OUP user menu

When Data Are Not Proof

Nikolai Bogduk MD, PhD, DSC
DOI: http://dx.doi.org/10.1111/j.1526-4637.2011.01216.x 1297-1298 First published online: 1 September 2011

In this issue, Prochazka, Ales, and Prachazkova reported their experience with intrathecal administration of midazolam for failed back surgery syndrome [1]. During the review process, there were disputes about this article between the referees and the authors. The referees welcomed the data but not the conclusions. The authors contend that they provide evidence of effectiveness of this intervention. In their correspondence, they cited other studies that also claimed effectiveness. However, the authors misunderstood the concerns of the referees. Post hoc, propter hoc is a common fallacy and illusion in clinical practice. Just because a patient improved after a treatment does not mean that they improved because of the purported effects of the treatment. In the present case, the authors contend that the improvements must have been due to the midazolam. Quoting others who make the same claim is not an evidence that midazolam is effective. Doing so only refers to others who suffer the same illusion.

For a drug to be regarded as effective, we need data on its attributable effect, i.e., how more often is the drug effective than nonspecific effects. It is palpably true as the authors have shown that patients improve after intrathecal administration of midazolam, but the data do not prove that the response was due—exclusively or in part—to the midazolam. The competing hypothesis is that the patients improved because they were cared for in a good manner, using a high-tech treatment. It may be that they responded to components of the treatment other than the midazolam. In other words, they responded because they had an injection, irrespective of what was injected. The history of interventional pain medicine is replete with examples of injections that work but not because of what was injected.

Patients with complex regional pain syndrome respond to sympathetic blocks, but equally as often, they respond to placebo blocks [2–8]. Patients respond to intra-articular injections of steroids into cervical or lumbar zygapophysial joints, but equally as often, they respond to injections of local anesthetic [9] or saline [10]. Patients respond to epidural injections of steroids but also to injections of saline into an interspinous ligament [11]. Patients respond to prolotherapy with dextrose but also to injections of normal saline [12]. These studies imply that the act of injection is far more significant than the agent injected and warn us to expect the same for new injections in the future.

These philosophical considerations limit the conclusions that might be drawn from the data of Prochazka, Ales, and Prachazkova [1]. However, they do not affect the value of their data. The merits of the study are that the authors report a large number of patients, which allows the properties of the intervention to be reliably quantified. We can see that intrathecal midazolam is not always effective. We see the reality that in practice, as well as in research, some patients go missing. Nevertheless, in those patients who do respond, we can see that the median survival time of the response is 5.5 weeks. As well, some 30% of patients respond for at least 12 weeks, and 10% respond for 24 weeks. These data allow us to make practical, moral, and scientific judgments.

Relief for 6 weeks, or even 12 weeks, is not a cure, and some would wonder if it is worth administering a therapy the response to which is so short-lived. In contrast, however, others could argue that there is precious little that can be done for patients with failed back surgery syndrome and therefore, that it is worth knowing that intrathecal midazolam can be drawn from a “bag of tricks” in the palliative management of these patients with a difficult problem.

Reassuring in this regard is the fact that intrathecal midazolam seems quite safe. So patients are not being exposed to risks if they choose to undergo a palliative treatment with short-lived effect.

However, the data also show that a proportion of patients can expect long-lasting effects. For a simple, safe, and inexpensive treatment, it could be worthwhile and cost-effective to offer the treatment just to harvest this modest yield of long-term responses.

Scientifically and morally, however, this is not an adequate end point. Scientifically, we do not know if the long-term responses are due to the drug or are a placebo effect. Pragmatically, one could argue that the effects on the patient are good and worth having even if they are a placebo, but the question then becomes: “do we need to add the midazolam?” We can harvest the placebo response with an injection of a less expensive agent with less risk of side effects. All that is required is to perform the injection in a convincing manner.

Proponents of contentious therapies might argue that it is unethical to use placebo as a treatment. I would respond that it is less ethical to use a costly, proprietary agent simply to create the illusion of an active treatment. If studies show that saline is no less effective, but safer, than the proprietary agent, the administration of saline is promoted to evidence-based practice.

I am not against harvesting a good outcome for patients, but I am hostile to wrongful claims about how and why this treatment works.


View Abstract