Local anesthetics for the treatment of neuropathic pain: on the limits of meta-analysis.
J. Rathmell,J. Ballantyne
DOI: https://doi.org/10.1213/01.ANE.0000184194.99110.80
2005-12-01
Abstract:I s meta-analysis a powerful predictive tool or a deceptive device that leads the average practitioner astray? The answer depends on the contributing studies, the methodology, and the interpretation of both contributing data and analytic results. When high quality, homogeneous trials are combined using sound methodology, meta-analysis can be a powerful predictive tool (1). When high-quality homogeneous trials are not available, as is often the case in the pain literature, the validity of a meta-analysis may depend more on the analysts’ interpretation of available data than on the strength of the methodology (2–5). In this issue of Anesthesia & Analgesia, Tremont-Lukats et al. (6) present a systematic review and meta-analysis examining the effectiveness of systemic administration of local anesthetics for relief of neuropathic pain. The methodology used by these authors is flawless, but have the analyses contributed new and clinically useful information? Research begins with an observation about possible cause and effect; the observation leads to formulation of a hypothesis that then becomes the basis of the research question. Meta-analysis should be no different from other forms of research, i.e., a question should be posed, and then the data should be sought to fit the question; the question should not be manipulated to fit the available data. Yet, when the available data are scanty, the temptation to manipulate the question is great. One way to ensure a result (and a publication) is to combine studies with unreasonable heterogeneity. Random effects statistical models used in meta-analysis are designed to account for heterogeneity, but they cannot account for excessive heterogeneity (7–9). When excessively heterogeneous trials are combined in a meta-analysis, complex issues tend to be oversimplified, and conclusions are misleading. The classic example in the anesthesia literature is the meta-analysis by Rodgers et al. (10) that finds a 30% reduction in mortality overall when neuraxial anesthesia is used during surgery. The conclusion of these authors is that the findings “support more widespread use of neuraxial blockade anesthesia.” Yet reduced mortality is found in only two surgical subsets, and the authors’ conclusion is grossly misleading (9). In this case, the question “Is neuraxial anesthesia always better than no neuraxial anesthesia?” was tailored to fit the vast amount of available data. Thus, it was not a reasonable question. In the present meta-analysis, the question “Is systemic administration of local anesthetic effective for the relief of neuropathic pain?” seems reasonable, but what does the question really entail? The effectiveness of an analgesic is not simply a matter of whether the analgesic can alter a pain score to a statistically significant extent. More important is whether the analgesia is clinically meaningful and whether the side effects are tolerable. The analyses of pain scores presented by Tremont-Lukats et al. (6) convincingly demonstrate that both short-term infusions of IV lidocaine and longer-term oral mexiletine administration produce a small improvement in pain (11 mm on a 0–100 mm scale) compared with placebo. Lidocaine, tocainide, and mexiletine were no better than other analgesics in 206 patients. The authors recognize that there is significant heterogeneity; in particular, differences in pain causes are an important source of clinical and statistical heterogeneity. They describe several subset analyses that highlight sources of heterogeneity. They also discuss the more difficult question of whether a mean difference of 11 mm represents a true clinical difference for patients. Although the comparisons between systemic local anesthetics and other analgesics may not have been worth combining in a metaanalysis, one could not wish for a better culling of the literature, better analyses, or a better interpretation of the data and analyses with regard to the analgesic Accepted for publication June 15, 2005. Address correspondence and reprint requests to James P. Rathmell, MD, Center for Pain Medicine, 62 Tilley Dr., South Burlington, Vermont 05403. Address e-mail to james.rathmell@uvm.edu.