For many things, you need 15 minutes of physician time, some de riguer labs to be drawn just to reasonably rule out other problems (in Western medicine, treatments fit the disease or condition, while in alternative medicine, each intervention cures almost everything so differential diagnosis is not much of a problem), and so on. CAM circumvents a lot of these costs. The interventions might be grouped into three groups: amazingly inexpensive (aromatherapy, acai berry juice); modest cost (some $30 per month ‘supplement’ to be taken regularly, or regular acupuncture sessions roughly equivalent to having a monthly physical trainer visit); and expensive (exotic chelation regimens, laetrile in Mexico, or a treatment episode at Dave Wakefield’s special Austin hide-away).

The efficacy measure matters. If the outcome is ‘energy level,’ ‘flexibility,’ ‘vitality,’ ‘quality of life,’ or ‘pain,’ then you will get at least a minimal amount of needle movement from placebo effects, nonspecific therapeutic factors, and so on.

This set-up, of a ratio compared to a ratio, can make a weakly efficacious intervention strongly favored by NICE. Efficacy often has only so much range across which to vary, but costs can go from a few dollars to tens of thousands.

Now that I make this comment, it occurs to me that a ‘correction factor’ could be thrown in.

Structural equation modeling suffers from this kind of set-up: what test statistic do you use? they are very sensitive to these contingencies. So, a lot of the SEM fit indices (CFI, NFI, NNFI, etc.), incorporate aspects to adjust for various aspects such as model parsimony (fewer predictors gives you less total explained variance but has the benefits of parsimony).

MAybe it is time to add some correction factor to the ICER. Maybe a threshold factor that is close to zero if the intervention as normally delivered does not deliver a clinically meaningful difference, and a graded factor multiplying clinically meaningful improvement.

Others may want reassurance that weakly statistically effecticacious, but possibly expensive and side-effect prone pharmaceuticals , or about equally weakly validated physiotherapeutic procedures such as massage, themselves rarely rigorously tested for cost-effectiveness, should always have first demand on their pocket or their pooled funds, rather than CAM-based programs that do look to be as cost-effective or better in pragmatic studies.

The oft asked question “what does it matter how it works?” is an entirely valid perspective for many. One of the burdens that modern medical science bears is, surely, advising on how tightening medical resources can be applied in the most cost-effective way. If we do not look to be taking that responsibility seriously others will take it off us.

I warned a few years ago that this was going to be the next major challenge for medical skepticism, and it is disappointing that one of the first attempts to “debunk” cost-effectiveness studies of CAM has elicited so little discussion.

If I tried to sell a “complementary” helicopter, or an “all natural” nail gun (aka ‘a rock’, presumably), nobody would buy it.

An “organic” Wurlitzer, geddit?
.

The authors admit (and I agree) that search strategies for CAM in the published medical literature are difficult.

It is difficult because the only reasonable definition is “that which is proven not to work or has not been evaluated” and nobody wants to do a study on “something proven not to work.” When I get into discussions about it with fellow med students and the occasional physician, I often start out by asking exactly that. To define what CAM is and how you can differentiate it from medicine that we practice. Inevitably they say “stuff like exercise, diet, and natural things” and I demolish that by citing the very scientific basis of diet and exercise and then pharmacognosy and the naturalistic fallacy. I’ve had many a person be surprised that they simply cannot define CAM in any meaningful way except that which reasonably precludes its use in medicine. Sometimes I get “well, what’s the harm” as a rebuttal. But that is a whole different story.