Table 2

Prevalence of bisexual behaviour among MSM by study and design characteristics

CategorySubgroupsNo of studiesPrevalence estimate (%) and 95% CIHeterogeneity I2 (%)Univariate meta-regressionMultivariate meta-regression
tp Valuetp Value
Data collection methodInterviewer-administered3228.4 (25.5 to 31.9)95.6NS
Self-administered1036.5 (24.6 to 48.6)97NS
Not reported or can not be classified into above two categories736.6 (28.2 to 45.9)96.1NS
SettingThrough a variety of recruitment methods;1526.7 (20.6 to 32.1)97.4Ref
VCT323.8 (14.7 to 33.6)96.6−0.390.696−0.220.825
MSM venues1935.1 (30.1 to 41.3)952.310.0211.850.064
Via the working group438.9 (26.0 to 52.7)97.41.910.0570.830.404
Not reported830.6 (26.2 to 35.3)85.90.870.3841.090.277
RegionsAll over China240.8 (18.9 to 64.8)NS
Southwest1227.3 (24.6 to 30.1)86.6NS
South China840.3 (30.4 to 50.6)95.8NS
East China1132.6 (24.0 to 41.8)96.8NS
North China1026.1 (18.5 to 34.3)96.6NS
Northeast631.0 (26.0 to 36.1)87.6NS
Recall window of bisexual behaviour*3 Months429.1 (12.7 to 48.8)96.9Ref
6 Months3429.0 (25.9 to 32.2)95.50.010.990
12 Months1041.7 (33.1 to 50.7)96.61.780.075
Grading of the QACO scoreSatisfactory3432.0 (28.1 to 35.9)96.9NS
Good1529.6 (23.9 to 35.5)95.1NS
Sample size<10004231.7 (28.0 to 35.6)95.9NS
≥1000728.5 (21.5 to 36.1)98.3NS
Data collection periodBefore 20062133.2 (26.4 to 40.4)97.6NS
2006 and later2330.9 (27.6 to 34.4)94.5NS
Not reported or can not be classified into above two categories524.5 (16.1 to 34.1)95.4NS
Sampling methodRespondent-driven sampling529.8 (24.5 to 35.8)86.7NS
All other non-probability sampling4431.8 (28.1 to 35.6)97NS
  • * One stduy taking 2 months was not included.

  • –, cannot be calculated; MSM, men who have six with men; NS, not significant; QACO, quality assessment checklist for observational studies; VCT, voluntary counselling and testing.