统计问题101问: 荟萃分析问题2

Question

Last week’s question described a meta-analysis of the effectiveness of parenteral corticosteroids for the relief of acute severe migraine headache in adults. Seven randomised controlled trials were identified in which single dose parenteral dexamethasone, administered alone or in combination with standard abortive therapy, was compared with placebo or any other standard treatment for acute migraine in adults. For each trial, the relative risk for recurrence of headache within 72 hours was obtained for the dexamethasone treatment arm compared with the placebo arm.

The results of the meta-analysis were presented in a forest plot (figure⇓). When combined with standard abortive therapy, single dose parenteral dexamethasone was more effective than placebo in reducing the recurrence of acute severe migraine headache in adults within 72 hours (relative risk 0.74, 95% confidence interval 0.60 to 0.90).

• Forest plot of the effectiveness of dexamethasone in preventing the recurrence of acute severe migraine headache in adults compared with placebo.

Which of the following statements, if any, are true?

·a) The vertical line in the centre of the forest plot represents the line of no effect.

·b) Each trial contributed equally to the total relative risk.

·c) A relative risk greater than one indicates increased risk of recurrence of headache within 72 hours if taking placebo compared with dexamethasone.

·d) For each trial, the relative risk for recurrence of acute severe migraine headache within 72 hours was significantly different from unity at the 5% level of significance.

·e) The meta-analysis showed an overall 26% reduction in recurrence of acute severe migraine headache within 72 hours for dexamethasone compared with placebo.

提示：这是一道多选题。

Answer

Answers a and e are true, whereas b, c, and d are false.

On the forest plot, the seven trials included in the meta-analysis are identified by the principal author on the left hand side. The sample relative risk and associated 95% confidence interval for each trial are shown on the right and are represented graphically in the centre. For each trial, the sample estimate is represented by a square and its associated 95% confidence interval by the horizontal line. The size of each square is directly proportional to the sample size of the study. The total overall effect is represented by a diamond; the centre of the diamond equals the average effect whereas the extremes equal the limits of the confidence interval. The vertical line in the centre of the graph is the “line of no effect”—that is, a relative risk of 1.0, which represents no difference in risks between dexamethasone and placebo (a is true).

The total overall estimate of the population relative risk was obtained by pooling the relative risks from the seven trials. The seven trials did not contribute equally to the pooled result (b is false)—the total estimate was not an average of the individual estimates. The contribution of each trial is indicated under the heading “Weight (%).” The percentage weight contributed by a trial is determined by the precision of its sample estimate of the population parameter, and trials with more precise estimates, typically those with larger samples, contributed more.

For the dexamethasone and placebo groups in each trial, the numbers of patients experiencing a recurrence of headache within 72 hours and the total group size are shown. It is standard practice to present a forest plot with the new or active treatment on the left and control or placebo group to the right. The presented relative risks are the risk of recurrence of acute severe migraine headache in adults for dexamethasone compared with placebo. Therefore, a relative risk greater than unity indicates increased risk of recurrence for dexamethasone relative to placebo (c is false), with the result favouring placebo, whereas a relative risk less than unity indicates a reduced risk, with the result favouring dexamethasone.

The vertical line on the forest plot represents a relative risk of 1.0—that is, a risk of recurrence of headache within 72 hours for dexamethasone treatment equal to that of placebo. As described in previous questions, if the 95% confidence interval for the population relative risk does not include unity (1.0), the risks of recurrence of headache within 72 hours for dexamethasone and placebo are significantly different from each other at the 5% level of significance. For all trials, except that reported by Innes (1999), the 95% confidence interval included unity so the sample relative risk of recurrence for dexamethasone compared with placebo was not significantly different from unity (d is false).

A 95% confidence interval is asymmetrical about the sample relative risk. The graphic display of the relative risks and confidence intervals were, however, plotted on a logarithmic scale in the forest plot so the confidence intervals appear to be symmetrical around the relative risk.

The total overall relative risk was 0.74 (95% confidence interval 0.60 to 0.90). Relative risks less than unity are subtracted from 1.0 and the resulting value interpreted as a reduced risk. Therefore, a 26% relative reduction in recurrence of acute severe migraine headache within 72 hours was demonstrated for dexamethasone relative to placebo (e is true). The results of the statistical test of the overall relative risk are displayed against “Test for overall effect.” The P value for the test of the statistical null and alternative hypotheses, described in a previous quiz, was reported as P=0.003. Because P was less than 0.05, the risk of recurrence of headache within 72 hours for dexamethasone was significantly different from that for placebo at the 5% level, which is confirmed by the associated 95% confidence interval not including unity.

The sample estimates of the population relative risk were combined to give a single overall estimate, thereby reducing a large amount of information to a manageable quantity. The total relative risk is a more precise estimate of the population relative risk than any of the individual sample estimates and can be generalised to a wider patient group. Meta-analyses have become integral to evidence based medicine, whereby the results of the best and most comprehensively available evidence, together with clinical expertise, are used in patient care.

所以答案是选择 a e

每天学习一点，你会更强大！