Supplementary Appendix

This appendix has been provided by the authors to give readers additional information about their work.

Supplement to: Le T, Wright EJ, Smith DM, et al. Enhanced CD4+ T-cell recovery with earlier HIV-1 antiretroviral therapy. N Engl J Med 2013;368:218-30. DOI: 10.1056/NEJMoa1110187

Le et al

1

Supplementary Appendix

This appendix has been provided by the authors to give readers additional information about

their work.

Supplement to:

Enhanced CD4+ T-Cell Recovery with Earlier HIV-1 Antiretroviral Therapy

Le T, Wright E, Smith DM, He W, Catano G, Okulicz J, Young J, Clark RA, Richman D, Little

S, and Ahuja SK.

Le et al

2

Table of Contents Page

Statistical Methods 3

Supplementary Tables

Table S1 8

Table S2 9

Table S3 12

Table S4 13

Table S5 14

Supplementary Figures

Figure S1 15

Figure S2 17

Figure S3 18

Figure S4 20

Supplementary References 21

Le et al

3

SUPPLEMENTARY STATISTICAL METHODS

(A) Nonlinear generalized estimating equations (GEE) modeling

Nonlinear generalized estimating equations (GEE) were used to model the cohort-level time-

trends (trajectory) of the squared root transformed CD4+ count and log10 viral load (VL). The

model was derived using exchangeable structure. For the nonlinear GEEs, we used spline-

smooth based on 4-month knots and then constructed 95% point-wise confidence bands. The

goodness-of-fit of the GEE models were assessed using the Wald statistic. Since the spline-

smoothed curves indicate the population-level dynamics of CD4+ T-cell counts, they do not

reflect the cross-sectional measurements of these parameters at the level of an individual. These

time-trends were derived using STATA.

The CD4+ and VL trajectories in study set 1 (i.e., during the therapy-free follow-up of

participants) were computed from the EDI for a maximum of 48 months. Participants were

censored on the last day of ART-free follow-up if therapy was initiated prior to 48 months, or

were censored at 48 months if they remained therapy-naïve until this time point.

The trajectories for CD4+ counts and VL in study set 2 were computed from the date of ART

initiation, anterograde for a maximum of 48 months after the initiation of ART (i.e., during the

time period the participants were on ART) and retrograde for a maximum of 16 months towards

the time of entry into the study cohort (i.e., during the time period the participants were therapy-

naïve). For those in study set 2, CD4+ and VL trajectories were stratified according to the CD4+

and viral load at ART initiation, as well as whether ART was initiated earlier or later. For these

analyses, participants were censored on the day they no longer maintained viral load suppression

Le et al

4

(defined as two consecutive viral load >75 copies/ml, at least 14 days apart), or on the last day of

follow-up if it was shorter than 48 months on ART, or at 48 months, if the patient maintained

viral load suppression on ART until this time point.

(B) Linear mixed-effects modeling: To mitigate the possibility that the GEE estimates may be

biased by the potential differences in the duration of follow-up after commencing ART, we used

linear mixed-effect models with random intercept and random slope (to account for the within-

subject correlation over time) to determine group-specific differences in CD4+ counts and viral

load after initiation of ART (i.e., with higher vs. lower pre-ART CD4+ or log10 VL as fixed

effects). For these analyses, we used the restricted maximum likelihood estimation to estimate

the least square means of changes in CD4+ counts and viral load during therapy. We used the

unstructured covariance because when evaluated by Akaike Information Criterion (AIC) this

method gave the lowest AIC when compared with other covariance structures [e.g., structure of

variance components and first-order autoregressive or compound symmetry]. The statistical

significance between the least square means of the CD4+ counts of the study groups (computed

by linear mixed-effects models) was determined. For ease in interpretation of the results derived

using square-root transformed CD4+ counts was converted to absolute CD4+ counts.

(C) Logistic regression, Cox proportional hazard modeling and Kaplan-Meier methods

Logistic regression was used to determine the likelihood (odds ratio) of attaining the primary and

secondary CD4+ endpoints on ART. A proportional-hazards Cox regression model, stratified

according to CD4+ count at ART initiation and timing of ART, was used to compare the rate of

attainment of the primary and secondary CD4+ endpoints on ART. All proportionality

Le et al

5

assumptions were assessed by checking the log cumulative survival plots for CD4+ count at

ART initiation (higher vs. lower) and timing of ART (earlier vs. later). We also examined for the

interaction between the CD4+ strata at time of ART initiation (higher vs. lower) and timing of

ART (earlier vs. later) on attainment of the primary and secondary CD4+ endpoints on therapy.

The model with interaction did not fit the data better than a model without interaction. As the

interaction was not a significant predictor, it was not included in the model. The Kaplan-Meier

method was used to estimate the cumulative probability of achieving the primary and secondary

CD4+ endpoints on ART.

By univariate logistic regression and Cox proportional modeling, we determined the covariates

that influenced the likelihood and rate of attaining the primary and secondary CD4+ endpoints on

ART. The covariates with a significant association were included in the multivariate models.

(D) Subgroup analyses

The objective of this study was to determine the conjoint influence of the timing of ART relative

to the EDI and CD4+ count at ART initiation on recovery of CD4+ T-cell counts. Thus, we

examined subgroups classified based on pre-defined pre-ART CD4+ and timing of ART relative

to the EDI.

The first subgroups were the CD4+ counts at ART initiation dichotomized according to whether

they were higher versus lower i.e., ≥500 or < 500 cells/mm3, respectively. This CD4+ threshold

was chosen because some guidelines panels recommend initiation of ART when the CD4+

counts fall to below 500 cells/mm3 (Reference 1).

Le et al

6

The next subgroups defined a priori were the timing of ART relative to the EDI. In this study,

we conceptualized that after infection there is a narrow restorative time window wherein the

immune system may be strategically poised for recovery, the likelihood and rate of which may

be augmented by commencement of ART earlier. Inspection of the trajectory of CD4+ counts in

therapy-naïve persons (Figure 2A) indicated that there were two inflection points in the CD4+

counts in therapy-naïve participants: (i) the 4 month time point post-EDI which marks when

CD4+ counts peak spontaneously, and (ii) the 12 month time point from EDI that marks when

CD4+ counts reach the threshold of 500 cells/mm3, which we found is similar to the cohort-level

CD4+ count present soon after HIV infection. In addition to these observed inflection points, we

calculated the time interval between baseline (study entry) and peak CD4+ counts, which was 3.5

(interquartile range, 2.6-5.2) months. On the basis of these calculations and the observed

trajectories, we termed the four months from EDI as the ‘restorative time window’.

These inflection points defined three time-windows of 0-4, >4-12 and >12 months from EDI. In

this schema, initiation of ART within 4 months from EDI was classified as ‘earlier’ initiation of

ART whereas initiation of ART after 4 months from the EDI was classified as ‘later’ initiation of

ART (i.e., >4 months from ART).

To mitigate possible confounding due to varying lengths of ART exposure, we also examined the

subset in study set 2 who received ART for ≥18 months (n=164). This duration was chosen

because 75 percent of persons who attained the primary post-ART CD4+ endpoint did so within

18.6 months of ART-initiation (time from ART-initiation to first CD4+ of ≥ 900 on ART was

6.8 (interquartile range, 1.9-18.6) months; Table 1).

Le et al

7

One post-hoc subgroup analyses was conducted wherein we asked whether initiation of ART

during the very earliest phases of infection associated with higher rates of CD4+ recovery. For

these analyses, participants who commenced ART earlier (i.e., among those who commenced

ART within four months from EDI) were stratified into two groups designated as Earlier I and II,

according to whether ART was commenced in the proximal or distal half of the 4-month time

window from EDI, respectively. The median was used to define proximal versus distal. For

participants who commenced ART earlier, the median time of ART initiation from EDI was 11.9

weeks (2.8 months). Thus, Earlier I comprised participants who commenced ART within 11.9

weeks (2.8 months) from EDI, whereas Earlier II comprised participants who commenced ART

between 11.9 weeks (2.8 months) to 17.2 weeks (4 months) from EDI.

We feel that we have reported the results of the subgroup analyses transparently and according to

the recommendations provided by Wang, et al.2

(E) Significance values

Reported P values are two-sided and set at the 0.05 significance value. The models were adjusted

for covariates but not adjusted for multiple comparisons in the prespecified subgroup analyses.

Le et al

8

Suppl. Table 1. Brief Summary of the Algorithm for Computing the Estimated Date of HIV-1 Infection* Class Definitions for stage of HIV-1 infection n (%) A1.0 If there is a first positive RNA† and negative enzyme immunoassay (EIA)

within 7 days of the first positive RNA, and no prior positive/indeterminate western blot (WB), then EDI = first positive RNA date − 11 days. (~Fiebig Stages I-II‡)

82 (17.5)

A2.0 If there is an indeterminate WB within 7 days of the first positive RNA, then EDI = first positive RNA date − 20 days. (~Fiebig Stages III-IV‡)

10 (2.1)

A3.0 If the last negative EIA or negative/indeterminate WB occurred ≤ 30 days before the first positive WB (with associated positive RNA), then EDI = midpoint of the positive WB date and the negative EIA or negative/indeterminate WB date (earlier of two) − 19 days. (~Fiebig Stage IV‡)

23 (4.9)

A3.1 If the first positive WB p31/32 band is absent, then EDI = first positive WB date − 89 days. (~Fiebig Stage V‡)

48 (10.3)

E1.0A If there is a detuned EIA (dtEIA) consistent with infection of ~3 mo within 30 days of the first positive WB and CD4 count > 200 or CD4% > 14 within 30 days of the first positive WB, then EDI = first dtEIA date − 70 days§. (Fiebig VI‡)

193 (41.2)

E1.0B If there is a dtEIA consistent with infection of ~3-6 mo within 30 days of the first positive WB and CD4 count > 200 or CD4% > 14), then EDI = first dtEIA date − 133 days§. (Fiebig VI‡)

55 (11.8)

E1.0C If there is a dtEIA consistent with infection of ~6-12 mo within 30 days of the first positive WB and CD4 count > 200 or CD4% > 14, then EDI = first dtEIA date − 170 days§. (Fiebig VI‡)

26 (5.6)

E2.0 If there is a first positive WB and a negative EIA within 365 days participant enrollment (Day 0), then EDI = midpoint between the last negative EIA and Day 0. (Fiebig VI‡)

31 (6.6)

* Each rule applied sequentially until EDI criteria satisfied. † Positive RNA was defined as a NAT/viral load exceeding the detectable level for a given assay. ‡ Indexed to algorithm from Fiebig et al.3 Note: no endpoint was defined for stage VI by Fiebig. § dtEIA threshold from Kothe et al.4

Total: 468

Le et al

9

Table S2. CD4+ T-cell Counts in Demonstrated or Presumed HIV Seronegative Individuals of European Descent and African Americans. Summary Statistics for CD4+ T-cell Counts*

Population No. of Reports

(No. of Individuals)†

CD4+ T-cell Counts (cells/mm3) Weighted Mean (95%

CI)‡ Median (IQR)§ Range¶

European 16 (11037) 1011 (1005-1017) 940 (834-1030) 796-1109 Mixed USA 8 (4083) 1006 (995-1018) 998 (882-1027) 771-1075 African American 2 (1006) 1077 (1054-1099) 1078 (1055-1100) 1055-1100 Combo 25 (16,126) 1014 (1008-1019) 952 (840-1036) 771-1109 Reports from which the summary statistics were derived

Race/ Ethnicity

Reason for

Study

Age (years), Mean

(Range) Gender

HIV Status**

n

Mean (SD or SE, Range) or Median (IQR) of Absolute

CD4+ T-cell Count (cells/mm3)††

Ref

European Austra-

lian GN Mean: 15 (Range: 10-37)

52% Female PN 2538 Mean: 1030 (SD: 270, Range:

210 -2530) 5

Austra-lian GN Mean: 14

(Range: 10-22) 48% Female PN 592 Mean: 1040 (SD: 300, Range:

200 -2800) 5

UK GN Mean: 50 (Range: 19-80) Female PN 396 Mean: 870 (SD: 330, Range:

390 -2380) 5

UK/

Belgium RF (Range: 7-17) 22% Female PN 22 Median: 800 (IQR: 700-1100) 6

UK/ Belgium RF (Range: 18-70) 55%

Female PN 101 Median: 800 (IQR: 700-1100) 6

Sweden RF (Range: 20-39) Combined PN 75 Mean: 1020 (SE: 39) 7 Sweden RF (Range: 20-39) Male PN 34 Mean: 929 (SE: 58) 7 Sweden RF (Range: 20-39) Female PN 41 Mean: 1096 (SE: 52) 7 Sweden RF (Range: 40-59) Combined PN 76 Mean: 834 (SE: 35) 7 Sweden RF (Range: 40-59) Male PN 39 Mean: 718 (SE: 32) 7 Sweden CT (Range: 40-59) Female PN 37 Mean: 956 (SE: 59) 7 Sweden RF (Range: 60-79) Combined PN 68 Mean: 796 (SE: 37) 7 Sweden RF (Range: 60-79) Male PN 36 Mean: 722 (SE: 47) 7 Sweden RF (Range: 60-79) Female PN 32 Mean: 880 (SE: 55) 7

Germany RF (Range: 19-85) Combined PN 100 Median: 870 (IQR: 490-1640) 8 Germany RF Male PN 50 Median: 830 8 Germany RF Female PN 50 Median 930 8

Switzer-

land RF Mean: 50 (Range: 24-68) Combined PN 70 Median: 691 (IQR: 309-1139) 9

Switzer-land RF Mean: 49

(Range: 23-70) Male PN 44 Median: 656 (IQR: 336-1126) 9

Switzer-land RF Mean: 51

(Range: 25-70) Female PN 26 Median: 761 (IQR: 314-1270) 9

Italy RF Mean: 37 (Range: 18-70) Combined PN 946 Mean: 940 (Range: 493-1666) 10

Italy RF - Male PN 532 Mean: 902 10

Le et al

10

Italy RF - Female PN 436 Mean: 989 10

England RF (Range: 11-79) Combined Neg. 600 Mean: 830 (SD: 288, Range: 410-1540)

11

England RF Mean: 30 (Range: 19-41)

Control Male Neg. 50 Mean: 840 (SD: 285) 11

England RF Mean: 31 (Range: 20-49)

Control Female Neg. 50 Mean: 1050 (SD: 377) 11

UK RF - Combined Neg. 234 Mean: 831 12

UK RF Mean: 31 (Range: 19-67) Male Neg. 91 Mean: 754 12

UK RF Mean: 28 (Range: 17-58) Female Neg. 195 Mean: 865 12

UK RF Mean: 33 (SD: 6, Range: 23-44) Male PN 32 Mean: 954 (SD: 272, Range:

460-1430) 13

USA RF Mean: 38 (Range: 31-45) Male PN 3467 Mean: 1100 (SD: 400) 14

France CT - - Neg. 61 Mean: 807 (SD: 378) 15 France CT - Male Neg. 16 Mean: 1109(SD: 399) 15

France CT - - PN 12 Mean: 844 (SD: 247) 16

Italy GN Mean: 41 (SD: 14.9) Male PN 263 Mean: 903 (SD: 308) 17

Italy GN Mean: 40 (SD: 15.9) Female PN 205 Mean: 1018 (SD: 319) 17

Dutch RF - - Neg. 1356 Mean: 993 (SD: 319, Range: 509-1761)

18

Dutch RF - - Neg. 678 Median: 930 (IQR: 490-1750) 19

Dutch CT (Range: 18-64) ~ 48% Female PN 59 Median: 908 (IQR: 513-1606) 20

Mixed USA Baltimore

/LA RF (Range: 18-60) - Neg. 2787 Mean: 1017 (SD: 329) 21

USA CT - - PN 19 Mean: 839 (SD: 276) 22

New

Mexico CT (Range: 21-53) Combined PN 20 Mean: 1075 (SD: 586) 23

New Mexico CT Mean: 76

(Range: 67-88 ) Combined PN 25 Mean: 924 (SD: 416) 23

New York CT Mean: 39

(SD: 6.7) - Neg. 34 Mean: 1013 (SD: 274) 24

USA CT Median: 25 (IQR: 18-30)

58% Female Neg. 24 Median: 785 (IQR: 662-860) 25

Le et al

11

USA CT Median: 49 (IQR: 45-66)

54% Female Neg. 24 Median: 869 (IQR: 658-1111) 25

USA RF (Range: 20-69) Combined PN 266 Mean: 1036 (SD: 296, Range: 294-1590)

26

California CT Median: 38 (IQR: 20-58)

65% Female PN 49 Mean: 771 (Range: 326-1344) 27

U.S. Air

Force RF Mean: 49 Male Neg. 883 Mean: 982 (Range: 417-1841)

African Americans

USA CT Median: 32 Female Neg. 513 Mean: 1055 (SE:15) 28

USA RF Mean: 38 (Range: 31-45) Male PN 493 Mean: 1100 (SD: 400) 14

IQR denotes Interquartile Range; SD denotes Standard Deviation; SE denotes Standard Error, CI denotes Confidence Intervals; UK denotes United Kingdom; USA denotes United States of America; and LA denotes Los Angeles. Mixed USA refers to CD4+ T-cell count data from the USA where the number of individuals who were European American or African American was not specified. Combo refers to the combined data from persons of European descent and African Americans. * Summary statistics were derived from the mean CD4+ T-cell count values reported in the studies outlined in the lower part of the table. From these data we derived the weighted mean (95% CI) and median (IQR) of the mean CD4 + T-cell counts from presumed negative (PN) or documented HIV-seronegative (Neg.) persons. † Total number of individuals in study reports. ‡ Calculated from studies that reported the mean CD4+ counts. § Median of the mean CD4+ counts reported. ¶ Range of mean CD4+ counts reported. | Reason for Study refers to the main purpose of why the study was conducted. Genetics (GN) implies that the study was conducted to test the genetic basis of inheritance of T-cell counts. Reference (RF) means that the study was conducted to obtain a reference CD4+ value specific for the healthy population investigated. Control (CT) denotes that the participants studied served as a control for a study in which the effects of a process (e.g. aging, HIV) on CD4+ counts was examined; in this instance, only the CD4+ values from the healthy individuals were used. ** PN: presumed HIV negative; Neg., denotes HIV-negative †† Unless otherwise specified all data are either mean (SD) or median (IQR). In some studies, standard error (SE) was reported.

Le et al

12

Table S3. Univariate Analyses for the Likelihood and Rate of Attainment of a CD4+ T-cell Count of ≥ 900 or ≥ 800 cells/mm3 while on ART among 213 Participants in Study Set 2. Endpoint of CD4+ ≥ 900 Endpoint of CD4+ ≥ 800 Predictive Factor

OR*or RR† (95% CI)

P Value

OR*or RR† (95% CI)

P Value

Univariate Sex – female vs. male OR 1.35 (0.40-4.57) 0.63 0.82 (2.24-2.78) 0.75

RR 1.18 (0.52-2.69) 0.69 0.83 (0.37-1.89) 0.66 Race – Caucasian vs. other OR 2.26 (1.10-4.65) 0.03 2.77 (1.38-5.59) 0.004

RR 1.59 (0.89-2.86) 0.12 1.90 (1.12-3.21) 0.017 Age at ART-initiation – yr OR 1.02 (0.99-1.06) 0.14 1.02 (0.99-1.06) 0.17

RR 1.01 (0.99-1.03) 0.31 1.01 (0.99-1.03) 0.34 Household number – >1 vs 1 OR 0.59 (0.26-1.33) 0.20 0.72 (0.32-1.60) 0.41

RR 0.85 (0.45-1.61) 0.62 0.94 (0.55-1.60) 0.82 Education- Bachelor's degree and higher vs. No Bachelor's degree

OR 1.39 (0.78-2.47) 0.26 1.42 (0.79-2.56) 0.25 RR 1.15 (0.76-1.73) 0.51 1.17 (0.82-1.69) 0.39

Income-groups† OR 1.08 (0.93-1.26) 0.32 1.04 (0.89-1.22) 0.63 RR 1.04 (0.93-1.16) 0.49 1.01 (0.92-1.11) 0.85

HBV co-infection OR 0.80 (0.11-5.85) 0.82 0.42 (0.04-4.14) 0.46 RR 0.90 (0.22-3.69) 0.88 0.74 (0.23-2.36) 0.61

HCV co-infection OR 1.23 (0.20-7.61) 0.83 0.84 (0.14-5.23) 0.85 RR 1.18 (0.29-4.85) 0.82 0.73 (0.23-2.34) 0.60

Co-infection- score (sum of events per person)‡

OR 0.80 (0.43-1.51) 0.50 1.27 (0.67-2.39) 0.46 RR 0.97 (0.60-1.59) 0.92 1.20 (0.86 -1.69) 0.29

Risk behavior-score (sum of risks per person)§

OR 1.09 (0.35-3.47) 0.88 1.52 (0.42-5.43) 0.52 RR 0.97 (0.42-2.26) 0.94 1.18 (0.56-2.46) 0.67

* Logistic regression and † Cox proportional hazard models were used to compute the odds ratio (OR) and the rate ratio (RR), respectively for the attainment of the CD4+ endpoints designated as primary and secondary CD4+ recovery (≥900 and ≥800 cells/mm3, respectively). † Income groups defined as 6 ordinal levels of $500-$999, $1000-$1999, $2000-$2999, $3000-$3999, $4000-$7999, and > $8000 per month of household income, and were used as a surrogate for access to health care. ‡ Co-infection score defined as sum of events (=1 if having infection) of HBV, HCV, syphilis, gonorrhea or/and chlamydia per person. § Risk-behavior score defined as sum of having risk of acquiring HIV infection (count=1) by sex with men, sex with women, sex with men and women, intravenous drug use, sharps exposure, blood products, and other risk factors per person.

Le et al

13

Table S4. Sampling Frequency of CD4+ Counts while on ART According to Whether ART Was Commenced Earlier vs. Later, or at Higher vs. Lower CD4+ Counts. Characteristic No. No. of CD4+

sampling Median (IQR) CD4+ measurements per participant

Median (IQR) interval (days) of sampling frequency of CD4+ counts

Overall 213 4,027 16 (9-26) 56 (36-65) Earlier ART − 0-4 mo 97 2,336 23 (14-33) 50 (30-56) Later ART − > 4 mo. 116 1,691 12 (7-20) 61 (56-78) Higher pre-ART CD4 − ≥500 cells/mm3 89 1,857 19 (12-27) 56 (36-61) Lower pre-ART CD4 − <500 cells/mm3 124 2,170 15 (9-26) 56 (41-69) Model 1* Model 2† Linear Mixed-effect Models No. Mean (95%CI)‡ P Value Mean (95%CI)‡ P Value Earlier ART − 0-4 mo 97 Low Sampling Frequency − ≥ median 48 26.52

(25.52-27.52) 0.49

26.33 (25.02-27.63) 0.56 High Sampling Frequency − < median 49 27.02

(26.01-28.03) 26.87

(25.73-28.02) Later ART − > 4 mo. 116 Low Sampling Frequency − ≥ median 58 23.68

(22.64-24.72) 0.42

23.78 (22.49-25.06) 0.55 High Sampling Frequency − < median 58 24.29

(23.25-25.32) 24.30

(23.28-25.31) Higher pre-ART CD4 − ≥500 cells/mm3 89 Low Sampling Frequency − ≥ median 49 28.38

(27.55-29.21) 0.53

28.69 (27.74-29.65) 0.11 High Sampling Frequency − < median 40 29.99

(27.06-28.92) 27.48

(26.54-28.42) Lower pre-ART CD4 − <500 cells/mm3 124 Low Sampling Frequency − ≥ median 60 22.64

(21.08-23.48) 0.11

23.59 (22.53-24.64) 0.91 High Sampling Frequency − < median 64 23.58

(22.77-24.39) 23.51

(22.74-24.28) Cox Proportional Hazard Models RR (95%CI)§ P Value RR (95%CI)§ P Value Earlier ART − 0-4 mo − High vs. Low sampling frequency

97 1.57 (0.95-2.60)

0.079 1.50 (0.79-2.88)

0.22

Later ART − 0-4 mo − High vs. Low sampling frequency

116 1.48 (0.78-2.82)

0.23 1.56 (0.68-3.58)

0.30

Higher CD4 (≥500 cells/mm3) − High vs. Low sampling frequency

89 1.14 (0.71-1.83)

0.58 0.70 (0.37-1.30)

0.25

Lower CD4 (<500 cells/mm3) − High vs. Low sampling frequency

124 1.82 (0.84-3.93)

0.13 0.88 (0.32-2.47)

0.81

* Model 1: Unadjusted. † Model 2: Adjusted for race, age at ART initiation, gender, viral load at ART initiation, length of follow-up, time-from EDI to ART initiation, and number of CD4+ measurements. ‡ Least square means of CD4+ counts during 48 months from ART initiation and 95% confidence intervals (CI) derived by linear mixed-effect models. § Cox proportional hazard models were used to compute the rate ratio (RR) for the attainment of primary CD4+ recovery (≥900 cells/mm3) while on ART.

Le et al

14

Table S5. Likelihood and Rate of Attainment of a CD4+ Count of ≥ 900 or ≥ 800 cells/mm3 within 18 Months of ART Initiation among 164 Study Participants in Study Set 2 Who Were on ART for at least 18 Months. Multivariate Model

Endpoint of CD4+ ≥ 900 Endpoint of CD4+ ≥ 800 OR*or RR†

(95% CI) P Value OR*or RR†

(95% CI) P Value

Multivariate Model 1: Time from EDI to ART-initiation -- mo‡ OR 0.89 (0.82-0.97) 0.01 0.92 (0.86-0.99) 0.02

RR 0.91 (0.85-0.97) 0.003 0.93 (0.89-0.98) 0.007 Multivariate Model 2: Time from EDI to ART-initiation of >4 vs. ≤4 mo >4 vs. ≤4 mo OR 0.34 (0.15-0.77) 0.01 0.41 (0.18-0.93) 0.03

RR 0.42 (0.25-071) 0.001 0.55 (0.35-0.87) 0.01 Multivariate Model 3: CD4+ count <500 vs. ≥ 500 cells/mm3 at ART-initiation OR 0.06 (0.03-0.14) <0.001 0.05 (0.02-0.12) <0.001

RR 0.13 (0.07-0.24) <0.001 0.13 (0.08-0.22) <0.001 Multivariate Model 4: Initiation of ART ≤4, >4-12 and >12 mo from EDI >4-12 vs. ≤4 mo OR 0.45 (0.18-1.12) 0.09 0.48 (0.19-1.22) 0.12

RR 0.48 (0.27-0.85) 0.01 0.59 (0.36-0.98) 0.04 >12 vs. ≤4 mo OR 0.19 (0.06-0.64) 0.007 0.31 (0.10-0.95) 0.04

RR 0.30 (0.12-0.73) 0.008 0.47 (0.23-0.96) 0.04 *Logistic regression and † Cox proportional hazard models were used to compute the odds ratio (OR) and rate ratio (RR), respectively for the attainment of the primary and secondary CD4+ endpoints of ≥900 or ≥800 cells/mm3, respectively. To mitigate possible confounding due to varying lengths of ART exposure, we examined the subset in study set 2 who received ART for ≥18 months (n=164). ‡ Multivariate Models depict the odds ratio (OR) and rate ratio (RR) for the associations of the length of time from EDI to ART initiation in months (model 1); initiation of ART earlier versus later, defined as starting ART within or after 4 months following EDI, respectively (model 2); a higher versus lower CD4+ count at ART-initiation (i.e., ≥500 versus < 500 cells/mm3, respectively) (model 3); the indicated time-intervals in which ART was initiated post EDI (model 4). In models 2 and 4, the referent group is initiation of ART within 4 months from EDI. All multivariate models were adjusted for race and length of time on ART in months; models 2 and 4 were also adjusted for the CD4+ T-cell counts at ART-initiation (i.e., < or ≥500 cells/mm3) whereas model 3 was also adjusted for whether ART was commenced earlier versus later.

Le et al

15

Figure S1. Viral Load Trajectories in Study Sets 1 and 2

The panels depict trajectories for plasma HIV-1 RNA (viral load, VL) with 95% point-wise

confidence bands. Panel A depicts the trajectory for the VL from the EDI in HIV-infected

individuals during the time periods they were therapy-naïve for a maximum of 48 months from

the EDI (study set 1). The time interval during which the VL reaches nadir corresponds to the

four month interval from EDI during which CD4+ counts rise spontaneously (shown in Figure

2A in the main text). Panel B depicts the VL trajectory of participants who received ART and

suppressed VL (study set 2) computed from the time of ART initiation (at arrow) anterograde

until last day of follow-up or loss of VL suppression to a maximum of 48 months and retrograde

for a maximum of 16 months towards the time of entry into the study cohort (i.e., during the time

period participants were therapy-naïve). Panels C and D show the VL trajectories from time of

ART initiation (indicated by arrow) according to whether the pre-ART CD4+ T-cell counts were

< or ≥500 cells/mm3 in participants who commenced ART earlier (panel C) or later (panel D).

We computed by linear mixed-effects models the differences in the least square mean log10 VL

(95% confidence intervals) after attainment of VL suppression between the study groups shown

in panels C and D (i.e., between those with higher vs. lower pre-ART CD4+); the mean log10 VL

difference in panel C was 0.05 (95% CI: −0.12 to 0.22; P value = 0.6) and in panel D was 0.10

(95% CI: 0.07 to 0.26, P value = 0.3).

Le et al

16

Le et al

17

Figure S2. CD4+ T-cell Count Trajectories among Participants in Study Sets 2 Who Did vs.

Did Not Achieve Primary and Secondary CD4+ Recovery.

Panels A and B depict CD4+ trajectories for participants in study set 2 who either achieved or

never achieved at least one CD4+ count that was ≥900 (A) or ≥800 (B) cells/mm3, respectively

on ART. The differences in the least square mean CD4+ counts from time of ART initiation

between the study groups shown in panels A and B were computed by linear mixed-effects

models, and adjusted for age, sex, race, VL at ART initiation, and length of follow-up from ART

initiation; the differences (95% confidence intervals) between the absolute CD4+ T-cell counts

of the study groups in panel A were 310 (272 to 349) and in panel B were 319 (279 to 359)

cells/mm3 (P<0.001 for both comparisons).

Le et al

18

Figure S3. Rate and Extent of CD4+ Recovery after ART Initiation According to Timing of

ART and CD4+ Count at ART Initiation (Higher versus Lower).

Panel A, schema illustrating the timing of ART (earlier versus later) at lower versus higher

CD4+ counts at ART initiation (< or ≥500 cells/mm3, respectively). Earlier versus later initiation

of ART was defined as commencement within or outside the time window of four months from

EDI, respectively. The median time for commencement of ART in those who initiated therapy

earlier was 11.9 weeks (2.8 months) from the EDI, and this time point was used to stratify

participants who started ART earlier into two groups designated as Earlier I and II. Thus, Earlier

I comprised participants who commenced ART within 11.9 weeks from EDI, whereas Earlier II

comprised those who commenced ART between 11.9 weeks (2.8 months) and 17.2 weeks (4

months) from the EDI. Panel B, Kaplan-Meier plots depict the rate of achieving primary CD4+

recovery (≥900 cells/mm3) in participants commencing ART earlier (combined Earlier I plus II)

or later, at higher vs. lower pre-ART CD4+ counts. Panel C, Kaplan-Meier plots depict the rate

of achieving primary CD4+ recovery in participants who started ART earlier at higher vs. lower

CD4 counts and were further classified as Earlier I vs. Earlier II. Panels D and E are the same as

panels B and C, respectively, except that they are for attainment of secondary CD4+ recovery

(≥800 cells/mm3) on ART. The significance values noted are for the comparisons of the

indicated study groups and were derived by Cox proportional hazard modeling.

Le et al

19

Le et al

20

Figure S4. Rate and Extent of CD4+ Recovery after Commencement of ART According to

Timing of ART (Earlier versus Later) and CD4+ T-cell Count at ART Initiation (Higher

versus Lower).

Panel A, the Kaplan-Meier plots for the attainment of a CD4+ count of ≥500 cells/mm3 on

therapy in the three color- and number-coded participant groups who initiated ART with CD4+

levels of <500 cells/mm3. The rate ratio (95% confidence interval and significance values) for

the comparisons between patient group 4 (referent group) versus group 5 were 0.48 (0.30 to 076;

P=0.002) and versus group 6 were 0.21 (0.12 to 0.38; P<0.001). Panel B, depicts the distribution

of the last CD4+ count in the indicated CD4+ stratum measured during VL suppression for the

six number-coded participant groups shown in Figure 3A (main text).

Le et al

21

References 1. Panel on Antiretroviral Guidelines for Adults and Adolescents. Guidelines for the use of antiretroviral agents in HIV-1-infected adults and adolescents. Department of Health and Human Services. wwwaidsinfonihgov 2011:1-166. 2. Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM. Statistics in medicine--reporting of subgroup analyses in clinical trials. N Engl J Med 2007;357(21):2189-94. 3. Fiebig EW, Wright DJ, Rawal BD, et al. Dynamics of HIV viremia and antibody seroconversion in plasma donors: implications for diagnosis and staging of primary HIV infection. Aids 2003;17(13):1871-9. 4. Kothe D, Byers RH, Caudill SP, et al. Performance characteristics of a new less sensitive HIV-1 enzyme immunoassay for use in estimating HIV seroincidence. J Acquir Immune Defic Syndr 2003;33(5):625-34. 5. Ferreira MA, Mangino M, Brumme CJ, et al. Quantitative trait loci for CD4:CD8 lymphocyte ratio are associated with risk of type 1 diabetes and HIV-1 immune control. Am J Hum Genet 2010;86(1):88-92. 6. Erkeller-Yuksel FM, Deneys V, Yuksel B, et al. Age-related changes in human blood lymphocyte subpopulations. J Pediatr 1992;120(2 Pt 1):216-22. 7. Wikby A, Mansson IA, Johansson B, Strindhall J, Nilsson SE. The immune risk profile is associated with age and gender: findings from three Swedish population studies of individuals 20-100 years of age. Biogerontology 2008;9(5):299-308. 8. Jentsch-Ullrich K, Koenigsmann M, Mohren M, Franke A. Lymphocyte subsets' reference ranges in an age- and gender-balanced population of 100 healthy adults--a monocentric German study. Clin Immunol 2005;116(2):192-7. 9. Bisset LR, Lung TL, Kaelin M, Ludwig E, Dubs RW. Reference values for peripheral blood lymphocyte phenotypes applicable to the healthy adult population in Switzerland. Eur J Haematol 2004;72(3):203-12. 10. Santagostino A, Garbaccio G, Pistorio A, et al. An Italian national multicenter study for the definition of reference ranges for normal values of peripheral blood lymphocyte subsets in healthy adults. Haematologica 1999;84(6):499-504. 11. Bofill M, Janossy G, Lee CA, et al. Laboratory control values for CD4 and CD8 T lymphocytes. Implications for HIV-1 diagnosis. Clin Exp Immunol 1992;88(2):243-52. 12. Maini MK, Gilson RJ, Chavda N, et al. Reference ranges and sources of variability of CD4 counts in HIV-seronegative women and men. Genitourin Med 1996;72(1):27-31. 13. Shahabuddin S. Quantitative differences in CD8+ lymphocytes, CD4/CD8 ratio, NK cells, and HLA-DR(+)-activated T cells of racially different male populations. Clin Immunol Immunopathol 1995;75(2):168-70. 14. Freedman DS, Gates L, Flanders WD, et al. Black/white differences in leukocyte subpopulations in men. Int J Epidemiol 1997;26(4):757-64. 15. Vuillier F, Lapresle C, Dighiero G. Comparative analysis of CD4-4B4 and CD4-2H4 lymphocyte subpopulations in HIV negative homosexual, HIV seropositive and healthy subjects. Clin Exp Immunol 1988;71(1):8-12. 16. Lebranchu Y, Thibault G, Degenne D, Bardos P. Abnormalities in CD4+ T lymphocyte subsets in patients with common variable immunodeficiency. Clin Immunol Immunopathol 1991;61(1):83-92.

Le et al

22

17. Amadori A, Zamarchi R, De Silvestro G, et al. Genetic control of the CD4/CD8 T-cell ratio in humans. Nat Med 1995;1(12):1279-83. 18. Tsegaye A, Messele T, Tilahun T, et al. Immunohematological reference ranges for adult Ethiopians. Clin Diagn Lab Immunol 1999;6(3):410-4. 19. Kassu A, Tsegaye A, Petros B, et al. Distribution of lymphocyte subsets in healthy human immunodeficiency virus-negative adult Ethiopians from two geographic locales. Clin Diagn Lab Immunol 2001;8(6):1171-6. 20. Gratama JW, Kluin-Nelemans HC, Langelaar RA, et al. Flow cytometric and morphologic studies of HNK1+ (Leu 7+) lymphocytes in relation to cytomegalovirus carrier status. Clin Exp Immunol 1988;74(2):190-5. 21. Giorgi JV, Cheng HL, Margolick JB, et al. Quality control in the flow cytometric measurement of T-lymphocyte subsets: the multicenter AIDS cohort study experience. The Multicenter AIDS Cohort Study Group. Clin Immunol Immunopathol 1990;55(2):173-86. 22. Wright JJ, Wagner DK, Blaese RM, Hagengruber C, Waldmann TA, Fleisher TA. Characterization of common variable immunodeficiency: identification of a subset of patients with distinctive immunophenotypic and clinical features. Blood 1990;76(10):2046-51. 23. Williams RC, Jr., Koster FT, Kilpatrick KA. Alterations in lymphocyte cell surface markers during various human infections. Am J Med 1983;75(5):807-16. 24. Yagi MJ, Joesten ME, Wallace J, Roboz JP, Bekesi JG. Human immunodeficiency virus type 1 (HIV-1) genomic sequences and distinct changes in CD8+ lymphocytes precede detectable levels of HIV-1 antibodies in high-risk homosexuals. J Infect Dis 1991;164(1):183-8. 25. Kalayjian RC, Landay A, Pollard RB, et al. Age-related immune dysfunction in health and in human immunodeficiency virus (HIV) disease: association of age and HIV infection with naive CD8+ cell depletion, reduced expression of CD28 on CD8+ cells, and reduced thymic volumes. J Infect Dis 2003;187(12):1924-33. 26. Tollerud DJ, Clark JW, Brown LM, et al. The influence of age, race, and gender on peripheral blood mononuclear-cell subsets in healthy nonsmokers. J Clin Immunol 1989;9(3):214-22. 27. Fahey JL, Prince H, Weaver M, et al. Quantitative changes in T helper or T suppressor/cytotoxic lymphocyte subsets that distinguish acquired immune deficiency syndrome from other immune subset disorders. Am J Med 1984;76(1):95-100. 28. Nowicki MJ, Karim R, Mack WJ, et al. Correlates of CD4+ and CD8+ lymphocyte counts in high-risk immunodeficiency virus (HIV)-seronegative women enrolled in the women's interagency HIV study (WIHS). Hum Immunol 2007;68(5):342-9.