{"id":98,"date":"2021-05-22T23:40:36","date_gmt":"2021-05-23T03:40:36","guid":{"rendered":"https:\/\/pressbooks.bccampus.ca\/rickyturgeon\/?post_type=chapter&p=98"},"modified":"2022-01-13T12:46:23","modified_gmt":"2022-01-13T17:46:23","slug":"results-of-the-meta-analysis-i-e-what-do-the-pooled-results-of-the-trials-show","status":"publish","type":"chapter","link":"https:\/\/pressbooks.bccampus.ca\/rickyturgeon\/chapter\/results-of-the-meta-analysis-i-e-what-do-the-pooled-results-of-the-trials-show\/","title":{"raw":"Results of the meta-analysis","rendered":"Results of the meta-analysis"},"content":{"raw":"As with [pb_glossary id=\"704\"]RCTs[\/pb_glossary], outcomes ought to be interpreted beyond just statistical significance to assess the magnitude of effect and clinical relevance. Interpretation also requires considerations beyond what is necessary when appraising [pb_glossary id=\"704\"]RCTs[\/pb_glossary]. It is also important to consider how many trials reported on a particular outcome, and what the quality of those specific trials were. Additionally, even if the trials are otherwise clinically and methodologically similar, statistical [pb_glossary id=\"107\"]heterogeneity[\/pb_glossary] identified by visual inspection and\/or formal statistical testing may preclude confidently combining trial results.\r\n

Checklist Questions<\/h1>\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
I2<\/sup> Value - What was the statistical [pb_glossary id=\"107\"]heterogeneity[\/pb_glossary]?<\/td>\r\n<\/tr>\r\n
Appropriate to pool the results & interpret the summary statistics?<\/td>\r\n<\/tr>\r\n
Fixed-effects or random-effects?<\/td>\r\n<\/tr>\r\n
Is the model used appropriate?<\/td>\r\n<\/tr>\r\n
Which effect measure was used? (e.g. [pb_glossary id=\"103\"]OR[\/pb_glossary], [pb_glossary id=\"109\"]RR[\/pb_glossary], [pb_glossary id=\"113\"]SMD[\/pb_glossary])<\/td>\r\n<\/tr>\r\n
What is the baseline risk for your patient from the individual trial they would fit best?<\/td>\r\n<\/tr>\r\n
What was the calculated absolute effect? (e.g. <\/span>ARR, NNT<\/a>)<\/span><\/td>\r\n<\/tr>\r\n
What proportion of the included studies report on this outcomes?<\/td>\r\n<\/tr>\r\n
If performed, what GRADE rating was assigned to each outcome?<\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n

Statistical heterogeneity: What was the statistical heterogeneity?<\/span><\/span><\/h1>\r\nFor information regarding the interpretation of forest plots refer to Appendix: Fundamental Statistics<\/a>.\r\n\r\n\r\n\r\n\r\n\r\n
Table 12. Different methods of assessing heterogeneity.<\/caption>\r\n
Methods to Assess Heterogeneity<\/strong><\/td>\r\nDescription<\/strong><\/td>\r\n<\/tr>\r\n
Visual assessment<\/td>\r\nAn intuitive visual evaluation of [pb_glossary id=\"107\"]heterogeneity[\/pb_glossary] (see examples below)<\/td>\r\n<\/tr>\r\n
Cochran\u2019s Q<\/td>\r\nA yes\/no test that shows statistical evidence of [pb_glossary id=\"107\"]heterogeneity[\/pb_glossary] if p <0.10 (analogous to the test for interaction used in subgroup analyses)<\/td>\r\n<\/tr>\r\n
I2<\/sup><\/td>\r\nI2<\/sup> ranges from 0-100% and represents the amount of variability in the point estimate across trials. Rule-of-thumb (one of many): I2<\/sup> <25% = minimal heterogeneity; I2<\/sup> >50% = substantial [pb_glossary id=\"107\"]heterogeneity[\/pb_glossary] (may not be appropriate to meta-analyze trials) (preferred over Cochran's Q)<\/strong><\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n
\r\n\r\nE.g. #1 A forest plot from a review (Koshman SL et al<\/a>.) evaluating the impact of pharmacist involvement in the care of patients with heart failure on all-cause hospitalization rate:\r\n\"\"\r\n<\/em>Plot 2. Pharmacist collaborative care vs. usual for patients with heart failure on the outcome of all-cause hospitalization.<\/em>\r\n\r\nVisually it can be seen that the [pb_glossary id=\"613\"]point estimates[\/pb_glossary] are directionally consistent and all the CIs<\/a> overlap. Consequently meta-analyzing the results for this outcome is appropriate. Notably, this is a case of appropriate [pb_glossary id=\"1101\"]meta-analysis[\/pb_glossary]<\/em> despite there being \"moderate\" statistical heterogeneity as measured by I2<\/sup> (34.4%), as discussed in the note below.<\/em>\r\n\r\n<\/div>\r\n
\r\n\r\nE.g. #2 A forest plot from a review of exercise for depression (Cooney GM et al.<\/a>) evaluating the effects of exercise plus treatment vs. treatment alone:<\/em>\r\n\"\"<\/em>\r\nPlot 3. Exercise plus treatment vs. treatment alone for patients with depression on the outcome of reduction in depression symptoms post-treatment.<\/em>\r\n\r\nVisually it can be seen that the [pb_glossary id=\"613\"]point estimates[\/pb_glossary] have unreasonable variation and the CIs<\/a> have minimal overlap. Consequently [pb_glossary id=\"107\"]heterogeneity[\/pb_glossary] is a concern and additional considerations are necessary, as discussed more below.<\/em>\r\n\r\n<\/div>\r\n
\r\n\r\nIf [pb_glossary id=\"107\"]heterogeneity[\/pb_glossary] is judged to be too high, this requires either:\r\n