Classical homeopathy is scientifically implausible as a therapy, because there is no substance of any medicinal value left in the functionally-infinitely diluted nostrum. Naturally, there is no hard evidence supporting the therapeutic use of homeopathy, in terms of clinical benefit to the patient. Absent such support, homeopathy-peddlers generally push affordability and low cost as homeopathy’s unique selling point (USP). A large retrospective cost-analysis study, based on nearly 45000 individual German patients, gives lie to that myth.
In a paper published last week in PLOS ONE, authors from the Charité University in Berlin – Julia Ostermann, Thomas Reinhold, and Claudia Witt – analyzed healthcare costs for patients choosing homeopathy in addition to their medicines, in comparison to those who stuck with only medicines, for their treatments. De-identified, pseudonymized data on patient claims and costs, as well as partial funding for the study, were provided by a German statutory health insurance company. From the abstract:
The aim of this study was to compare the health care costs for patients using additional homeopathic treatment (homeopathy group) with the costs for those receiving usual care (control group) […] Patients in both groups were matched using a propensity score matching procedure based on socio-demographic variables as well as costs, number of hospital stays and sick leave days in the previous 12 months. Total cumulative costs over 18 months were compared between the groups with an analysis of covariance (adjusted for baseline costs) across diagnoses and for six specific diagnoses (depression, migraine, allergic rhinitis, asthma, atopic dermatitis, and headache).
One of the reasons the study caught my attention was the presence of Claudia Witt as one of the authors – in fact, one of the two who conceived and designed the study. Critics of alternative medicine may be well-acquainted with Witt, a physician trained and licensed in Germany, who holds (and has held in the past) various professional appointments in complementary and alternative medicine (fashionably referred to as “Integrative Medicine”) in the University of Zurich, Charité University in Berlin, and the University of Maryland School of Medicine in Baltimore. She is highly invested into popularizing altmed modalities, including homeopathy and acupuncture. (N.B. For a masterful, science-based rebuttal of the linked acupuncture study, read this post by David Gorski in Science-based Medicine.)
I am glad that the authors followed the data to where it went in presenting the results. There is some hand-waving in the discussion and study interpretation section, but I’ll come to that shortly. All the 44550 subjects were insured through the German health-insurer Techniker Krankenkasse (TK). Regardless of their treatment-modality use at the time of inclusion, the defined ‘patients’ in the study opted for homeopathy available to them under the integrated care contract (ICC) model of German healthcare; defined ‘controls’ were TK-insured patients who didn’t subscribe to the ICC model (leading to one caveat in the design, as authors mention: controls could have been treated by homeopaths external to the ICC). Because in Germany the ICC is technically available to every insured person, randomization of cases and controls was not feasible; instead, the authors chose to match homeopathy users and controls using a propensity score calculated based on multiple covariates.
The study was designed to compare overall healthcare costs across diagnoses between the groups, and in addition, looked at data from specific confirmed diagnoses of depression, migraine, tension headache, atopic dermatitis, allergic asthma and rhinitis. Data were analyzed for a total of 30 continuous months, 12 months before study inclusion for the baseline and 18 months after. Consideration and calculation of overall healthcare costs included costs of outpatient care (as charged by a homeopath or a physician), medication, inpatient stay, productivity loss (including sick leaves as a reason), contractual costs of the ICC, among others.
Out of over 4 million patients who satisfied the inclusion criteria, data from 44550 patients (22275 matched case/control pairs) from a wide geographical area were analyzed. The authors found that (emphasis in boldface, mine):
- Adjusted cumulative total costs were significantly higher in the homeopathy group 18 months after the study onset, compared to controls.
- Higher costs in the homeopathy group were primarily driven by productivity loss and outpatient care costs; outpatient care was costlier in homeopathy group.
- Over 18 months, homeopathy users took on an average 19 sick days, as opposed to 16 in the control group; for people with clinical depression, on an average homeopathy users took 48 sick days (whereas controls took about 42), required significantly more frequent hospital stays, and incurred about 18% more in adjusted cumulative total expenses.
- Higher expenses in the homeopathy group was a trend seen across almost all the different specific diagnoses.
- The largest cost difference was seen in the initial 1-3 months, attributable to costs for initial homeopathic consultations.
- Although the mean number of prescriptions were lower (10.4) in the homeopathy group versus the controls (10.7), overall costs of medication – as well as the mean number of diagnoses – after 18 months were significantly higher in the homeopathy group, compared to controls. And this is even after group differences decreased over time.
So, homeopathy is affordable and low cost… NOT, as per the observed numbers. So, is there a way to spin the data in an altmed-friendly way? Despite some attempted hand-waving by the authors, it doesn’t appear so.
The authors correctly point out that this study, based on claims and cost data, did not allow them to evaluate the effectiveness of the ICC homeopathy, or infer the severity of disease in individual patients. However, they also state the concept that “health care costs can serve as a proxy for a medical condition or for the severity of a disease” – perhaps implying that homeopathy patients were a priori sicker at entry, or got sicker as the study continued? I don’t know; in the Interpretation section, the authors cite a reference that says many patients seeks homeopathy are chronically ill. However, in this study, their propensity score matching system at entry resulted in a fairly homogeneous and comparable pair of groups based on covariates tested. So I don’t understand the significance of this remark in this context. Besides, they do have month-to-month cost data for the approximately 4 out of every 10 patients who matched according to a confirmed diagnosis. Surely analyzing that data for specific conditions could provide an idea about the severity of individual conditions? I didn’t see such an analysis in the main paper or the supplementary data.
The authors also remark: “… approximately two-thirds of our sample comprised female participants, reflecting the current literature on predictions for complementary medicine usage” – but exact same proportions were included in the control group, too. So how is this observation relevant in this context?
The authors lament that they could not assess the cost of treatment over longer times, since the homeopathy costs seemed to come down closer to the level of non-homeopathy control group at 18 months – although there is no snapshot statistical analysis provided for at-18-month end-of-study costs (a data point I’d have loved to see). I do think that an important parameter to elaborate would have been the group attrition rate at 18 months; that is, if patients didn’t get relief from additional homeopathy at the end of the study period, did they stop using it? Within the demographic of the participants, what were the factors that kept the patients adherent to homeopathy if they were not getting benefit from it? The authors do touch upon this scenario in Interpretation in order to explain the observations, but not to any clarity.
This question is made murkier by the fact that in Germany, nearly 2% of German physicians are homeopaths – which means, there are homeopaths who were conventionally trained in medicine, and therefore, are allowed to prescribe real medicine. About 1 in 4 in the homeopathy group in this study received their medication from homeopaths. The authors use this situation as a possible reason for driving up overall healthcare costs for homeopathy patients, indicating that a conventionally trained homeopath could potentially diagnose undeclared or hidden problems in a homeopathy-seeking patient and urge them to get proper medical care, which is likely costlier. Wouldn’t a more likely and simpler explanation be that in these patients, homeopathy didn’t work and the homeopath, having real medical training, had the good sense to prescribe real medicine? [In my infancy, I was treated by such a physician, our family doctor who practised both homeopathy and “allopathy” (a nonsensical term for real medicine favored by homeopaths) as and when needed – which likely allowed him to cater to a large clientèle. Hmmm… That wily and charismatic old goat!]
The other prior work in this area (that the authors cite in Interpretation) seem rather short on methodological validity and homogeneity of observations – as the authors themselves point out. In that lore, this work will hopefully provide some definitive evidence that the usual homeopathy-associated claim of lower costs is, at best, wishful thinking – on part of the homeopaths. In a 2012 essay, Scott Gavura, a trained pharmacist, wrote a scathing critique of the concept of “cost-effectiveness” as applied to ineffectual altmed modalities most commonly by its ardent proponents; today, in the light of the described study, that essay is well-worth a revisit.