Skip to main content

Advertisement

Table 3 Meta-analysis methods of therapeutic meta-analyses published in 2013 and including non-randomized studies (n = 188)

From: Meta-analyses including non-randomized studies of therapeutic interventions: a methodological review

Characteristic Included NRSI only Included NRSI&RCTs All meta-analyses
(n = 69) (n = 119) (n = 188)
No (%) No (%) No (%)
Studies combined    
Type of studies included    
 Only prospective cohort studies 2 (3) 3 (3) 5 (3)
 Only cohort studies 18 (26) 34 (29) 52 (28)
 Including also case–control studies 18 (26) 28 (23) 46 (24)
 Including all types of NRSI 5 (7) 18 (15) 23 (12)
 Other 28 (41) 39 (33) 67 (36)
 “Observational studies” 6 (9) 22 (18) 28 (15)
 “Prospective and retrospective studies” 11 (16) 12 (10) 23 (12)
 “Retrospective studies” 11 (16) 5 (4) 16 (9)
Did not clearly report design for each study 43 (62) 87 (73) 130 (69)
Meta-analysis combining results of NRSI and RCTs NA 88 (74) NA
 Results also presented separately NA 44/88 (50) NA
Meta-analysis combining studies of different designs 39 (57) 61 (51) 100 (53)
Meta-analysis model    
Network meta-analysis 0 2 (2) 2 (1)
Crude or adjusted estimates used    
 Combined crude and adjusted estimates 10 (14) 12 (10) 22 (12)
 Adjusted 8 (12) 13 (11) 21 (11)
 Both crude and adjusted estimates separately 4 (6) 4 (3) 8 (4)
 Crude 2 (3) 4 (3) 6 (3)
 Not reported/unclear 45 (65) 86 (72) 131 (70)
If adjusted, list of confounding factors 15/22 (68) 19/29 (66) 34/51 (67)
Type of model    
 Random-effects model 39 (57) 56 (47) 95 (50)
 Fixed-effects model, unless high heterogeneity, then random-effects model 19 (28) 33 (28) 52 (28)
 Both fixed- and random-effects models 5 (7) 21 (18) 26 (14)
 Fixed-effects model 3 (4) 6 (5) 9 (5)
 Not reported/unclear 3 (4) 3 (3) 6 (3)
Heterogeneity assessment & exploration    
Heterogeneity assesseda 66 (96) 116 (97) 182 (97)
 I2 57 (83) 107 (90) 164 (87)
 Cochran Q χ2 test 41 (59) 74 (62) 115 (61)
 τ2 4 (6) 2 (2) 6 (3)
Heterogeneity exploreda 54 (82) 103 (87) 157 (84)
 Meta-regression analysis 13 (19) 21 (18) 34 (18)
 Subgroup analysis 44 (64) 82 (69) 126 (67)
 Sensitivity analysis 35 (51) 74 (62) 109 (58)
Reporting bias assessmenta 49 (71) 78 (66) 127 (68)
 Standard funnel plot 41 (59) 70 (59) 111 (59)
 Egger’s test 32 (46) 36 (30) 68 (36)
 Begg’s test 18 (26) 24 (20) 42 (22)
Among MAs with ≥10 studies, reporting bias assessed 24/29 (83) 58/76 (76) 82/105 (78)
  1. MAs meta-analyses; NRSI non-randomized studies of intervention; RCTs randomized controlled trials
  2. aMore than one item could have been reported for each meta-analysis