Statistical Significance Versus Practical Significance

Editorial tools are used. In drug therapy management, we need the operators (pharmacists) of the tools to embrace clinical practice guidelines and the principles of continuous quality improvement. Compensating pharmacists at the lowest possible price does not foster this attention and commitment. Farris, Kumbera, Halterman, and Fang in this issue of the Journal, describe a proprietary program developed around the concept of compensating pharmacists based upon measurable patient outcomes.37 This is the holy grail of managed care pharmacy, compensation for clinical pharmacists based upon valued outcomes. ■■ Statistical Significance Versus Practical Significance Researchers know the value of sample size in demonstrating statistical significance between groups. Small sample size requires a large absolute difference between groups to demonstrate statistical significance. For example, the data used to obtain FDA approval of pravastatin for the indication of stroke involved a study of 9,014 patients with a history of myocardial infarction or unstable angina and total cholesterol levels of 155 to 271 mg per deciliter. The study found that the risk of stroke was 3.7% among patients given pravastatin versus 4.5% among patients given placebo over a follow-up period of 6 years, an absolute difference of 0.8% and relative risk reduction for stroke of 19%.38 Subsequent scrutiny of the study findings and the literature citations identified flaws. Of the 4 studies cited by the authors to support the reported reductions of 25% to 30% in the rate of stroke, the West of Scotland trial showed no reduction in the rate of stroke—the risk after 5 years of follow-up was reported as 1.6 percent for both the pravastatin and placebo groups (P=0.57), and a second cited study (the Air Force-Texas Coronary Atherosclerosis Prevention Study, or AFCAPS/ TexCAPS) did not list stroke as an individual end point. More to the point here, the 0.05 statistical difference reported by the authors using univariate analysis for pravastatin versus placebo was actually 0.10 upon application of multivariate analysis, arguably the more appropriate measure.39 Aside from the statistical debate, in which a small absolute difference may or may not be statistically “significant,” there is the very important matter of practical significance. The study results suggested that it would require 750 patients to be treated for one year to prevent one nonfatal stroke, or more than $750,000 in discounted drug cost, an unfavorable cost-effectiveness outcome and a very unfavorable cost outcome, compared to the risk reduction possible with aspirin.40 Previous research and a meta-analysis of clinical studies found no association between cholesterol levels and risk of stroke. In this issue of the Journal, Arocho, Solis, Wade, Goldberg, and Tang discuss the very important subject of underlying characteristics of groups that might explain differences in the dependent measure found in study results.41 But within their analysis is the ever-present matter of practical versus statistical significance. As part of their findings of underlying differences among calcium channel blocker patients, they report 86% of amlodipine users had at least one professional service, compared to 84% of felodipine users (P<0.05), in a study involving 17,667 patients. They also report a statistically significant (P=0.0007) difference between average daily dose, 5.8 mg for amlodipine patients versus 5.93 mg for felodipine patients. They report average age but not median age and do not address the fact that 39.2% of amlodipine patients in their study were less than age 70 versus 33.9% of felodipine patients less than age 70. Readers should note that this study was funded by the manufacturer of amlodipine. Managed care pharmacists would be interested in the formulary status of these two drugs at the time that the drug and medical claims were incurred in the health plans employed in the study, a fact not disclosed by the authors. Nevertheless, the authors are correct in highlighting the importance of managed care pharmacists investigating the underlying differences in patient populations, particularly in drug-todrug comparisons. ■■ Prevalence and Costs of Atopic Dermatitis Atopic dermatitis (AD), or eczema, is a chronic, relapsing inflammatory skin disease affecting about 10% of the Western population, with a prevalence estimate of 4% to 20%.42 The high range of the prevalence estimate is in children, and 80% to 90% of AD cases are diagnosed by age 5.43 The disease is physically distressing to the patient and often has an adverse effect on the quality of life.44 Fivenson, Arnold, Kaniecki, Cohen, Frech, and Finlay, in this issue of the Journal, provide estimates of some of the direct and indirect costs of AD among patients identified in one health plan, expressed in 1997 dollars.45 Readers might note that about two thirds of the patients were cared for only by a dermatologist, suggesting that these patients were the more severe cases of AD in that population. Combined with the small prevalence, approximately 1.3% of plan members, the data suggest that this patient population is more representative of the moderate-to-severe population of AD patients. Yet, only 3 patients were rated as “severe” by provider-assessment. Nearly two thirds of the patients in this cohort were younger than 16 years, and 19.5% were younger than 4 years. There was one inpatient visit and no emergency room visits among 3,576 patientmonths in the study. The only inpatient visit occurred in a patient with a “mild” case of AD, according to provider-assessed severity. ■■ Bias Management—The Path to Better Evidence In this issue and on the Journal Web site (www.amcp.org) is a copy of the editorial policy, disclosure statement, and attestation required of all authors of articles published in JMCP, effective with the January/February 2002 issue. The detail is designed to obtain all disclosures necessary to permit readers to evaluate potential biases that may affect the objectives of the research, interpretation of data, conclusions, or other findings and to attribute to each listed author the specific contribution to the article. Frederic R. Curtiss, PhD, RPh, CEBS, Editor-in-Chief 404 Journal of Managed Care Pharmacy JMCP September/October 2002 Vol. 8, No. 5 www.amcp.org  (...truncated)