Table 5.

Factors associated with the presence of thrombocytopenia

Results of uni- and multivariable log-binomial regression models
Covariate	Univariable (n = 400)		Multivariable‡ (n = 397*)
	PR (95% CI)	p value	PR (95% CI)	p value
Sex
Male	1		1
Female	0.33 (0.16 to 0.69)	0.003	0.32 (0.15 to 0.68)	0.003
HAART status at recruitment§
Not on HAART	1		1
On HAART	0.35 (0.16 to 0.76)	0.009	0.32 (0.14 to 0.75)	0.008
CD4 count cells/μl
<50	1
50 to < 200	0.40 (0.17 to 0.95)	0.037
≥ 200**	0.48 (0.18 to 1.33)	0.159
BMI
under weight	1
normal	0.90 (0.25 to 3.20)	0.873
over weight	0.67 (0.16 to 2.75)	0.575
Incomplete information*	-	-
Age, years
18 to < 30	1
30 to < 35	2.36 (0.76 to 7.34)	0.139
35 to < 41	1.74 (0.57 to 5.27)	0.328
≥ 41	1.74 (0.56 to 5.40)	0.335
Education level
None	1
Minimum Primary	1.05 (0.13 to 8.63)	0.961
Minimum Secondary	1.02 (0.12 to 8.62)	0.983
Tertiary	0.58 (0.05 to 7.00)	0.667
Don’t know	-	-
Employment status
Unemployed	1
Employed	2.62 (0.61 to 11.3)	0.196
Marital status
Not married	1
Married	1.50 (0.73 to 3.09)	0.274
Presence/history of oral candidiasis
No	1
Yes	1.44 (0.66 to 3.15)	0.358
Presence/history of fever
No	1
Yes	1.01 (0.50 to 2.06)	0.975

^‡Multivariable model only includes variables for which data are shown below because all others were lacking significance.

§All participants on HAART were categorized into one group for this analysis because none of the 15 patients on AZT had thrombocytopenia.

AZT = azidothymidine or zidovudine.

*Three participants with missing BMI could not be included in the log-binomial regression model because the outcome did not vary in this group.

**CD4 categories 200 to <350 and ≥ 350 were combined for this analysis because none of the participants with CD4 count ≥ 350 had thrombocytopenia.