Supplementary Appendix Table of Contents
I. Methods
A. Derivation of base case parameters from the IPT trial
II. Estimates of the impact of IPT and ART in a clinic cohort by baseline CD4 categories A. Impact of ART on TB incidence
B. Impact of ART on all-cause mortality C. Impact of 6H and 36H on TB incidence D. Impact of 6H and 36H on all-cause mortality
III. Sensitivity Analysis
A. Influential parameters B. Threshold analysis
I.A. Derivation of base case parameters from the IPT trial
Base case statistics for the annual rates of tuberculosis and death were derived from data gathered from enrollees of the Botswana IPT trial (Samandari, Lancet 2011). In order to present the most conservative estimates, the intent-to-treat analysis results were used. TB and death rates for the 36-month trial duration were calculated using a person-time approach, and 95% confidence intervals were derived using survival analysis assuming an exponential distribution. These rates and confidence intervals were calculated for each CD4, TST, ART, and IPT subgroup.
II.A. Impact of ART on TB incidence
A key table describing the impact of ART on TB incidence for our analysis is shown below. It shows the estimated proportionate decline in TB incidence as ratios for three different scenarios of CD4 thresholds of ART initiation. Within each CD4 threshold category, the patients were divided into PLWH below or above the threshold and a ratio was calculated that represents the percent reduction in TB as compared to the corresponding rates (above and below the CD4 threshold) observed in the Botswana trial. The observed TB incidence rates (A and B) were derived from participants in the 6-month or 36-month arms of the IPT Trial and C-H were calculated using those values while the ratios (%
declines) were determined from the articles listed in Table II.B’.
Table II.A.1: Decline in TB incidence CD4
Threshold
Status at entry into program Annual incidence
of TB
% decline in TB comparing no ART to
no ART groups
% decline in TB comparing ART to
ART group
Equation % Equation %
CD4<200 ART received (CD4≤200) A derived Botswana trial derived Botswana trial ART not immediate (CD4>200) B derived Botswana trial derived Botswana trial
CD4<250 ART received (CD4≤250) C r4= (C-A)/A -21%
ART not immediate (CD4>250) D r1=(D-B)/B -20%
CD4<350 ART received (CD4≤350) E r5= (E-A)/A -26%
ART not immediate (CD4>350) F r2= (F-B)/B -50%
CD4<500 ART received (CD4≤500) G r6= (G-A)/A -35%
ART not immediate (CD4>500) H r3= (H-B)/B -65%
The estimates of the ratios r1–r6 were found or imputed from the following references. Sometimes the references that had the most relevant evidence to our study context were selected over less relevant studies. Studies considered more relevant were those examining TB incidence among PLWH living in community settings in TB-endemic countries.
Table II.A.2: Calculated values for decline in TB incidence Ratio Selected base
case value
Reference in the literature and calculated value of the ratio Range tested in sensitivity model
r1 -20% Lawn (-16%) 0.04-0.30
r2 -50% Badri (-46%), Holmes (-57%), Lawn (-35%), Girardi (-40%) 0.10-0.97
r3 -65% Lawn (-65%), Girardi (-46%) 0.13-0.97
r4 -21% Lawn (-21%), Gupta (-23%) 0.04-0.32
r5 -26% Badri (-24%), Miranda (-29%), Lawn (-31%), Girardi (-22%), Moreno (-24%), CAUSAL (-21%), Gupta (-36%)
0.05-0.38 r6 -35% Lawn (-34%), Girardi (-37%), Cohen (-37%), Chaisson from Cohen 2011 (-51%), CAUSAL
(-34%), Gupta (-43%)
0.07-0.52
II.B. Impact of ART on all-cause mortality
A key table describing the impact of ART on mortality for our analysis is shown below. It shows the estimated proportionate decline in mortality as ratios for three different scenarios of CD4 thresholds of ART initiation. Within each CD4 threshold category, the patients were divided into PLWH below or above the threshold and a ratio was calculated that represents the percent reduction in mortality as compared to the corresponding rates (above and below the CD4 threshold) observed in the Botswana trial. The observed TB incidence rates (A and B) were derived from participants in the 6-month or 36-month arms of the IPT Trial and C-H were calculated using those values while the ratios (%
declines) were determined from the articles listed in Table II.B.2.
Table II.B.1: Decline in all-cause mortality CD4
Threshold
Status at entry into program Annual incidence
of death
% decline in death comparing no ART to
no ART groups
% decline in death comparing ART to
ART group
Equation % Equation %
CD4<200 ART received (CD4≤200) J derived Botswana trial derived Botswana trial
ART not immediate (CD4>200) K derived Botswana trial derived Botswana trial
CD4<250 ART received (CD4≤250) L r4= (L-J)/J -22%
ART not immediate (CD4>250) M r1=(M-K)/K -18%
CD4<350 ART received (CD4≤350) N r5= (E-J)/J -28%
ART not immediate (CD4>350) P r2= (P-K)/K -54%
CD4<500 ART received (CD4≤500) Q r6= (G-J)/J -44%
ART not immediate (CD4>500) R r3= (R-K)/K -71%
The estimates of the ratios r1-r6 were found or imputed from the following references. Sometimes the references that had the most relevant evidence to our study context were selected.
Table II.B.2: Calculated values for decline in all-cause mortality Ratio Selected base
case value
Reference in the literature and calculated value of the ratio Range tested in sensitivity model
r1 -18% Schim & French (imputed -17.7%) 0.04-0.27
r2 -54% Schim & French (imputed -53.6%) 0.11-0.81
r3 -71% Schim & French (imputed -70.8%) 0.14-1.07
r4 -22% Sterne (imputed -22%) 0.04-0.34
r5 -28% averaged Sterne (imputed -39%), Hoggs (-31%), Egger (-18%), May (-23%) 0.06-0.41
r6 -44% Sterne (imputed -44%) 0.09-0.66
II.C. Impact of 6H and 36H on TB incidence
A key table for the impact of IPT on TB incidence is shown below.
Table II.C: Impact of IPT on TB incidence
TST+ TST-
H0 H6 H36 H0 H6 H36
CD4<200 ART i1 3.65% 1.06% i3 0.73% 1.01%
CD4≥200 no immediate ART i2 2.40% 0.36% i4 1.16% 0.86%
The values for TB in H6 and H36 columns are based upon 3-year annualized Botswana IPT trial data by applying SAS’s lifereg procedure on the intent-to-treat cohort. The TB benefit of H0 is back-calculated from the above values using the meta-analysis of Akolo, Cochrane 2010. The reduction in TB for TST-positive PLWH was RR 0.36 (95%CI 0.22, 0.61) but as established in the Botswana IPT trial and other studies (Johnson, AIDS 2001), this benefit declined within 6 months of IPT completion. Since the period of observation is 3 years, the benefit of 6H was assumed not to endure beyond the first year: (Year1 = 64% decline; Year2 & Year3 = 0% decline). So the following calculation was made: 2.36/3.00=0.787 or 21% decline for 3 years.
Despite the lack of statistical significance, it has been repeatedly observed that TST-negative PLWH gain a modest reduction in TB incidence: RR 0.86, 95% CI 0.59-1.26. However, the 14% decline in TB is unlikely to last beyond the first year, and similar to the above calculation, the 3-year benefit was estimated at 2.86/3.00=0.95 or 5% decline for 3 years.
Therefore:
i1 = 3.65%/(1-0.21)=4.62%;
i2 = 2.40%/(1-0.21)=3.04%;
i3 = 0.73%/(1-0.05)=0.77%;
i4 = 1.16%/(1-0.05)=1.22%
The benefit of 6H and 36H in reducing TB was varied across its 95% confidence limits of the model estimates to assess the stability of the final policy rankings.
Two studies presented at the XIX International AIDS Conference (2012 Washington, D.C.) provided information about the combination of IPT + ART. Teferi, et al. described a programmatic retrospective analysis from Namibia, “Effect of primary isoniazid preventive therapy on tuberculosis incidence rate among HIV-infected adults enrolled in HIV care in northern Namibia: a retrospective cohort study” (Abstract # WEPE066).
Participants either received (n=363) or did not receive (n=752) 6-months IPT. CD4 and TST status at ART initiation were not specified. The 6H arm had 1.14/100py TB and the no IPT arm had 3.07/100py (aHR 0.52, 95%CI 0.28-0.95), i.e. 48% reduction. The benefit of IPT was lost after 1.5 years. In the same study, 82% of participants were receiving concomitant ART. The study found that there was no mortality benefit. Rangaka et al. described in “Randomized controlled trial of isoniazid preventive therapy in HIV-infected persons on antiretroviral therapy in South Africa”
(Abstract # THLBB03) an RCT in which ART recipients (100%) received either 12H (n=662) or placebo (n=667).,There was a 48% decline in TB incidence (HR 0.63, 95%CI 0.41-0.94); however, the benefit of IPT was lost after 1.5 years. There was no mortality benefit in this study as well.
While neither of these studies is directly comparable with the Botswana IPT trial and did not stratify by TST status or baseline CD4, they confirm that there is an added benefit of 6- or 12-months IPT for PLHIV receiving ART and that the magnitude of effect we calculate is conservative in terms of the benefit of 6-months IPT. In summary, we calculated a 21% decrease in TB if 6H vs no IPT; Teferi, et al. reported 48% decline; and Rangaka, et al. reported a 38% decline.
II.D. Impact of 6H and 36H on all-cause mortality
A key table for the impact of IPT on death incidence is shown below.
Table II.D: Impact of IPT on all-cause mortality
CD4 level and ART Status TST+ TST-
H0 H6 H36 H0 H6 H36
CD4<200 ART d1 0.91% 1.06% d3 1.09% 1.86%
CD4≥200 no immediate ART d2 2.18% 0.54% d4 0.85% 1.25%
The values for death in H6 and H36 columns are based upon 3-year annualized Botswana IPT trial data by applying SAS’s lifereg procedure on the intent-to-treat cohort. The reduction in death for TST+ PLWH was RR 0.74 (95%CI 0.55, 1.00) in the Akolo, Cochrane 2010 meta-analysis and it is likely that this marginal benefit would decline within 6 months of IPT completion. Since the period of observation is 3 years, the benefit of 6H was assumed not to endure beyond the first year: (Year1 = 26% decline; Year2 & Year3 = 0% decline). So the following calculation was made:
2.74/3.00=0.91 or a 9% decline for the 3 years compared to no IPT.
For TST- PLWH, there was no effect on death: RR 1.02, 95% CI 0.90-1.16 so there was no decrease in death from 6H.
Therefore:
d1 = 0.91%/(1-0.09)=1.00%;
d2 = 2.18%/(1-0.09)=2.40%;
d3 = 1.09%/(1-0.00)=1.09%;
d4 = 0.85%/(1-0.00)=0.85%
The benefit of 6H and 36H in reducing death was varied across its 95% confidence limits of the model estimates to assess for stability of the final policy rankings.
The two studies mentioned above that were presented at the XIX International AIDS Conference (2012 Washington, D.C.), Teferi et al. and Rangaka et al. also provided more information about the effect of the combination of IPT + ART on mortality. Neither presentation included a report of an observed decline in the risk of death as a result of IPT (but baseline TST and CD4 were not shown and both had trends favoring IPT).
There are other published references that have observed a mortality benefit for IPT and we reference those in the manuscript.
III. Sensitivity Analysis
We conducted sensitivity analysis on the model considering both outcomes—incident TB disease and all-cause mortality. Given the large number of variables, the first step was to identify which parameters had the greatest influence on model output, followed by systematically identifying thresholds for which changes in base values would alter potential policy recommendations. The primary goal of this approach was to determine the stability of modeled results if estimates of intervention impact (e.g., effect of ART or IPT) were different than expected, and also assess generalizability of results given marginal differences in population or epidemiological context (e.g., baseline CD4 distribution or TB prevalence).
III.A Influential model parameters
To identify influential model parameters, we generated tornado diagrams for each outcome. The tornado analysis was conducted by varying each model parameter across a range of values at specified intervals. At each interval, model outcomes were recalculated and compared with the result when only base values were used. The magnitude of difference between the result at each interval change and the base value defined the comparative influence of each parameter. The table below outlines the most influential parameters for both outcome measures, including the percent change in results attributable to each parameter and the cumulative percent change when considering more than one parameter.
We chose the net health benefits (NHB) equation to determine comparative cost-effectiveness of each alternative for the purposes of the sensitivity analysis. NHB combines costs, effectiveness and willingness-to-pay (WTP) into a single score that can be used to rank alternatives (Stinnett and Mullahy, MDM, 1998). Assuming a static WTP for one additional unit of health benefit—in this case one additional case of TB or
death averted—the alternative with the highest NHB is the optimal strategy. Using NHB varied each parameter individually and assessed the influence on our results. The benefits of this approach included (1) the ability to compare changes to parameters on cost-effectiveness considering all polices simultaneously and (2) avoiding instability associated with sensitivity analysis using ICERs. The single drawback of this approach was that a WTP would have to be specified for each outcome to run the analysis. In the primary analysis we intentionally avoided assigning a WTP to an additional case of TB or death averted because this is not well described in the literature and unknown in the context of Botswana. Employing a conservative approach, we ran the sensitivity analysis using the NHB equation considering a range of WTP thresholds for each outcome: $10,000, $50,000, and $1,000,000 (Table III.A).
To vary base values, we established ranges for each of the 101 model parameters. Ranges for baseline CD4 distribution, TB prevalence, sensitivity and specificity of tuberculin skin testing (TST), rates of TB and HIV, and effectiveness of ART and IPT were either estimated using available literature, calculated using confidence intervals of Trial results, or estimated by clinicians. Cost inputs were varied by 50% above and below their base values. A full list of model parameters and ranges can be found in Table 2 of the manuscript.
Table III.A: Influential model parameters
Incident TB Outcome
Willingness-to-pay Parameter Brief Description Percent Influence
Cumulative Percent Influence
$10,000
ppART
Per-patient cost of ART 55 55
Threshold_less250
Percent of population below threshold CD4 250 cell/ml 30 85
$50,000
Effect of ART on TB
Reduction in incident TB due to provision of ART 50 50
Effect of CD4 on TB below thresh
Reduction in incident TB due to increase in average CD4 15 66
$1,000,000
Effect of ART on TB
Reduction in incident TB due to provision of ART 64 64
Effect of CD4 on TB below thresh
Reduction in incident TB due to increase in average CD4 27 91 All-cause Mortality Outcome
$10,000
ppART
Per-patient cost of ART 46 46
Threshold_less250
Percent of population below threshold CD4 250 cell/ml 26 72
Effect of ART on mort
Reduction in mortality due to provision of ART 19 91
Effect of CD4 on mort below thresh
Reduction in mortality due to increase in average CD4 4 94
$50,000
Effect of ART on mort
Reduction in mortality due to provision of ART 63 63
Effect of CD4 on mort below thresh
Reduction in mortality due to increase in average CD4 18 81
ppART
Per-patient cost of ART 7 88
Threshold_less250
Percent of population below threshold CD4 250 cell/ml 3 91
$1,000,000
Effect of ART on mort
Reduction in mortality due to provision of ART 66 66
Effect of CD4 on mort below thresh
Reduction in mortality due to increase in average CD4 24 90
III.B Threshold Analysis
After identifying the most influential model parameters, we subjected each of these variables to additional sensitivity analysis. For every parameter in table III.A.1, values were varied in 1 percent increments across their specified range. At each of the 100 intervals, policies were rank-ordered by effectiveness, dominated policies were excluded, and ICERs were recalculated. This resulted in an additional 800 model runs.
Taking this approach we were able to (1) conclude if at any value in the range of an influential parameter could reorder the cost-effectiveness rank of policy alternatives observed in the base case, and (2) if rank-order changes occurred, identify the threshold at which the value would alter the initial findings.
Considering the incident TB outcome, the rank of alternatives did not change at any values in parameter ranges. The one notable occurrence was that increases to the effectiveness of ART eventually resulted in the policy ART500 becoming undominated—meaning it was no longer more costly and less effective than other alternatives and could be considered when choosing policies for TB-HIV mitigation. Though undominated in upper ranges of ART effectiveness, ART500 remained the most expensive option per TB case averted compared with the other polices in this analysis. A similar situation with the policy ART500 was observed if assuming an increase in average CD4 count in a PLHIV population is more effective in preventing TB disease than considered in the model.
Considering the mortality outcome, there was no change rank-order of policy alternatives for any value in parameter ranges; however, at some parameter values, policies became dominated and were excluded from the list of potential policy alternatives. If ART effectiveness or increases
to average CD4 cell count on all-cause mortality is less than expected, ART500 would be dominated and should not be considered a cost- effective alternative. Conversely, if the effect of ART or increase in average CD4 count is more effective than assumed in the model, the policy TST_H6H36 would be dominated and excluded. Finally, TST_H6H36 was dominated when we assumed 47% or more of the population had CD4 counts below 250 cells/ml or the annual per-patient cost of ART was lower than $271.
Table III.B summarizes findings of the threshold analysis by parameter. Table III.B-2 provides model output at each of the threshold values.
Table III.B: Key Threshold Values
Incident TB Outcome Parameter
(Output in Table III.B-2) Base Value Threshold Value
Percent increase (decrease)
Implication for results
Effect of ART on TB (1) 0.59 0.71 20
If the provision of ART is ≥20% more effective in preventing incident TB than assumed in the model, ART500 would no longer be dominated and may be considered a potential alternative
Effect of CD4 on TB below thresh
(2) 0.21 0.26 25
If the reduction in the rate of incident TB as a result of increases in average CD4 cell count is ≥25% greater than assumed, ART500 would no longer be dominated and may be considered a potential alternative All-cause Mortality Outcome
Effect of ART on TB (3) 0.62 0.44 (28)
If the provision of ART is ≤28% less effective in preventing all-cause mortality than assumed, ART500 would become dominated and should be excluded
Effect of ART on TB (4) 0.62 0.73 18
If the provision of ART is ≥18% more effective in preventing mortality than assumed in the model, TST_H6H36 would become dominated and should be excluded
Effect of CD4 on TB below thresh
(5) 0.22 0.20 (11)
If the reduction in the rate of incident TB as a result of increases in average CD4 count was ≤11% less than assumed, ART500 would become dominated and should be excluded
Effect of CD4 on TB below thresh
(6) 0.22 0.24 10
If the reduction in the rate of incident TB as a result of increases in average CD4 cell count is ≥10% greater than assumed, TST_H6H36 would become dominated and should be excluded
Threshold_less250 (7) 0.41 0.47 15 If the percent of the population with CD4 < 250 cell/ml was 47% or more, TST_H6H36 would become dominated and should be excluded
ppART (8) 601 271 (55) If the annual cost of ART per patient was $271 or less, TST_H6H36 would become dominated and should be excluded
Table III.B-2 Model Outputs At Threshold Values
Incident TB Outcome
ICER Threshold
Output
Adjusted Parameter
Value Alternatives Total Cost Difference Total Cases Difference ICER
1 0.71 TST_H6 $5,860,031 255.96 22,895
ALL_H6 $5,920,860 249.04 23,775 (Ext Dom)
TST_H36 $6,058,569 $198,539 158.01 97.94 38,342 2,027
TST_H6H36 $6,235,771 $177,202 132.54 25.47 47,047 6,957
ALL_H36 $6,770,294 134.19 50,453 (Dominated)
ART350 $7,758,404 139.8 55,496 (Dominated)
ART500 $9,861,720 $3,625,949 131.12 1.42 75,211 2,547,848
2 0.26 TST_H6 $5,868,364 291.57 20,127
ALL_H6 $5,930,545 290.41 20,421 (Ext Dom)
TST_H36 $6,063,988 $195,624 181.16 110.41 33,474 1,772
TST_H6H36 $6,240,978 $176,990 154.78 26.37 40,320 6,711
ALL_H36 $6,777,110 163.17 41,535 (Dominated)
ART350 $7,780,576 234.22 33,218 (Dominated)
ART500 $9,867,042 $3,626,064 153.78 1 64,164 3,608,044
All-cause Mortality Outcome
3 0.44 TST_H6 $5,852,166 397 14,759
ALL_H6 $5,915,713 384 15,398 (Ext Dom)
TST_H36 $6,044,978 $192,812 320 77 18,890 2,520
TST_H6H36 $6,221,803 $176,825 316 4 19,663 49,438
ALL_H36 $6,746,951 459 14,699 (Dominated)
ART350 $7,766,938 354 21,917 (Dominated)
ART500 $9,866,774 $3,644,970 316 0 31,199 20,552,872
4 0.73 TST_H6 $5,888,719 241 24,397
ALL_H6 $5,951,707 231 25,719 (Ext Dom)
TST_H36 $6,082,522 $193,803 161 81 37,860 2,401
TST_H6H36 $6,259,357 157 39,858 (Ext Dom)
ALL_H36 $6,804,328 215 31,575 (Dominated)
ART350 $7,814,967 132 59,202 (Ext Dom)
ART500 $9,915,157 $3,832,635 81 79 121,811 48,354
5 0.20 TST_H6 $5,871,061 316 18,561
ALL_H6 $5,934,319 305 19,443 (Ext Dom)
TST_H36 $6,064,386 $193,325 238 79 25,520 2,457
TST_H6H36 $6,241,216 $176,830 234 4 26,668 49,151
ALL_H36 $6,776,611 333 20,342 (Dominated)
ART350 $7,790,266 246 31,620 (Dominated)
ART500 $9,883,714 $3,642,498 234 0 42,234 239,418,214
6 0.24 TST_H6 $5,877,098 291 20,217
ALL_H6 $5,940,264 280 21,216 (Ext Dom)
TST_H36 $6,070,587 $193,488 211 79 28,727 2,438
TST_H6H36 $6,247,418 208 30,077 (Ext Dom)
ALL_H36 $6,786,087 293 23,167 (Dominated)
ART350 $7,800,714 198 39,398 (Ext Dom)
ART500 $9,905,242 $3,834,656 130 82 76,474 46,882
7 0.47 TST_H6 $6,852,241 294 23,330
ALL_H6 $6,918,457 284 24,363 (Ext Dom)
TST_H36 $7,037,297 $185,057 224 70 31,418 2,654
TST_H6H36 $7,215,893 221 32,684 (Ext Dom)
ALL_H36 $7,757,275 315 24,638 (Dominated)
ART350 $7,796,494 218 35,841 (Ext Dom)
ART500 $9,896,548 $2,859,251 172 52 57,630 54,710
8 271 TST_H6 $4,024,304 301 13,368
ALL_H6 $4,087,470 290 14,086 (Ext Dom)
TST_H36 $4,217,765 $193,461 222 79 19,004 2,446
TST_H6H36 $4,394,587 218 20,127 (Ext Dom)
ALL_H36 $4,935,624 309 15,965 (Dominated)
ART350 $5,251,186 218 24,140 (Ext Dom)
ART500 $6,657,279 $2,439,514 172 50 38,767 48,576
These findings suggest the results of our analysis were stable and policy recommendations within the bounds specified in the sensitivity analysis.
The threshold analysis should assist policy-makers and practitioners in determining how applicable these recommendations are for other TB- endemic settings.
REFERENCES
1. Akolo, C., et al., Treatment of latent tuberculosis infection in HIV infected persons. Cochrane Database Syst Rev, 2010(1): p. CD000171.
2. Badri et al. Effect of highly active antiretroviral therapy on incidence of tuberculosis in South Africa: a cohort study. Lancet 2002.
3. CAUSAL. Impact of antiretroviral therapy on tuberculosis incidence among HIV-positive patients in high-income countries. HIV-CAUSAL Collaboration. Clin Infect Dis. 2012 May;54(9):1364-72. Epub 2012 Mar 28. PMID:22460971
4. Chaisson talk at WHO STOP TB satellite meeting on TB-HIV at XIX International AIDS Conference, Washington, DC, 2012.
5. Cohen MS et al. N Engl J Med. 2011 Aug 11;365(6):493-505. Epub 2011 Jul 18. Prevention of HIV-1 infection with early antiretroviral therapy.
6. Egger et al. ART Cohort Collaboration. Prognosis of HIV-1-infected patients starting highly active antiretroviral therapy: a collaborative analysis of prospective studies. Lancet 2002
7. French et al. Immunologic And Clinical Stages in HIV-1–Infected Ugandan Adults Are Comparable and Provide No Evidence of Rapid Progression but Poor Survival With Advanced Disease. JAIDS 1999
8. Girardi et al. Incidence of Tuberculosis among HIV-infected patients receiving highly active antiretroviral therapy in Europe and North America. Antiretroviral Therapy Cohort Collaboration. Clin Infect Dis. 2005 Dec 15;41(12):1772-82. Epub 2005 Nov 11. PMID:16288403 9. Gupta A et al. Tuberculosis incidence rates during 8 years of follow-up of an antiretroviral treatment cohort in South Africa: comparison
with rates in the community. PLoS One. 2012;7(3):e34156. doi: 10.1371/journal.pone.0034156. Epub 2012 Mar 30. PMID:22479548 10. Hoggs et al. Rates of Disease Progression by Baseline CD4 Cell Count and Viral Load After Initiating Triple-Drug Therapy. JAMA 2001 11. Holmes et al. CD4 Decline and Incidence of Opportunistic Infections in Cape Town, South Africa: Implications for Prophylaxis and
Treatment. JAIDS 2006.
12. Johnson, J.L., et al., Duration of efficacy of treatment of latent tuberculosis infection in HIV-infected adults. AIDS, 2001. 15(16): p. 2137- 47.
13. Lawn et al. Short-term and long-term risk of tuberculosis associated with CD4 cell recovery during antiretroviral therapy in South Africa.
AIDS 2009, 23:1717–1725
14. May et al. ART Cohort Collaboration. Prognosis of HIV-1-infected patients up to 5 years after initiation of HAART: collaborative analysis of prospective studies AIDS 2007 21:1185-1197
15. Miranda et al. Impact of antiretroviral therapy on the incidence of tuberculosis: the Brazilian experience, 1995-2001. PLoS One. 2007 Sep 5;2(9):e826. PMID:17786198
16. Moreno et al. Incidence and risk factors for tuberculosis in HIV-positive subjects by HAART status. Int J Tuberc Lung Dis. 2008 Dec;12(12):1393-400
17. Rangaka, M.X., et al., Randomized controlled trial of isoniazid preventive therapy in HIV-infected persons on antiretroviral therapy.
Abstract THLBB03 in AIDS 2012. 2012: Washington, DC USA.
18. Samandari, T., et al., 6-month versus 36-month isoniazid preventive treatment for tuberculosis in adults with HIV infection in Botswana: a randomised, double-blind, placebo-controlled trial. Lancet, 2011. 377(9777): p. 1588-98.
19. Schim van der Loeff et al. Mortality of HIV-1, HIV-2 and HIV-1/HIV-2 duallyinfected patients in a clinic-based cohort in The Gambia. AIDS 2002
20. Sterne et al. When to Start Consortium. Timing of initiation of antiretroviral therapy in AIDS-free HIV-1-infected patients: a collaborative analysis of 18 HIV cohort studies. Lancet 2009
21. Stinnett, A.A. and J. Mullahy, Net health benefits: a new framework for the analysis of uncertainty in cost-effectiveness analysis. Medical Decision Making, 1998. 18(2 Suppl): p. S68-80.
22. T.Teferi, et al. Effect of primary isoniazid preventive therapy on tuberculosis incidence rate among HIV-positive adults enrolled in HIV care in northern Namibia: a retrospective cohort study. 19th International AIDS Conference: Abstract no. WEPE066