3.1 Studies identification and retrieval
The combined literature search strategy retrieved a total of 189 potential studies, of which 32 were screened for full-text review and 11 studies were eligible to be included in the systematic and meta-analysis (Fig. 1)
3.2 Systematic review
Characteristics of included studies
Eleven articles found in different parts of Ethiopia published from 2010 to 2019 from international peer-reviewed and national journals included to estimate incidence and predictors of tuberculosis among HIV infected patients. A total sample size of 8,875 participants were involved in the final analysis. Among this, 1, 036 new tuberculosis cases were found among HIV patients from 2010 to 2019 published years. The sample size of included studies ranged from 271 (13) to 4210 (21) subjects. All of the reviewed studies were retrospective follow-up studies with a follow-up duration for assessing the outcome ranging from 42 months to 84 months (11-15, 21-26)
Findings from a review of studies showed that the highest and lowest incidence rate of tuberculosis among HIV patients was in Afar region (12 ) and in the Amhara region (23 ) which was 8.64 and 2.22 per 100 person-year observation respectively (12, 23) (see Table 1). The majority of the studies reported that positively significant predictors for the incidence of tuberculosis were ambulatory and Bedridden vs. working, being anemic individuals, WHO clinical staging III and IV, being males, CD4 less than 200 cells/mm3, past history of tuberculosis, not on CPT and IPT.
Table 1: Characteristics of included studies with the outcome of the studies (n = 11)
The author with the publication year
|
Study region
|
Study design
|
Study population
|
Follow-up time(months)
|
Sample size
|
Total PY(O)
|
IR per 100 PYO
|
Study quality
|
Ayalaw et al,2015
|
Amhara
|
Retrospective follow-up study
|
Children
|
72
|
271
|
1100.5
|
4.73
|
Good
|
Endalamaw et al,2018
|
Amhara
|
Retrospective follow-up study
|
Children
|
60
|
352
|
1294.7
|
2.63
|
Good
|
Beshir MT et al,2019
|
Oromia
|
Retrospective follow-up study
|
Children
|
60
|
428
|
1109.6
|
6.04
|
Good
|
Alemu YM. Et al,2016
|
Amhara
|
Retrospective follow-up study
|
Children
|
60
|
645
|
1854.0
|
4.26
|
Good
|
Jerene et al,2017
|
AA&SNNP
|
Retrospective follow-up study
|
Adults
|
84
|
660
|
2843.5
|
2.25
|
Good
|
Ahmed A, et al,2017
|
Afar
|
Retrospective follow-up study
|
Adults
|
84
|
451
|
1377.4
|
8.64
|
Good
|
Temesgen,2017
|
Amhara
|
Retrospective follow-up study
|
Adults
|
60
|
492
|
1285.5
|
6.46
|
Good
|
Tamiso,2016
|
SNNP
|
Retrospective follow-up study
|
Adults
|
72
|
496
|
1977.6
|
5.36
|
Good
|
Assefa et al,2014
|
Amhara
|
Retrospective follow-up study
|
Adults
|
42
|
400
|
1181.8
|
2.20
|
Good
|
Alene et al,2013
|
Amhara
|
Retrospective follow-up study
|
Adults
|
60
|
470
|
1724.1
|
7.89
|
Good
|
Kassa et al, 2012
|
AA
|
Retrospective follow-up study
|
Adults
|
60
|
4210
|
8792.3
|
3.07
|
Good
|
3.3 Meta-analysis
Given a considerable heterogeneity of the outcome across eleven included studies, this meta-analysis aims to determine the incidence of tuberculosis and identify different predictor factors of active TB among HIV positive patients taking ART in Ethiopia.
3.3.1 Pooled incidence of active TB among HIV positive patients
All eleven articles provided information on the incidence of tuberculosis among PLHIV in Ethiopia (11-15, 21-26) . Based on the results of random-effects method, the overall pooled incidence rate of TB per 100-person year observation among HIV infected patients was 4.8 (95% CI: 3.69–5.83) and the heterogeneity was considerable (I2 = 94.1%, Q=170, DF=10, variance= 2.97, z=8.7, p <0.0001). Fig (2).
3.3.2 Subgroup analyses of the incidence rate of tuberculosis in people living with HIV (PLWHIV)
We conducted subgroup analysis based on the pre-defined category of study setting, year of publication, study population, length of Follow-up time and sample size classification used to assess the incidence of tuberculosis (Table 2).
In our subgroup analysis of the incidence of tuberculosis among PLWHIV in Ethiopia with studies in Hospital-based was slightly lesser than facility-based studies (4.7; 95% CI: 3.46, 5.97 Vs 5.0; 95% CI: 1.92, 8.01).
Furthermore, we conducted a subgroup analysis based on the year of publication date used to assess the incidence of tuberculosis. The pooled incidence rate estimates of tuberculosis for the studies conducted within 2010 – 2014 was significantly lesser than studies in 2015-2019 (4.3; 95 % CI: 1.95, 6.66) Vs 5.0; 95% CI: 3.54, 6.40) (Table 2).
We also performed a subgroup analysis based on the total sample size which the studies were conducted. The pooled incidence of tuberculosis among people living with HIV in Ethiopia was significantly higher for sample size less than 500 (5.4; 95% CI: 3.80, 7.08) compared to those articles with a sample size of greater than or equal to 500 (3.1; 95% CI: 2.23, 4.00).
Finally, we conducted subgroup analysis based on the follow-up period for sampled population.so that the pooled incidence of tuberculosis among HIV infected ART user individuals was significantly higher for the follow-up period less than sixty months (5.1; 95%CI: 3.64,6.46) compared to longer period of follow-up time (4.1; 95% CI: 1.87,6.27) (Table 2).
In the analyses stratifying summary estimates of TB incidence rates in the above categories, heterogeneity remained high for each stratum. This implied that these variables did not explain most of the heterogeneity observed in the TB incidence rate (Table 2.)
Table 2: Incidence of tuberculosis in people with HIV in Ethiopia: Subgroup meta-analysis and heterogeneity analysis
Subgroup types
|
Observation(N)
|
IR per 100 PY(O)
|
95%CI
|
I2 (%)
|
Q-statistic
|
Tau2
|
Df
|
P value
|
Year of publication
|
|
2010_2014
|
3
|
4.3
|
[1.95,6.66]
|
96.3
|
53.9
|
4.11
|
2
|
P<0.0001
|
2015_2019
|
8
|
5.0
|
[3.54, 6.40]
|
93.6
|
109.5
|
3.88
|
7
|
P<0.0001
|
Study Setting
|
|
Hospital based
|
8
|
4.7
|
[3.46,5.97]
|
93.3
|
104.9
|
2.95
|
7
|
P<0.0001
|
Facility based
|
3
|
5.0
|
[1.92,8.01]
|
96.8
|
63.4
|
6.95
|
2
|
P<0.0001
|
Study population
|
|
Children
|
4
|
4.3
|
[2.96,5.71]
|
83.7
|
18.5
|
1.61
|
3
|
P<0.0001
|
Adults
|
7
|
5.0
|
[3.51,6.50]
|
95.9
|
148.1
|
3.77
|
6
|
P<0.0001
|
Sample size range
|
|
Less than 500
|
8
|
5.4
|
[3.80,7.08]
|
93.8
|
112.6
|
5.20
|
7
|
P<0.0001
|
Greater than/equal 500
|
3
|
3.1
|
[2.23,4.00]
|
85.8
|
14.1
|
0.51
|
2
|
P=0.001
|
Follow-up period(months)
|
|
Less than or equal to 60
|
8
|
5.1
|
[3.64,6.46]
|
94.7
|
132.8
|
5.80
|
7
|
P<0.0001
|
Greater than 60
|
3
|
4.1
|
[1.87,6.27]
|
94.1
|
34.0
|
3.53
|
2
|
P<0.0001
|
We also performed a subgroup analysis based on the Study population used to determine the incidence rate of tuberculosis among HIV patients. Four studies were in children and the rest were about adult people. The pooled incidence rate tuberculosis in people living with HIV was 4.3 (95% CI: 2.96, 5.71) and 5.0 (95% CI: 3.51, 6.50) for the studies conducted children and adults respectively (see Fig 3)
3.3.3 Sensitivity analysis
We performed a leave-one-out sensitivity analysis for the sake of further investigating the potential sources of heterogeneity in the analysis of the incidence rate of tuberculosis in PLWHIV. Our sensitivity analysis showed that our findings were not influenced by a single study that all the point estimates of the leave-one-out are within the confidence interval of the combined estimate and it is stable. Our pooled estimated incidence varied between 4.4 (3.39-5.38) and 5.0 (3.85–6.23) after the deletion of a single study (see Table 3 and fig 4).
Table 3: sensitivity analysis for the incidence of tuberculosis among HIV infected patients in Ethiopia.
Study omitted Estimate [95% CI]
Ayalaw et al (2015) 4.8 (3.63, 5.90)
Endalamaw et al (
1) 5.0 (3.82, 6.16)
Beshir MT et al (2019) 4.6 (3.53, 3.74)
Alemu YM. Et al (2016) 4.8 (3.65, 5.99)
Jerene et al (2017) 5.0 (3.85, 6.23)
Ahmed A, et al (2017) 4.4 (3.39, 5.38)
Temesgen et al (2017) 4.6 (3.51, 5.68)
Dalbo M. Tamiso A (2016) 4.7 (3.57, 5.82)
Assefa et al (2014) 5.0 (3.87, 6.19)
Alene et al (2013) 4.4 (3.43, 5.43)
A. Kassa et al (2012) 5.0 (3.63, 6.32)
Combined 4.8 (3.69, 5.83)
3.3.4 Publication bias
The funnel plot and Egger’s regression tests showed that there is no evidence of substantial publication bias for the incidence of tuberculosis among HIV infected patients in Ethiopia (see additional files 1 and 2).
3.3.5 Pooled estimated effects of predictors on the incidence of Tuberculosis among HIV patients
Meta-analysis was conducted to identify pooled estimates of predictors for the incidence of tuberculosis among HIV infected individuals after the initiation of ART in Ethiopia. Among pooled estimates of predictors Bedridden functional status, anemia status, WHO clinical staging, cotrimoxazole preventive therapy, Isoniazid, preventing therapy, CD4 cell count and Gender were found to be significant predictors for the incidence of tuberculosis whereas Ambulatory Functional status, previous history of tuberculosis, past opportunistic infections and family size were not statistically significant pooled predictors estimates for the incidence of tuberculosis among HIV infected individuals in Ethiopia.
The hazard of developing tuberculosis among HIV infected individuals for the pooled estimates of four observations for anemia was 2.3 times more likely as compared to those who had no anemia (HR:2.30; 95%CI:1.75,3.02; I2(p-value) =31.2%(0.18); Publication bias: β(p-value) =0.82(0.60)).
Incidence of tuberculosis among HIV infected individuals was higher for those on WHO clinical staging III and IV than I and II by meta-analysis of 10 studies (HR: 2.26; 95%CI: 1.70, 3.02; I2 (test) = 47.2 %( p=0.048); publication bias: β (p-value) =-0.13(0.90)).
The Hazard of developing tuberculosis among HIV infected individuals for those not using CPT was 2.16 times more likely than CPT users for five articles (HR: 2.16; 95%CI:1.23,3.72; I2(test)= 55.3%( p= 0.062); publication bias: β(p-value) = -1.37 (0.54)).
The hazard of developing TB among male HIV infected individuals was 37% higher among combined of five articles by meta-analysis compared to those female patients (HR: 1.37; 95%CI:1.16,1.63; I2(test)= 0(P=0.59); publication bias: β(p-value) = -2.88 (0.01)).
Since the heterogeneity of the variables not using IPT and CD4 count was high, further analysis by Meta-regression required to manage it (see table 5).
Table 5: a meta-analysis of predictors of incidence of tuberculosis among HIV positive patients in Ethiopia (2010-2019)
Variables
|
Observation
|
HR
|
95%CI
|
p-value of Q
|
I2 (%)
|
Tau2
|
Q-statistic
|
P value of
estimate
|
Functional status
|
|
Working(ref)
|
4
|
1
|
|
|
|
|
|
|
Ambulatory
|
1.46
|
[0.84,2.51]
|
p=0.001
|
78.8
|
0.29
|
18.83
|
P=0.18
|
Bedridden
|
2.01
|
[1.21,3.35]
|
p= 0.137
|
42.7%
|
0.13
|
6.98
|
p = 0.007
|
Anemia status
|
|
Not anemic(ref)
|
8
|
1
|
|
|
|
|
|
P<0.001
|
Anemic
|
2.30
|
[1.75,3.02]
|
P=0.179
|
31.2
|
0.05
|
10.17
|
WHO clinical staging
|
|
Stage I/II(ref)
|
10
|
1
|
|
|
|
|
|
P<0.001
|
Stage III/IV
|
2.26
|
[1.70,3.02]
|
p= 0.048
|
47.2
|
0.09
|
17.05
|
Cotrimoxazole preventive therapy
|
|
Yes(ref)
|
5
|
1
|
|
|
|
|
|
p = 0.007
|
No
|
2.16
|
[1.23,3.72]
|
p= 0.062
|
55.3
|
0.20
|
8.95
|
Isoniazid preventing therapy
|
|
Yes(ref)
|
7
|
1
|
|
|
|
|
|
P=0.001
|
No
|
3.67
|
[1.73,7.76]
|
P<0.001
|
79.1
|
0.74
|
28.7
|
CD4 cell count
|
|
Greater than 200
|
6
|
1
|
|
|
|
|
|
P=0.002
|
Less than or equal to 200
|
2.12
|
[1.31,3.43]
|
P<0.001
|
90.8
|
0.32
|
54.38
|
Previous history of Tuberculosis
|
|
No
|
3
|
1
|
|
|
|
|
|
|
Yes
|
1.49
|
[0.77,2.89]
|
p= 0.004
|
73.5
|
0.39
|
15.11
|
P=0.24
|
Gender
|
|
Female
|
5
|
1
|
|
|
|
|
|
|
Male
|
1.37
|
[1.16, 1.63]
|
p =0.588
|
0
|
0
|
2.82
|
P<0.001
|
Family size
|
Small family size(<5)
|
|
1
|
|
|
|
|
|
|
Large family size(≥5)
|
3
|
1.18
|
[0.89,1.56]
|
p = 0.29
|
19.0
|
0.01
|
2.47
|
P= 0.26
|
Meta-regression for High heterogeneity variables.
Isoniazid preventive therapy was a statistically significant predictor for the incidence of tuberculosis among HIV infected individuals for seven articles. The risk of developing tuberculosis among HIV infected individuals for those not using IPT was 3.47 times more likely than IPT users ((HR: 3.67; 95%CI:1.73,7.76; I2(test)= 79.1%(p<0.001); publication bias: β(p-value) =3.47(0.14)). Meta-regression revealed that not using IPT in adults’ study group was higher than children for the incidence of tuberculosis (HR: 2.12; 95% CI: 1.17, 3.86).
CD4 count at baseline of ART initiation was a significant pooled factor for developing tuberculosis among six articles. The risk of tuberculosis incidence among HIV infected individuals was 2.12 times more likely for those who had less than 200 cells/ mm3 at baseline than greater CD4 cell count (HR: 2.12; 95%CI:1.31,3.43; I2(test)= 90.8%( P<0.001); publication bias: β (p-value) = -3.16 (0.57)).
Table 6: Meta-regression of selected predictors to the incidence of tuberculosis among HIV positive patients in Ethiopia (2010-2019)
Study level variables
|
Low CD4 count
|
Not using IPT
|
HR with 95% CI
|
I2
|
Tau2
|
Adj.R2
|
HR with 95% CI
|
I2
|
Tau2
|
Adj.R2
|
Study population
|
Children
|
Ref.
|
|
|
|
|
|
|
|
Adults
|
2.12[1.17,3.86]
|
90.6%
|
0.28
|
0
|
2.94[0.50,17.02]
|
71.4%
|
0.52
|
25.97%
|
Study setting
|
Health facility based
|
Ref.
|
|
|
|
|
|
|
|
Hospital based
|
0.96[0.21,4.44]
|
92.6%
|
0.36
|
26.5%
|
0.31[0.06,1.63]
|
69.2%
|
0.44
|
37.93%
|
Follow-up time
|
Less/equal to 5 years
|
Ref.
|
|
|
|
|
|
|
|
Greater than 5 years
|
2.18[0.73,6.50]
|
78.4%
|
0.15
|
46.05%
|
0.53[0.06,5.08]
|
68.4%
|
0.68
|
3.61%
|