27/02/2012 Systematic Reviews Statistical tests for heterogeneity Professor Declan Devane Chair of Midwifery National University of Ireland Galway Email: [email protected] Assessing statistical heterogeneity • If there is substantial heterogeneity among studies in a systematic review, it might be inappropriate to do a meta-analysis. • How do we know if there is ‘substantial’ heterogeneity? 1. Visual inspection of a forest plot of studies included in the review; 2. Assessment of results of tests for statistical heterogeneity. 1 27/02/2012 Statistical tests for heterogeneity • Cochran Q (Chi-square, X2) • I2 • Tau2 Cochran Q (Chi-square, x2) • It assesses whether observed differences in results are compa7ble with chance alone. • If the Cochran X2 (Cochran Q) is sta7s7cally significant (usually p<0.1) there is definite heterogeneity. – The level of significance for X2 (Cochrane Q) is oGen set at 0.1 due to the low power of the test to detect heterogeneity • Rule of thumb: if X2 is bigger than df (which is 1 less than number of studies in forest plot) then heterogeneity is present 2 27/02/2012 Study or Subgroup Bowsher 1997 Kieburtz 1998 Lampl 2002 Leijon 1989 Max 1988 Pilowsky 1982 Shlay 1998 Turkington 1980 Vrethem 1997 Watson 1982 Amitriptyline Placebo Events Total Events Total Weight 32 38 22 34 14.6% 23 34 9 41 12.4% 17 20 16 19 14.6% 10 15 1 15 4.7% 16 34 4 25 9.7% 4 12 3 12 7.7% 31 71 28 65 14.0% 19 19 0 20 2.7% 24 33 8 33 12.2% 16 24 2 24 7.2% Risk Ratio M-H, Random, 95% CI 1.30 [0.98, 1.73] 3.08 [1.65, 5.74] 1.01 [0.77, 1.32] 10.00 [1.46, 68.69] 2.94 [1.12, 7.73] 1.33 [0.38, 4.72] 1.01 [0.69, 1.49] 40.95 [2.65, 633.88] 3.00 [1.58, 5.68] 8.00 [2.06, 31.07] Total (95% CI) 300 288 100.0% Total events 192 93 Heterogeneity: Tau² = 0.43; Chi² = 59.65, df = 9 (P < 0.00001); I² = 85% Test for overall effect: Z = 3.14 (P = 0.002) Risk Ratio M-H, Random, 95% CI 2.23 [1.35, 3.69] 0.01 0.1 1 10 100 Favours Placebo Favours Amitriptyline Cochran Q (Chi-square) • The X2 (Cochran Q) test has low power to detect heterogeneity if there are few studies in the meta-‐analysis and may, conversely, give a highly significant result if it comprises many large studies, even when the heterogeneity is unlikely to affect the conclusions. 3 27/02/2012 I2 • An Index, I2, which does not depend upon the number of studies, the type of outcome data or the choice of treatment effect (e.g. rela7ve risk) can be used to quan7fy the impact of heterogeneity and assess inconsistency. • I2 = 100x(Q – df)/Q I2 • I2 represents the percentage of the total varia7on across studies due to heterogeneity; it takes values from 0% to 100%, with the value of 0% indica7ng no observed heterogeneity. If there is evidence of sta7s7cal heterogeneity, we should proceed cau7ously, inves7gate the reasons for its presence and modify our approach accordingly, perhaps by dividing the studies into subgroups of those with similar characteris7cs. 4 27/02/2012 Study or Subgroup Bowsher 1997 Kieburtz 1998 Lampl 2002 Leijon 1989 Max 1988 Pilowsky 1982 Shlay 1998 Turkington 1980 Vrethem 1997 Watson 1982 Amitriptyline Placebo Events Total Events Total Weight 32 38 22 34 14.6% 23 34 9 41 12.4% 17 20 16 19 14.6% 10 15 1 15 4.7% 16 34 4 25 9.7% 4 12 3 12 7.7% 31 71 28 65 14.0% 19 19 0 20 2.7% 24 33 8 33 12.2% 16 24 2 24 7.2% Risk Ratio M-H, Random, 95% CI 1.30 [0.98, 1.73] 3.08 [1.65, 5.74] 1.01 [0.77, 1.32] 10.00 [1.46, 68.69] 2.94 [1.12, 7.73] 1.33 [0.38, 4.72] 1.01 [0.69, 1.49] 40.95 [2.65, 633.88] 3.00 [1.58, 5.68] 8.00 [2.06, 31.07] Total (95% CI) 300 288 100.0% Total events 192 93 Heterogeneity: Tau² = 0.43; Chi² = 59.65, df = 9 (P < 0.00001); I² = 85% Test for overall effect: Z = 3.14 (P = 0.002) Risk Ratio M-H, Random, 95% CI 2.23 [1.35, 3.69] 0.01 0.1 1 10 100 Favours Placebo Favours Amitriptyline I2 • Thresholds for the interpretation of I2 can be misleading, since the importance of inconsistency depends on several factors. A rough guide to interpretation is as follows: – 0% to 40%: might not be important; – 30% to 60%: may represent moderate heterogeneity; – 50% to 90%: may represent substantial heterogeneity; – 75% to 100%: considerable heterogeneity*. 5 27/02/2012 Tau2 • aA estimate of the between-study variance in a random-effects meta-analysis is given as (known as tau-squared (T2 or Tau2)). • If > 1, suggests presence of substantial statistical heterogeneity. Study or Subgroup Bowsher 1997 Kieburtz 1998 Lampl 2002 Leijon 1989 Max 1988 Pilowsky 1982 Shlay 1998 Turkington 1980 Vrethem 1997 Watson 1982 Amitriptyline Placebo Events Total Events Total Weight 32 38 22 34 14.6% 23 34 9 41 12.4% 17 20 16 19 14.6% 10 15 1 15 4.7% 16 34 4 25 9.7% 4 12 3 12 7.7% 31 71 28 65 14.0% 19 19 0 20 2.7% 24 33 8 33 12.2% 16 24 2 24 7.2% Risk Ratio M-H, Random, 95% CI 1.30 [0.98, 1.73] 3.08 [1.65, 5.74] 1.01 [0.77, 1.32] 10.00 [1.46, 68.69] 2.94 [1.12, 7.73] 1.33 [0.38, 4.72] 1.01 [0.69, 1.49] 40.95 [2.65, 633.88] 3.00 [1.58, 5.68] 8.00 [2.06, 31.07] Total (95% CI) 300 288 100.0% Total events 192 93 Heterogeneity: Tau² = 0.43; Chi² = 59.65, df = 9 (P < 0.00001); I² = 85% Test for overall effect: Z = 3.14 (P = 0.002) Risk Ratio M-H, Random, 95% CI 2.23 [1.35, 3.69] 0.01 0.1 1 10 100 Favours Placebo Favours Amitriptyline 6 27/02/2012 7 27/02/2012 8 27/02/2012 Thank you Professor Declan Devane Chair of Midwifery National University of Ireland Galway Email: [email protected] 9
© Copyright 2026 Paperzz