The Placebo Effect in Medicine

History

The term placebo - Latin for I will please - finds its origin in the late Middle Ages: it comes from the recitation of the psalm 116 Placebo Domino in regione vivorum (I will please the Lord in the land of the living)[i], during Christian celebrations of the dead. With the change of having paid singers sing the devotion instead of relatives, the phrase "singing a placebo to someone" came to be used as a paid, hypocritical substitute.

The placebo term found its way into medical jargon through the English physician Alexander Sutherland, at the end of the 18th century. In order to give in to the urgings of dissatisfied patients during a therapy and to calm them down, little medicinal substances were prescribed supplementary, which were called placebo, in the sense of a substitute action (Jütte, 2013). In medical dictionaries the term can be found for the first time in the beginning of the 19th century.

---

[i] Lat. Translation after Jerome verse Placebo domino in regione vivorum Ps. 114,9

Placebo term

Modern understandings of the placebo concept can be traced back to the most widely cited work, The powerful Placebo (Beecher HK, 1955). In this work, the magnitude of a placebo response, as a positive effect not due to the therapeutic intervention, was located at about 35% of all uses.

Since then, this one-third limit has persisted in the scientific literature as a magical constant of nature and a seminal benchmark for assessing therapeutic interventions and drug effects.

However, a comprehensive and precise definition of the term placebo does not yet exist (as of 2021). This fact is primarily based on different explanatory approaches that are difficult to reconcile with each other, their different research approaches and the associated criticism.

In addition, in the baggage of a helpful placebo effect also sits the reverse, damaging effect of a nocebo effect, which also requires a more detailed explanation.

Placebo Research

Placebo research has been striving for a deeper understanding since the middle of the 20th century until today. Different models have been developed, which are first presented in a short overview.

Placebo as a statistical phenomenon

The currently prevailing view describes a placebo simply as the opposite of verum. This view is based on the practice of the double-blind, randomized controlled trial (RCT), which is considered the gold standard for proving a drug substance.

Accordingly, a therapeutic agent is considered to be verum if its efficacy:

• a recognized valid theoretical foundation is/can be assumed,
• under controlled conditions, clear differences between verum and an inert substance as non-verum can be statistically proven.

Consequently, according to this model, everything would be placebo that triggers a demonstrably positive effect but does not have an accepted model of effect and is not verum. Similarly, anything that is not considered verum and conversely produces a negative effect would be a nocebo (see below). In clinical jargon, this is often simplistically concluded:

A placebo would be a placebo drug or a placebo intervention.

A consistent derivation for nocebo does not yet exist in clinical parlance.

The condition of double blinding of the RCT with selected group and control group shows the importance that statistical medicine ascribes to the individual, the encounter and the interaction between practitioner:in and patient:in. In this model, however, these factors are evaluated exclusively as distorting confounding factors that must be eliminated as far as possible in order to assess efficacy.

The limitation of this explanatory approach is based on the fact that current pharmacological research assumes purely linear cause-effect chains to explain healing effects.

Placebo as a contextual concept

Therapeutic effects can be due to the interaction between practitioner:in and patient:in and can be modulated by expectations without a drug having been given or an intervention being or having been intended. Since this occurs in every encounter, it would be more accurate to use the broader term and instead of placebo, it would be better to speak of a placebo effect.

From this perspective, all non-verum interventions that lead to improvement, and can be distinguished from the natural course, are by definition placebo effects. According to this reasoning, all types of applications always have a placebo component, but this cannot be sharply defined.

This leads to the interesting question whether it is not in principle wrong to define single components of a treatment process as the only acting factors. The processes that contribute to successful healing in the treatment of sick people quickly become so complex when considered individually that they can hardly be meaningfully integrated into a narrow conceptual system. The context in which the process takes place must therefore be analyzed as well.

Analysis of contextual conditions

Current placebo-nocebo research therefore seeks an interdisciplinary working concept of general acceptance, focusing in its analysis on the contextual factors of treatment:

• Effects of expectations (outcome related and self-efficacy),
• conditioning,
• Induction or reduction of anxiety,
• Handler-patient interactions.

Numerous interesting works show that these factors have a significant impact on the recovery process. Here are some examples:

• Faster recovery after hysterectomy, less postoperative fever, less digestive disturbances, and shorter hospital stays when optimistic content is conveyed (Evans & Richardson 1988).
• A better response to analgesics when injected overtly and visibly vs. covertly, applied incidentally (Pollo et al., 2001).
• Warm, friendly and supportive practitioners are more successful in all interventions (Di Blasi et al., 2001).
• Surgical procedures, such as adhesiolysis for chronic abdominal pain, have no proven advantage compared to pure laparoscopy (Swank et al., 2003).

The list of just surgical interventions that have not been shown to be causally effective after the introduction of a placebo intervention is long.

Most explanations and working hypotheses in this approach to contextual research are limited to neuropsychological explanatory models. It is easy to understand and should be undisputed that such an operationalization carries the risk of unintentionally excluding important, yet unexplained areas of explanation.

Placebo as an effective agent

Few papers attempt the explicit efficacy of a placebo intervention through open administration of an inert substance.

Patients with irritable bowel syndrome showed impressive improvements after 3 weeks of openly given placebo medication compared to a control group without treatment (Kaptchuk et al., 2010). Outcome: global improvement (p=0.02). Therapy was initiated after 15 min of conversation with the following statements:

• The placebo effect is powerful
• The body can respond automatically
• A positive attitude is helpful, but not necessary
• It is important to take the tablets reliably

A meta-analysis comparing six antihypertensive drugs with placebo in hypertensive patients showed that 30% of the participants in the placebo group achieved the target blood pressure, compared to 58% in the verum group. In both groups, therapy had to be discontinued because of serious adverse drug reactions, 12% in the placebo group and 13% in the verum group. In the placebo group, headache and joint pain were twice as common, and cold feet about three times as common. In contrast, more patients complained of drowsiness in the verum group (11.8% vs. 6.6%) - (Preston et al., 2000).

In a careful analysis of analgesic effects, in addition to a verum group, the placebo response was also compared to an untreated control group (no treatment group) and showed clearly detectable effects in the placebo group (Hróbjartsson & Gøtzsche, 2001).

From this it is easy to see that placebo effects can also be demonstrated in physiological and biochemical processes. A purely psychological explanatory model is not sufficient here.

Further reviews have shown that biological and psychological effects of placebo-induced analgesia can also be detected by functional imaging, which maps decreasing neuronal activity in the thalamus, gyrus cinguli, and insular region (Price et al., 2008). In pain management, the efficiency of positive expectancy of efficacy depends on whether patients are cognitively capable of making such an assessment of the intervention. If this is not always possible, higher doses of analgesics must be applied (Benedetti, 2006).

Limits of the placebo effect

In asthma treatment, for example, objectifiable and subjective complaints have been shown to respond differently to verum and placebo. Improvement in one-second capacity (FEV1) was achieved significantly (>20%) only with albuterol, whereas the subjective sense of improvement was 50% in all treatment types compared with the no treatment group (21%) (Wechsler et al., 2011).

The statement that there are always and everywhere placebo effects is not provable here and should therefore not be used in a generalized way, no matter how often it is quoted in the literature.

Summary of placebo research

The significant work on placebo research shows evidence of significant effects primarily in diseases in which the subjective experience plays a major role in the development of the symptoms. This applies above all to chronic pain conditions. This concerns above all the influence on the pain processing and somatoform complaint complexes.

If one follows all the current medical research approaches, the therapeutic benefit of an intervention can be described as the sum of the verum treatment plus the natural course plus the placebo effect. This sounds simple, but such a definition remains fuzzy.

Especially with regard to the main focus of placebo research so far, one has to ask whether a distinction between verum and placebo, as it is often made, is meaningful. Very many results in different papers include multiple therapeutic effects that cannot be explained by verum administration. It follows that the linear cause-effect chain already mentioned is an unnecessary limitation of the pharmacological explanatory model.

To delimit unexplained positive reactions sweepingly as a non-effective agent with a placebo concept basically has no logical justification. It even seems contradictory in the context of complex reactions.

Placebo as an effect paradox

However a definition of placebo is made, it is based on a basic assumption:

Placebo is an agent to which an investigator attributes an effect that he does not understand or expect, or cannot explain with his models of effect. Once such a phenomenon is understood, it is no longer a placebo.

This logical paradox is also supported factually. In a comprehensive follow-up study, the effects classified as placebo effects in Beecher's work [1] could all be attributed to other factors (Kienle, 1995).

As conditions for the recognition of a placebo effect it was assumed in this analysis that

• a placebo was given at all,
• the described effect represented an actual effect of the placebo administration and could not have come about by other means, and
• the placebo effect had therapeutic relevance in the underlying work.

The review shows that, contrary to Beecher's statements and previous interpretations, there was no evidence of a placebo effect in any of the underlying studies. Rather, a variety of factors were shown to have faked a placebo effect:

• Spontaneous improvements
• Spontaneous fluctuations (one-sided mention of improvement with concealment of deterioration rates)
• Regression to the mean (see below for supplementary chapter)
• Selection effects and scaling effects (statistical bias)
• Adjunctive therapy effects (not considered)
• Psychotherapeutic effects (not considered)
• Irrelevant test criteria
• Accompanied responses
• Conditioned responses
• Uncritical use of anecdotes
• Citation chains without reference
• Incorrect interpretation of non-specific symptoms

"...In view of these results, the question arises in conclusion whether the existence of a placebo effect is not merely a medical historical illusion..." (Kienle, 1995)

In this respect, it would make sense from a scientific point of view to dispense with the placebo term and to speak instead of unspecific - in the sense of not yet understood - effects that act on a disease process and require further comprehensive research.

This seems to be justified above all because all reactions declared as placebo effects (especially in the sense of context-related factors, spontaneous improvements and regression effects) always also exist with verum administration and do not only appear in research papers. Especially in everyday practice, such distinctions usually cannot be made because of the large differences in contextual conditions. This is especially true if the differentiation between placebo and verum has already produced only a small statistical difference in research studies. The difficulty is then to justify the small difference purely mathematically, since the conditions for the placebo effect are not known.

Nocebo effect

A systematic nocebo intervention for research purposes is ethically excluded.

However, there are numerous documented situations in which a negative expectation can be aroused, maintained or forced in the patient by the behavior of the practitioner. This is quite well documented especially in relation to pain and fear of intervention and non-specific fears versus feeling "at the mercy" and "nothing helps anyway" (Wei et al., 2018).

Especially in clinical practice, the negative influence in the careless handling of patients by clinical staff as a nocebo phenomenon is common and seemingly more significant in its effects than positive placebo effects (Horsfall, 2016).

Homeopathy and Placebo

For homeopathic therapy the described effects of placebo research are of weighty importance. First and foremost, it is a matter of developing concepts to prove a medicinal effect.

For example, when clear evidence appears in clinical research that allergies and asthma in children can be effectively treated by individualized homeopathy, a simple contextual factor rationale is insufficient as an alternative rationale (Cf. chapter "Placebo as a Contextual Concept").

NOTES: The focus and status of clinical efficacy research are presented and discussed in detail and critically in the chapter on clinical homeopathy research.

In addition, the concepts of placebo research presented at the beginning also offer help, especially for homeopathic case analysis in everyday practice, in order to effectively reduce therapeutic misconceptions.

The central question is: what is context and what is most certainly a medicinal reaction? How can this be systematically documented and evaluated in order to be clearly identified?

Another point is that potentized remedies prescribed according to the homeopathic principle are generally liked to be dubbed as placebo in the literature critical of homeopathy, on the grounds that a satisfactory scientific explanatory model is lacking. (see above definition according to statistical placebo model). From the described point of view this is certainly logical, but it is easily overlooked that just this model, as explained, cannot provide an explanation for the placebo effects per se. In this generality it is therefore a self-referential and unscientific argument.

In addition, all phenomena described as placebo effects, according to a holistic understanding of disease (More on the understanding of disease in homeopathy), which thinks beyond models with a linear causal-effect-chain concept, always represent parts of a complex reaction to be analyzed. At the same time, the converse conclusion that "everything that happens after taking a homeopathic drug can also be attributed to taking the drug" is also not true.

A differentiated knowledge of this branch of research is therefore essential for conscientious scientific work in homeopathic practice.

Regression to the mean

Simplified, the term means: Measured values tend to be closer to a mean value in further measurements. This statistical phenomenon also applies, for example, to the measurement of anatomical characteristics.

This statement has great importance for the interpretation of therapy` effects in clinical efficacy studies, to distinguish normalizations even without therapy influence from actual therapy effects, and to determine the efficacy of an intervention. This is especially crucial for prospective observational studies as applied in clinical care research on homeopathy.

In addition, the understanding likewise helps to delineate apparently therapeutic effects from effects actually caused by the intervention, even in everyday practice.

The effect is ubiquitous and must be considered for all types of interventions.

In the case of linear correlations, it is easy to show that such a regression effect exists. It can be easily determined here, e.g. when measuring the height of children and their parents. As a rule, children of short parents tend to be slightly taller and, conversely, children of tall parents tend to be slightly shorter.

This means that the overall results in both extreme groups are closer to the mean value of the average heights. However, the children in both subgroups remain taller than average compared to the general average for tall parents, with fluctuations, and the short children are usually smaller than average.

This also makes it clear that regression to the mean is not a unidirectional phenomenon.

The comparisons show that although it is possible to make a certain prediction, this is different in both groups. The statistical prediction possibilities are illustrated in the percentile curves to assess the body sizes.

Non-linear correlations

In the case of non-linear correlations, i.e. all study phenomena that can only be statistically represented in a scatter diagram with curve-like progressions, the regression effect is not so easily applicable for forecasts:

For example, coffee promotes concentration, but too much coffee can have the opposite effect, and this can vary greatly from one individual to another in terms of coffee quantity and depend on several individual factors and variations that are not recorded in this way.

In such non-linear events, the range of variation can also have a peak with a tendency away from the center, which can also change over time. Therefore, simple predictions, as they exist for linear correlations, are not possible.

In general, the more complex the correlations are, the less the simple regression assumption applies to the result.

Difference of measured values in studies

Irritation about the significance of the regression effect is particularly frequent in the evaluation of repeated measurements in medical studies when it comes to assessing characteristics over time.

In a simplified approach, the difference between the measured values obtained before and after the intervention (so-called pre-post design) is compared in order to derive a statement on efficacy.

A frequent criticism, especially in homeopathy studies, is that due to the regression effect of the post values, the result would have to be adjusted, since there would already be better results without therapy (non-treatment group). A comparison for the evaluation of efficacy should therefore only be made with a correspondingly adjusted value, otherwise the overall statement of the study would be questionable.

However, this argumentation, which is usually aimed at statistically lowering the effect afterwards, does not take into account that the regression effect also occurs in the opposite direction. After all, patients' disease may worsen due to their natural course of disease, even if the therapy effect is effective.

Thus, a one-sided adjustment of outcome measures, as required, distorts the significance of the result. Here are a few typical considerations and examples:

• Testing of a new treatment method should not be done only on those who have the highest level of suffering at the time of the study. The other extreme group, patients with only low levels of distress, must also be examined for regression effects, since they can also obscure evidence of efficacy in the evaluation of the work.
• The same problem occurs when patients are in different phases of their individual disease despite comparably similar measurements. Patients who present at the apex of an exacerbation at the time of participation in a study may show improvements in follow-up measurements that then statistically show a regression effect, without automatically concluding that this is due to the intervention. It may then rather be the natural course of the disease.
• The same also applies to participants who are in a stage of relative symptom relief at the starting point of a deterioration when they take part in a study and then notice a worsening in the course of the study. These exacerbations after treatment are not automatically attributable to the intervention itself or a lack of efficacy. These can also correspond to the natural course of the disease.

Summary

Regression effects are more significant as a probable bias the more the assessment of treatment effects is a one-sided assessment of a group under extreme conditions.

To what consequences a misinterpretation of the regression effect can lead is impressively illustrated by the example of 'intelligence' and 'socio-economic status' (SOE) presented by Furby (1973). Both characteristics are positively correlated.

Due to the regression effect, lower intelligence is found on average in children of highly intelligent parents. Due to the positive correlation of intelligence and SOE, the children of parents with high SOE are also less intelligent than their parents. Eysenck (1971, as cited in Furby, 1973) interpreted this regression effect as a negative correlation of intelligence and SÖS-determined developmental factors. Accordingly, does high status make children stupid?

If we consider the regression effect in the opposite direction, parents of highly intelligent children are less intelligent than their children:

Since there is a positive correlation between intelligence and SOE, children with above-average intelligence come more often from a home with above-average SOE. Analogous to Eysenck's interpretation, it could be concluded that above-average SES makes children even more intelligent than their parents, which means that SES is an important positive factor for the children's intelligence development. So does high status make children smart? (Nachtigall & Suhl, 2002)

The example also shows that two correlating characteristics can lead to the opposite conclusion, depending on the interpretation. The evaluation error lies in reducing the correlation to a simple linear relationship between the two characteristics, when in fact the relationship is more complex.

Simple pre-post comparisons of fewer measured values, although popular, are not a particularly suitable means of reliably estimating therapeutic effects. Therefore, for more accurate statements on the estimation of regression effects, it is true that

• different control groups are necessary
• the natural course of the disease must be included
• the results of the studied population must not be generalized without further ado,
• conclusions about the effectiveness of the intervention / treatment, require a complex analysis of the entire event in the long-term course.

In summary, the regression effect must be taken into account in experimental design and in statistical analysis, especially of longitudinal studies for the design of control groups. In this sense, it is a statistical quantity.

However, it is at the same time a mistake to assume that in longitudinal studies all measured values tend toward the middle over time, especially when complex, non-linear relationships are involved, such as the assessment of individual disease developments.

This also applies to treatments in everyday practice, which show a significant range of variation in the individual picture and can rarely be reduced to a few parameters, even in acute events.

For this reason, the generally pejorative argument "regression to the mean" towards proofs of efficacy in homeopathic treatments or in homeopathic studies, if it is statistically linearly justified and cited as a directed variable, is based on an error in thinking.

Sources

[Comment by the author:] brief description of the main content of the respective work, its significance and benefit

General works

[1] Beecher, H. K. (1955, December 24). THE POWERFUL PLACEBO. Journal of the American Medical Association, 159(17), 1602.https://doi.org/10.1001/jama.1955.02960340022006

[Comment by the author: The most cited basic work.]

[2] Jütte, R. (2013, April). The early history of the placebo. Complementary Therapies in Medicine, 21(2), 94-97. https://doi.org/10.1016/j.ctim.2012.06.002

[Comment by the author: Summary of all critical points focusing on the incompatibility of pharmacological analyses and psychotherapeutic efficacy.]

[3] Kienle, G. S. & Kiene, H. (1997, December). The powerful placebo effect: fact or fiction? Journal of Clinical Epidemiology, 50(12), 1311-1318. https://doi.org/10.1016/s0895-4356(97)00203-5.

[4] Kienle, G. S. (1995, May 1). Der sogenannte Placebo-Effekt. Illusion, Fakten, Realität. Schattauer, F.K. Verlag

[Comment by the author: Comprehensive critique of the placebo concept, the article describes the argumentation in brief, the book is only available antiquarially (2021).]

[5] Rintelen, H., Kruse, J., Langewitz, W., Söllner, W., Köhle, K., Herzog, W., & Joraschky, P. (2016, November 28). Uexküll, Psychosomatische Medizin: Theoretische Modelle und klinische Praxis (German Edition) (8th ed.). Urban & Fischer.

[Comment by the author: Current summary of Langewitz W. in Current psychosomatic view of placebo concepts.]

[6] Uexküll, T (1994). Preprint from Das Placebo Phänomen. Jahrbuch Kritische Medizin 21 – Arzt-Konsument-Verhältnisse. Argument Verlag. Retrieved 10/12/2022 from http://www.med.uni-magdeburg.de/jkmg/?attachment_id=1791

[Comment by the author: The value of this longer reprint lies in the analysis of the placebo effect from a semiotic point of view, together with a comprehensive list of sources.]

Spezifische Arbeiten

[7] Wei H, Zhou L, Zhang H, Chen J, Lu X, Hu L. The Influence of Expectation on Nondeceptive Placebo and Nocebo Effects. Pain Res Manag. 2018 Mar 19;2018:8459429 https://doi.org/10.1155/2018/8459429

[8] Horsfall L. The Nocebo Effect. SAAD Dig. 2016 Jan;32:55-7. PMID: 27145562.

[9] Evans, C. & Richardson, P. (1988, August). IMPROVED RECOVERY AND REDUCED POSTOPERATIVE STAY AFTER THERAPEUTIC SUGGESTIONS DURING GENERAL ANAESTHESIA. The Lancet, 332(8609), 491–493. https://doi.org/10.1016/s0140-6736(88)90131-6

[10] Pollo, A., Amanzio, M., Arslanian, A., Casadio, C., Maggi, G. & Benedetti, F. (2001, July). Response expectancies in placebo analgesia and their clinical relevance. Pain, 93(1), 77–84. https://doi.org/10.1016/s0304-3959(01)00296-2

[11] Blasi, Z. D., Harkness, E., Ernst, E., Georgiou, A. & Kleijnen, J. (2001, March). Influence of context effects on health outcomes: a systematic review. The Lancet, 357(9258), 757–762. https://doi.org/10.1016/s0140-6736(00)04169-6

[12] Swank, D., Swank-Bordewijk, S., Hop, W., van Erp, W., Janssen, I., Bonjer, H. & Jeekel, J. (2003, April). Laparoscopic adhesiolysis in patients with chronic abdominal pain: a blinded randomised controlled multi-centre trial. The Lancet, 361(9365), 1247–1251. https://doi.org/10.1016/s0140-6736(03)12979-0

[13] Kaptchuk, T. J., Friedlander, E., Kelley, J. M., Sanchez, M. N., Kokkotou, E., Singer, J. P., Kowalczykowski, M., Miller, F. G., Kirsch, I. & Lembo, A. J. (2010, 22. December). Placebos without Deception: A Randomized Controlled Trial in Irritable Bowel Syndrome. PLoS ONE, 5(12), e15591. https://doi.org/10.1371/journal.pone.0015591

[14] Preston, R. A., Materson, B. J., Reda, D. J. & Williams, D. W. (2000, 22. May). Placebo-Associated Blood Pressure Response and Adverse Effects in the Treatment of Hypertension. Archives of Internal Medicine, 160(10), 1449. https://doi.org/10.1001/archinte.160.10.1449

[15] Hróbjartsson, A. & Gøtzsche, P. C. (2001, 24. May). Is the Placebo Powerless? New England Journal of Medicine, 344(21), 1594–1602. https://doi.org/10.1056/nejm200105243442106

[16] Price, D. D., Finniss, D. G. & Benedetti, F. (2008, 1. January). A Comprehensive Review of the Placebo Effect: Recent Advances and Current Thought. Annual Review of Psychology, 59(1), 565–590. https://doi.org/10.1146/annurev.psych.59.113006.095941

[17] Benedetti, F. (2006, May). Placebo analgesia. Neurological Sciences, 27(S2), s100–s102. https://doi.org/10.1007/s10072-006-0580-4

[18] Wechsler, M. E., Kelley, J. M., Boyd, I. O., Dutile, S., Marigowda, G., Kirsch, I., Israel, E. & Kaptchuk, T. J. (2011, 14. July). Active Albuterol or Placebo, Sham Acupuncture, or No Intervention in Asthma. New England Journal of Medicine, 365(2), 119–126. https://doi.org/10.1056/nejmoa1103319

[19] Furby, L. (1973, Februar). Interpreting regression toward the mean in developmental research. Developmental Psychology, 8(2), 172–179. https://doi.org/10.1037/h0034145[21] Nachtigall C., Suhl U. (2002). Der Regressionseffekt Mythos und Wirklichkeit. methevalreport. Psychologische Methodenlehre und Evaluationsforschung am Institut für Psychologie der Friedrich-Schiller-Universität Jena. retrieved 12.10.2022 from https://www.metheval.uni-jena.de/materialien/reports/report_2002_02.pdf

All papers mentioned in this document are listed. The archive contains these and possibly other papers.