Table S1. Number of eligible individuals by cohort, HIV-CAUSAL and CNICS Collaborations,

Transcription

1 Cohort Table S1. Number of eligible individuals by cohort, HIV-CAUSAL and CNICS Collaborations, No. of antiretroviraltherapy naïve individuals No. initiated cart regimen in 2000 or later No. with confirmed virologic suppression (two consecutive HIV RNA 200 copies/ml) No. with confirmed virologic suppression within 12 months of initiating cart FHDH 26,696 14,717 10,802 8,742 8,243 VACS 11,492 6,895 4,856 3,401 3,229 CHIC 20,225 12,272 10,347 9,064 8,310 SHCS 4,110 2,913 2,649 2,503 2,329 PISCIS 6,333 4,453 3,736 3,283 3,130 GEMES PRIMO 1,300 1, SEROCO CORIS 5,526 3,215 2,542 2,253 2,146 CORIS-MD ATHENA 9,657 6,209 5,612 5,325 4,706 UKREG 1,771 1, AMACS 2,468 1,576 1,220 1, AQUITAINE 1, CNICS 7,285 5,715 4,438 3,864 3,517 All Cohorts 100,120 61,552 49,166 42,026 39,029 No. 18 years or older, no history of AIDS, CD4 cell count within the previous three months, and no pregnancy (when information was available) Appendix 2. 1

2 To estimate the inverse probability weights, we fit a pooled multinomial regression model for monitoring (a 3-level categorical variable) in the original, unexpanded study population. In a given month, monitoring took the value 0 if no CD4 or RNA measurement was recorded, 1 if either a CD4 or RNA measurement was recorded but not both, and 2 if both a CD4 and RNA measurement were recorded. The model included the previously listed covariates as well as the most recent measurement of the following time-varying covariates: CD4 cell count (restricted cubic splines with 5 knots at 200, 350, 500, 650, and 1000 cells/µl), HIV RNA ( 200, , 1,000-9,999, 10,000 copies/ml), diagnosis of an AIDS-defining illness (when the outcome was all-cause mortality), proportion of months of follow-up from baseline to the current observation with a CD4 cell count measurement (<1/9, 1/9 - <1/6, 1/6 - <1/3, 1/3), and months since the last CD4 cell count measurement (0-1, 2-3, 4). During months in which a CD4 cell count or HIV RNA measurement was not recorded, we carried forward the most recently recorded measurement. Cut points were chosen based on the distribution of the data. Because of the dynamic nature of the strategies under consideration, we computed partially stabilized weights and weights corresponding to monitoring strategies where individuals are monitored with a uniform probability during the grace period. The uniform weights performed best in a check of the balance of baseline CD4 cell count across the monitoring strategies in the pseudopopulation created by the inverse probability weights and were therefore used in the primary analysis (Appendix Table 2) 21. The weights were truncated at the strategy-specific 99 th percentile (approximately 31 for the 3-month strategy, 90 for the 6-month strategy, and 183 for the 9-12 month strategy), but truncation had little effect on the estimates. 2

3 Table S2. Contributions to the weights at different time points for strategies with monitoring at the end of the grace period (partially stabilized) and strategies with uniform monitoring, HIV-CAUSAL and CNICS Collaborations Time point Before grace period j<0 During grace period 0 j<m End of grace period j=m Monitor Unstabilized None 1 p_d_0 Double 1 1 p_d_1 None 1 1 p_d_1 Double 1 p_d_2 Partially stabilized* p_w_n p_d_0 p_w_n 1 p_d_1 p_w_n 1 p_d_1 p_w_n p_d_2 Type of weights Uniform monitoring** (used in primary analysis) 1 p_d_0 1/(m + 1 j) p_d_2 1 1 [ m + 1 j ] 1 p_d_2 1/(m + 1 j) p_d_2 Product of partially stabilized and uniform monitoring p_w_n p_d_0 p_w_n 1/(m + 1 j) p_d_2 1 p_w_n (1 [ m + 1 j ]) 1 p_d_2 p_w_n 1/(m + 1 j) p_d_2 *Corresponds to strategies where monitoring occurs at the end of the grace period **Corresponds to strategies where monitoring occurs with uniform probability during the grace period m, length of grace period j, position in grace period (month) p_d_2, conditional probability of having CD4 and RNA monitored p_d_1, conditional probability of having one measurement p_d_0, conditional probability of not being monitored p_w_n, conditional probability of not being censored, estimated from a model fit in the expanded population after censoring 3

4 Table S3. Association of prognostic factors with CD4 cell count monitoring, HIV-CAUSAL and CNICS Collaborations Odds ratio (95% CI)* Baseline characteristic CD4 cell count (cell/µl) <200 1 (reference) 200 to < (1.04, 1.07) 350 to < (1.08, 1.11) (1.09, 1.13) Sex Male 1 (reference) Female 1.05 (1.03, 1.06) Race White 1 (reference) Black 0.96 (0.94, 0.97) Other/unknown 1.05 (1.03, 1.06) Age (years) <35 1 (reference) (1.05, 1.07) > (1.13, 1.15) Origin North America or Western Europe 1 (reference) Sub-Saharan Africa 0.96 (0.94, 0.97) Other 0.94 (0.93, 0.96) Unknown 0.96 (0.94, 0.98) Acquisition Group Heterosexual 1 (reference) Homo/bisexual 1.04 (1.03, 1.05) Injection Drug User 1.00 (0.98, 1.02) Other/Unknown 1.00 (0.98, 1.02) Calendar Year (reference) (0.82, 0.84) (0.69, 0.70) (0.52, 0.54) Months to Suppression (reference) (0.90, 0.92) (0.83, 0.85) Years since HIV diagnosis <1 1 (reference) 1- < (0.95, 0.97) 5 or more or unknown 0.94 (0.93, 0.95) Time-varying characteristic Most recent CD4 cell count (cell/µl) <200 1 (reference) 200 to < (0.84, 0.88) 350 to < (0.76, 0.79) (0.70, 0.73) Most recent HIV RNA (copies/ml) 4

5 200 1 (reference) (1.18, 1.24) 1,000-9, (1.08, 1.14) 10, (1.14, 1.19) Diagnosis of AIDS-defining illness no 1 (reference) yes 1.12 (1.09, 1.16) Average proportion of months with a CD4 measurement <1/9 1 (reference) 1/9 to < 1/ (0.78, 0.85) 1/6 to < 1/ (1.27, 1.37) 1/ (1.89, 2.05) Time since last CD4 cell count at previous month 0 or 1 months 1 (reference) 2 or 3 months 2.81 (2.75, 2.86) 4 or more months 2.30 (2.25, 2.34) Average proportion of months with an RNA measurement <1/9 1 (reference) 1/9 to < 1/ (1.70, 1.94) 1/6 to < 1/ (2.28, 2.58) 1/ (3.01, 3.41) Time since last RNA measurement at previous month 0 or 1 months 1 (reference) 2 or 3 months 2.82 (2.77, 2.88) 4 or more months 3.01 (2.95, 3.08) *The odds ratios were obtained by fitting a pooled logistic regression model in the original, unexpanded population for CD4 monitoring (yes/no). Each of the listed baseline and time-varying covariates, as well as follow-up time, were included in the model. An analysis of prognostic factors and RNA monitoring produced similar results. 5

6 Table S4. Treatment switching by monitoring strategy, HIV-CAUSAL and CNICS Collaborations Monitoring strategy Outcomes, cases Unadjusted hazard ratio (95% CI) Personmonths Baselineadjusted hazard ratio (95% CI) IPW-adjusted hazard ratios (95% CI) 1, , (reference) 1.00 (reference) 1.00 (reference) 1, , (0.90, 0.98) 0.93 (0.90, 0.97) 0.81 (0.67, 0.96) 1, , (0.89, 0.96) 0.92 (0.88, 0.95) 0.91 (0.67, 1.24) IPW, inverse probability weighted. Table S5. Additional results for clinical outcomes by monitoring strategy, HIV-CAUSAL and CNICS Collaborations

7 Outcome and monitoring strategy Outcomes, cases Unadjusted hazard ratio (95% CI) Personmonths Baselineadjusted hazard ratio (95% CI) Baselineadjusted hazard ratio (95% CI) with time-varying covariates in outcome model* All-cause mortality , (reference) 1.00 (reference) 1.00 (reference) , (0.82, 1.19) 0.96 (0.80, 1.15) 0.85 (0.66, 1.09) , (0.84, 1.21) 0.96 (0.80, 1.14) 0.84 (0.66, 1.06) AIDS-defining illness or death , (reference) 1.00 (reference) 1.00 (reference) , (0.95, 1.18) 1.00 (0.90, 1.12) 0.98 (0.84, 1.16) , (0.96, 1.18) 0.99 (0.89, 1.10) 0.97 (0.84, 1.13) *These hazard ratios were obtained by including the baseline and time-varying covariates in a model for the outcome, rather than by fitting a weighted regression model for the outcome conditional on the baseline covariates. Table S6. Mortality by monitoring strategy using alternative monitoring strategies and definitions of virologic suppression, HIV-CAUSAL and CNICS Collaborations

8 Monitoring strategy Deaths, cases Person-months IPW-adjusted mortality hazard ratios (95% CI) No change in monitoring frequency while HIV RNA >200 copies/ml, CD4 200 cells/µl, or after diagnosis of an AIDS-defining illness , (reference) , (0.64, 3.29) , (0.75, 5.03) 3-6 months while HIV RNA >200 copies/ml, CD4 200 cells/µl, or after diagnosis of an AIDS-defining illness , (reference) , (0.41, 1.71) , (0.54, 2.97) Baseline defined as two consecutive HIV RNA 50 copies , (reference) , (0.49, 2.55) , (0.46, 1.97) Baseline defined as one HIV RNA 200 copies , (reference) , (0.52, 1.35) , (0.46, 1.74) IPW, inverse probability weighted Table S7. Virologic failure by monitoring strategy using alternative monitoring strategies and definitions of virologic suppression and failure, HIV-CAUSAL and CNICS Collaborations

9 Monitoring strategy and definition of virologic failure No. RNA>200 / No. RNA>50 (no. with RNA measurements at 18 ± 2 months) IPW-adjusted virologic failure risk ratios (95% CI) Failure defined as RNA>200 Failure defined as RNA>50 Primary analysis (baseline defined as two consecutive HIV RNA 200 copies) 193 / (reference) 1.00 (reference) (4,497) 30 / (0.46, 1.19) 0.64 (0.49, 0.84) (919) 33 / 79 (447) 2.35 (1.56, 3.54) 1.18 (0.88, 1.59) No change in monitoring frequency while HIV RNA >200 copies/ml, CD4 200 cells/µl, or after diagnosis of an AIDS-defining illness 193 / (reference) 1.00 (reference) (4,497) 41 / (0.86, 1.86) 0.89 (0.69, 1.14) (687) 21 / 37 (183) 2.40 (1.53, 3.76) 1.21 (0.88, 1.66) 3-6 months while HIV RNA >200 copies/ml, CD4 200 cells/µl, or after diagnosis of an AIDS-defining illness 283 / (reference) 1.00 (reference) (4,933) 78 / (0.78, 1.71) 0.77 (0.60, 0.97) (1,250) 63 / (2.15, 4.80) 1.36 (1.02, 1.81) (714) Baseline defined as two consecutive HIV RNA 50 copies 137 / (reference) 1.00 (reference) (3,772) 24 / (0.58, 1.59) 0.91 (0.66, 1.25) (822) 18 / (1.39, 3.69) 1.69 (1.19, 2.40) (346) Baseline defined as one HIV RNA 200 copies 230 / (reference) 1.00 (reference) (4,550) 60 / (0.77, 1.45) 0.95 (0.79, 1.15) (896) 49 / 110 (527) 1.61 (1.11, 2.35) 1.20 (0.97, 1.49) IPW, inverse probability weighted Figure S1. Example of censoring process for an individual with CD4 cell count >200 cells/µl monitored for the first time after baseline at the 9 th month of follow-up. 9

10 Replicate 1: monitor once every 3±1 months Replicate 2: monitor once every 6±1 months month month censor monitor follow-up time Replicate 3: monitor once every 9-12 ±1 months month This individual had data consistent with strategies (i)-(iii) until the fourth month of follow-up, strategies (ii)-(iii) until the 7 th month of follow-up, and strategies (iii) until the 9 th month of follow-up. Therefore the replicate following strategy (i) was censored at the 4 th month of follow-up and the replicate following strategy (ii) at the 7 th month. After the 9 th month of follow-up only the replicate assigned to strategy (iii) will remain uncensored (for as long as the individual s data remain consistent with the strategy). Importantly, replicates can also be censored if they are monitored more frequently than indicated by their strategy. For example, if replicate 3 is monitored again at month 10, he or she would be censored from the 9-12 month strategy at that time. Figure S2. 18-month survival and AIDS-free survival curves by monitoring strategy, HIV-CAUSAL and CNICS Collaborations

11 AIDS-free survival Survival Every 3 months Every 6 months 0.98 Every 9-12 months Months since baseline Every 3 months Every 6 months Every 9-12 months Months since baseline Figure S3. 18-month mean CD4 cell curves by monitoring strategy excluding (a) intravenous drug users and individuals with an unknown mode of transmission and (b) individuals presenting late to care (initiating cart at CD4 cell count <200 cells/µl), HIV-CAUSAL and CNICS Collaborations

12 Mean CD4 cell count (cells/µl) Mean CD4 cell count (cells/µl) (a) Every 3 months Every 6 months Every 9-12 months Months since baseline (b) Every 3 months Every 6 months Every 9-12 months Months since baseline 12