Table 3.
Sensitivity analyses and stratified meta-regression for assessing heterogeneity among case–control studies looking at the exposures tonsillectomy in childhood and “surgery” premorbid 1
|
Exposure variables |
Tonsillectomy - childhood |
Surgery - premorbid |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| |
Studies |
Sub-group |
Meta-regression |
Studies |
Sub-group |
Meta-regression |
||||
| n | Total ( n ) | OR (95% CI) 2 | p | β (95% CI) 3 | n | Total ( n ) | OR (95% CI) 2 | p | β (95% CI) 3 | |
|
Population-level characteristics | ||||||||||
|
Latitude |
12 |
8,836 |
1.32 (1.08-1.61)* |
0.79 |
1.00 (0.96-1.03) |
15 |
2,470 |
1.19 (0.83-1.70) |
0.89 |
1.00 (0.97-1.03) |
| (geographic) | ||||||||||
|
Female: Male |
10 |
8,218 |
1.39 (1.15-1.68)* |
0.45 |
1.20 (0.71-2.04) |
14 |
2,362 |
1.22 (0.84-1.79) |
0.91 |
0.98 (0.64-1.48) |
|
Mean age of MS onset |
9 |
2,818 |
1.17 (0.85-1.61) |
0.51 |
0.98 (0.90-1.06) |
10 |
1,551 |
1.12 (0.77-1.63) |
0.53 |
0.94 (0.77-1.16) |
|
Study-level characteristics | ||||||||||
|
NOS score |
11 |
8,480 |
1.29 (1.03-1.60)* |
0.77 |
0.98 (0.85-1.13) |
15 |
2,470 |
1.19 (0.83-1.70) |
0.58 |
0.89 (0.58-1.37) |
|
NOS score (< 6 vs. ≥ 6) |
12 |
8,836 |
1.32 (1.08-1.61)* |
0.60 |
0.88 (0.51-1.50) |
15 |
2,470 |
1.19 (0.83-1.70) |
0.84 |
0.84 (0.13-5.57) |
|
Language (English vs. non-English) |
12 |
8,836 |
1.32 (1.08-1.61)* |
0.10 |
0.62 (0.34-1.12) |
15 |
2,470 |
1.19 (0.83-1.70) |
0.73 |
0.78 (0.17-3.61) |
|
Pub type (published vs. unpublished) |
12 |
8,836 |
1.32 (1.08-1.61)* |
0.09 |
0.50 (0.22-1.13) |
15 |
2,470 |
1.19 (0.83-1.70) |
N/A4 |
|
|
Number of covariates 5 (≤ 2 vs. > 2) |
12 |
8,836 |
1.32 (1.08-1.61)* |
0.04 |
1.64 (1.02-2.62)* |
15 |
2,470 |
1.19 (0.83-1.70) |
0.73 |
0.88 (0.39-1.96) |
|
Sample size (≤ 100 vs. > 100) |
12 |
8,836 |
1.32 (1.08-1.61)* |
0.54 |
0.52 (0.05-5.27) |
15 |
2,470 |
1.19 (0.83-1.70) |
0.56 |
0.76 (0.29-2.01) |
| MS cases (≤ 100 vs. > 100) | 12 | 8,836 | 1.32 (1.08-1.61)* | 0.37 | 1.32 (0.68-2.60) | 15 | 2,470 | 1.19 (0.83-1.70) | 0.70 | 0.87 (0.40-1.90) |
1 All analyses weighted by sample size.
2 OR (95% CI) were calculated using the random effects model in RevMan.
3 β (95% CI) were calculated using the study level log OR and the SE of the estimate (calculated in RevMan) by univariate random effects meta-regression in STATA 12 using the megareg command.
4 All studies were published – analysis not warranted.
5 Number of potentially confounding variables adjusted for in individual studies.
CI = confidence interval, MS = multiple sclerosis, NOS = Newcastle-Ottawa Scale, OR = odds ratio, Pub = Publication, SE = Standard error.