look at the placebo effect in different ways. In addition to results from clinical trials on drug treatments, we also have a proposed mechanism and data consistent with that mechanism. The ability to create a model predicting strength of the placebo effect, the increase in strength of placebo effect as quality of the study increases, and a proposed mechanism all together make the theory that there is a real placebo effect robust.

Beyond the placebo effect, the nocebo effect could result in a normally effective drug becoming ineffective. View towards the treatment therefore needs to be taken into account and we need to make sure that the placebo component of the drug’s effect does not wane, or at least compensate for its waning, and we need to make sure that view towards treatment is not negatively effecting outcome.

Research Ideas

There are a few research ideas that arise from this discussion. First, if placebo and nocebo can have a considerable impact on certain treatments, we should know this. Skepticism towards standard medicine is increasing and that might have a negative impact on efficacy of certain treatments. To test this effect, we can perform a few clinical studies that measure how positive or negative the patient’s view is towards medicine in general and towards the treatment. If perception has a significant effect, then we would expect the placebo to result in greater positive results among those who positively score medical treatment and weaker, or maybe even negative effects, among those who negatively score medical treatment.

A few additional questions can also be asked, such as whether or not the patient believed he received an actual treatment or a placebo. In order to reduce the risk of how the question of skepticism influences results, some patients should be asked at the beginning of the test, some at the end, and some at both points. This variation will reduce noise caused by expectation generated by the question being asked. The sample size would have to be fairly large in order to make sure that enough data on each sub-sample is collected. While pain medication may show the most robust results, continuing with a less obvious condition like hypertension might be preferred. There are a number of high blood pressure medications which are considered relatively safe and therefore are reasonable for large scale clinical trials.

Because perception of efficacy is important, the patients should be blinded from seeing their blood pressure results.