By Louis Kuritzky, MD, Clinical Assistant Professor, University of Florida, Gainesville. Dr. Kuritzky is a consultant for GlaxoSmithKline and is on the speaker's bureau of GlaxoSmithKline, 3M, Wyeth-Ayerst, Pfizer, Novartis, Bristol-Myers Squibb, AstraZeneca, Jones Pharma, and Boehringer Ingelheim.
Noninferiority and Randomized Trials
When one reviews the literature it is likely expected that the statistical analyses performed, the complexity of which is often beyond the ken of the practicing clinician, have been appropriately conducted and accurately reflect both the content and intent of the results. Unfortunately, sometimes this is not the case.
Noninferiority trials are usually designed to show that product B, which entered the therapeutic arena after product A, does not fall below a prespecified margin of efficacy compared to product A, in which case product B may be fairly described as "not inferior to product A." This is not the same as saying that product A and B are 'equivalent.' To make that statement, an equivalency trial must be performed, which demonstrates that the efficacy of product B is within prespecified boundaries above and below when compared to product A.
The authors reviewed 116 noninferiority trials and 36 equivalence trials from literature published between 2003-2004. Discouragingly, all but 33 trials had deficits in sample size or its calculation, utilization of confidence intervals, or P value calculation. When subjected to the authors' prespecified criteria for adequacy of study design that further included a requirement that authors justify the margins chosen to determine noninferiority, only 7 trials (4.6%) ultimately were fully satisfactory. Twelve percent of the 33 trials which satisfied overall quality concerns were considered 'highly misleading:' ie, equivalence or non-inferiority was claimed when results were insufficiently robust to do so. The authors suggest that their observations call for greater rigor and uniformity both in design and reporting of noninferiority or equivalence trials.
Henanff AL, et al. JAMA 2006;295:1147-1151.
Thyroid Status, and Cardiovascular Risk
Concern has been raised that subtle perturbations of thyroid function might lead to adverse cardiovascular outcomes. Specifically, subclinical hyperthyroidism has had a reputation of being etiologically involved with a greater incidence of atrial fibrillation. To study the relationship between thyroid dysfunctions and cardiovascular outcomes, investigators from the Cardiovascular Health Study followed subjects (n = 3,233) for new onset atrial fibrillation, coronary heart disease, stroke, cardiovascular mortality, and all-cause mortality, looking for a relationship between thyroid status and outcomes.
Study subjects were comprised of mature, community-dwelling men and women (age > 65). Because one purpose of the study was to elucidate the impact of thyroid dysfunction that was not clinically apparent, subjects already on thyroid treatment at baseline were excluded from analysis. At baseline, 18% of subjects had a clinically occult abnormality of thyroid function, the most common of which was subclinical hypothyroidism (ie, normal T4 with elevated TSH).
Only 1.5% of individuals had subclinical hyperthyroidism (ie, TSH subnormal, with normal T3 and T4). However, the risk of developing atrial fibrillation over the study period was twice as great in persons with subclinical hyperthyroidism as in those without thyroid dysfunction. Similarly, the other thyroid dysfunctions (subclinical and overt hypothyroidism) were not associated with adverse CV outcomes.
Mid-life adults with subclinical hyperthyroidism are at increased risk of atrial fibrillation.
Cappola AR, et al. JAMA. 2006;295:1033-1041.
Glucosamine, Chondroitin Sulfate, and Knee Osteoarthritis
Osteoarthritis (oa) is the most common chronic disabling condition in America. Because there are no disease-modifying drugs to treat OA, intervention relies upon symptomatic relief and improvement of function, which are usually concordant. NSAIDs and Coxibs (eg, celecoxib) have been a mainstay of therapy, but heightened awareness of the GI toxicity of traditional NSAIDs, coupled with publicity about cardiovascular risks of coxibs (ie, rofecoxib, valdecoxib) have tempered enthusiasm for their routine use.
Popular opinion, and a modest number of favorable clinical trials, have encouraged utilization of glucosamine and chondroitin for OA. Their apparent lack of toxicity, in the face of recent reappraisal of NSAID/coxib safety profiles, is additionally reassuring. Indeed, metaanalysis of existing trials has also been supportive. A single, large, randomized controlled trial was devised to help clarify uncertainties remaining after the metaanalysis indicated remaining methodologic uncertainties.
Subjects with knee OA (n = 1583) were randomized to glucosamine, chondroitin (or both), celecoxib, or placebo.
Neither glucosamine nor chondroitin, nor the combination was statistically superior to placebo in the overall analysis. In the subgroup with moderate-to-severe pain at baseline (22% of the total population), the combination of glucosamine and chondroitin was superior to placebo. That the overall population did not benefit from chondroitin or glucosamine, coupled with the fact that amongst responders, response time was quicker with celecoxib, suggests a secondary role, if any, for the glucosamine and chondroitin.
Clegg DO, et al. N Engl J Med. 2006;354:795-808.