Scholarly article on topic 'Increased non-AIDS mortality among persons with AIDS-defining events after antiretroviral therapy initiation'

Increased non-AIDS mortality among persons with AIDS-defining events after antiretroviral therapy initiation Academic research paper on "Health sciences"

CC BY
0
0
Share paper
OECD Field of science
Keywords
{""}

Academic research paper on topic "Increased non-AIDS mortality among persons with AIDS-defining events after antiretroviral therapy initiation"

RESEARCH ARTICLE

Increased non-AIDS mortality among persons with AIDS-defining events after antiretroviral therapy initiation

April C Pettit1©, Mark J Giganti2, Suzanne M Ingle3, Margaret T May3, Bryan E Shepherd2, Michael J Gill4, Gerd Fatkenheuer5, Sophie Abgrall67, Michael S Saag8, Julia Del Amo9, Amy C Justice10,11, Jose M Miro12,

q A *1 ^ *1 /-, 1 "7 Q 1 o

Matthias Cavasinni , Francois Dabis , Antonella D Monforte , Peter Reiss , Jodie Guest , David Moore ■ ,

0(~\ 91 oo oo OA. O^

Leah Shepherd , Niels Obel , Heidi M Crane , Colette Smith , Ramon Teira , Robert Zangerle , Jonathan AC Sterne3, Timothy R Sterling1 and for the Antiretroviral Therapy Cohort Collaboration (ART-CC) investigators

Corresponding author: April C Pettit, 1161 21st Avenue South, A2200 MCN, Nashville, Tennessee 37232, USA. Tel: 001 615 343 0574. (april.pettit@vanderbilt.edu) These data were presented in part at the 19th International Workshop on HIV ObservationalDatabases (IWHOD), 26 to 28 March 2015, Catania, Italy, Abstract #0052 and the 20th International Workshop on HIV Observational Databases (IWHOD), 7 to 9 April 2016, Budapest, Hungary, Abstract #0046.

Abstract

Introduction: HIV-1 infection leads to chronic inflammation and to an increased risk of non-AIDS mortality. Our objective was to determine whether AIDS-defining events (ADEs) were associated with increased overalland cause-specific non-AIDS related mortality after antiretroviraltherapy (ART) initiation.

Methods: We included HIV treatment-naive adults from the AntiretroviralTherapy Cohort Collaboration (ART-CC) who initiated ART from 1996 to 2014. Causes of death were assigned using the Coding Causes of Death in HIV (CoDe) protocol. The adjusted hazard ratio (aHR) for overalland cause-specific non-AIDS mortality among those with an ADE (all ADEs, tuberculosis (TB), Pneumocystis jiroveci pneumonia (PJP), and non-Hodgkin's lymphoma (NHL)) compared to those without an ADE was estimated using a marginalstructuralmodel.

Results: The adjusted hazard of overallnon-AIDS mortality was higher among those with any ADE compared to those without any ADE (aHR 2.21, 95% confidence interval(CI) 2.00 to 2.43). The adjusted hazard of each of the cause-specific non-AIDS related deaths were higher among those with any ADE compared to those without, except metabolic deaths (malignancy aHR 2.59 (95% CI 2.13 to 3.14), accident/suicide/overdose aHR 1.37 (95% CI 1.05 to 1.79), cardiovascular aHR 1.95 (95% CI 1.54 to 2.48), infection aHR (95% CI 1.68 to 2.81), hepatic aHR 2.09 (95% CI 1.61 to 2.72), respiratory aHR 4.28 (95% CI 2.67 to 6.88), renalaHR 5.81 (95% CI 2.69 to 12.56) and centralnervous aHR 1.53 (95% CI 1.18 to 5.44)). The risk of overalland cause-specific non-AIDS mortality differed depending on the specific ADE of interest (TB, PJP, NHL).

Conclusions: In this large multi-centre cohort collaboration with standardized assignment of causes of death, non-AIDS mortality was twice as high among patients with an ADE compared to without an ADE. However, non-AIDS related mortality after an ADE depended on the ADE of interest. Although there may be unmeasured confounders, these findings suggest that a common pathway may be independently driving both ADEs and NADE mortality. While prevention of ADEs may reduce subsequent death due to NADEs following ART initiation, modification of risk factors for NADE mortality remains important after ADE survival.

Keywords: AIDS-defining events; non-AIDS mortality; tuberculosis; Pneumocystis jiroveci pneumonia; non-Hodgkin's lymphoma; marginal structural model

AdditionalSupporting Information may be found online in the Supporting Information tab for this article.

Received 23 March 2017; Accepted 10 November 2017

Copyright © 2018 The Authors. Journalof the International AIDS Society published by John Wiley & sons Ltd on behalf of the InternationalAIDS Society. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

1 | INTRODUCTION

In the era of antiretroviraltherapy (ART), there have been striking decreases in acquired immune deficiency syndrome

(AIDS)-related mortality [1-3] and prolongation of life expectancy [4-7] for people living with HIV infection (PLWH). With this, there has been a shift in distribution of causes of death toward non-AIDS-defining events (NADEs) [8-12]. Factors

contributing to non-AIDS mortality include immunodeficiency [13,14], ART toxicity [15,16], increasing age [17-22], and lifestyle factors such as tobacco use and obesity [18,21,23]. However, chronic inflammation and immune activation produced by chronic HIV infection as wellas many AIDS-defining events (ADEs) are now recognized as significant drivers in the pathogenesis of non-AIDS-related deaths [14,24-28].

A previous AntiretroviralTherapy Cohort Collaboration (ART-CC) study sought to examine variation in mortality associated with specific ADEs among patients receiving ART. The study found that overallmortality rates after an ADE were dependent on the specific ADE diagnosis and a classification scheme for ADE severity (mild, moderate, severe) was proposed [29]. This study demonstrated that not allADEs have the same consequences with regards to overallmortality. However, studies to determine if this finding holds true for mortality due to non-AIDS deaths are lacking.

Our objective was to estimate the overalleffect of ADEs as well as three specific ADEs of differential severity on the risk of NADE mortality after ART initiation in high-income settings. We sought to appropriately controlfor time-updated diagnosis of ADEs and immune status (CD4+ count and HIV-1 RNA). Moreover, we evaluated the overalland ADE-specific effects on differing subtypes of NADE deaths.

2 | METHODS

2.1 | Patient population

We conducted an observationalcohort analysis among PLWH enrolled in the ART-CC from 1996 to 2014 (http://www.b ris.ac.uk/art-cc) [30]. Data were contributed by 19 cohorts in Europe and North America. Patients were included if they were ART-naive, had CD4+ and HIV-1 RNA values available at ART initiation, and were >18 years of age at ART initiation. Ethics approval was obtained from all participating study sites, the NationalHealth Service Health Research Authority South West-Cornwalland Plymouth Research Ethics Committee, United Kingdom (REC reference 12/SW/0253). Informed consent or a waiver of informed consent were obtained as required by local site ethics committees. Patients were assigned a random study number at the local sites and a limited dataset was transmitted to the ART-CC Data Coordinating Center.

2.2 | Study definitions

Person-time was contributed beginning at ART initiation (study baseline) untilthe earliest of loss to follow-up, death, or administrative censoring (May 2012 to December 2015, depending on cohort). ART was defined as a regimen that contained >3 drugs including nucleoside reverse transcriptase inhibitors (NRTIs), protease inhibitors (PIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), or integrase inhibitors. Baseline CD4+ and HIV-1 RNA values were defined as the first value <90 days before or <7 days after ART initiation. Loss to follow-up (LTFU) was defined as >12 months between available laboratory results; these patients were censored at 12 months after the date of their last results. Persons were not allowed to re-enter the analysis after meeting LTFU criteria.

Information on cause of death was recorded using the InternationalClassification of Diseases Ninth Revision (ICD-9) code, the ICD-Tenth Revision (ICD-10) code, a classification based on the Coding of Death in HIV (CoDe) project (http:// www.cphiv.dk/Tools-Standards/CoDe/About), or free text. Causes of death were classified using an adapted version of the CoDe protocol into mutually exclusive categories. Clinicians classified deaths using information on death (ICD-9/ ICD-10 codes or free text), patient characteristics at ART initiation (age, sex, HIV transmission risk group, ADEs, and hepatitis C status), time from ART initiation to death, ADEs after starting ART, latest CD4 (within six months of death), and whether a patient was on ART at the time of death. A computer algorithm and a clinician classified deaths using ICD codes when these codes were available. Otherwise, two clinicians independently classified each death. Disagreements between clinicians and/or computer-classifications were resolved via paneldiscussion as per the CoDe protocol[1,31].

The exposure of interest was an ADE after ART initiation, defined according to the US Centers for Disease Controland Prevention [32]. Diagnosis of ADEs after ART initiation and the dates of diagnosis were validated by localsite investigators. In addition to evaluating ADEs overall, two mild ADEs (tuberculosis (TB) and Pneumocystis jiroveci pneumonia (PJP)) and one severe ADE (non-Hodgkin's lymphoma (NHL)) according a previous classification scheme [29], were evaluated to elucidate differences in NADE mortality based on ADE severity.

The outcome of interest was NADE mortality. We evaluated both overallnon-AIDS deaths as wellas cause-specific nonAIDS deaths, including cardiovascular, hepatic, metabolic, nonAIDS malignancy, non-AIDS infection, renal, respiratory, central nervous, accident/suicide/overdose, and other NADE deaths not falling into one of the previous categories.

2.3 | Statistical analysis

For continuous variables, median and interquartile range (IQR) are shown. For categoricalvariables, number (n) and percent (%) are shown. Crude NADE mortality rates per 1000 person-years (p-y) of follow-up were calculated for the overallpopulation as wellas for persons with ADE and without an ADE after ART initiation.

To modelthe association between ADEs and NADE mortality while appropriately adjusting for baseline and time-dependent confounders, marginalstructuralmodels were constructed [33]. Patient exposure status was assessed monthly until censoring or death. Patients were classified as unexposed for each month prior to their first ADE after ART initiation and exposed in each subsequent month. As we were interested in ADEs after ART initiation, patients with an ADE prior to ART initiation were classified as unexposed at baseline. Inverse probability weights (IPW) for the primary modelof interest were determined based on four components. Each component was a pooled logistic regression model predicting probability of the following events each month: 1) ADE diagnosis; 2) LTFU; 3) administrative censoring; 4) a competing event, including death due to other NADEs (in the case of cause-specific NADE deaths), death due to ADEs, and death from unknown causes. These models included time since ART initiation, age at ART initiation, sex, cohort, year of ART

initiation, baseline ART regimen (PI-based, NNRTI-based, other), HIV transmission risk group (heterosexual, men who have sex with men (MSM), injection drug use (IDU), blood transfusion, other/unknown), presence of ADE at or prior to ART initiation, baseline CD4+ count, baseline HIV-1 RNA, and current CD4+ count and HIV-1 RNA; restricted cubic splines were used for allcontinuous variables. Both current CD4+ count and HIV-1 RNA were time-dependent variables. If a patient had more than one measurement during a given month, the value from the latest date was utilized; if a patient did not have a measurement during a given month, the value from the previous month was carried forward.

A weighted pooled logistic regression modelincluding robust standard errors was used to estimate the hazard ratio (HR) for NADE mortality comparing those who did and did not experience the ADE. This weighted pooled logistic regression model approximates a marginal structural Cox regression model when certain conditions are met (short intervals and low event rates) [33,34]. The weights from the four components above were multiplied to obtain the single time-updated weight incorporated into this model. This modelincluded current ADE status (current or previous ADE diagnosis), time since initialvisit date in months, age at ART initiation, sex, cohort, year of ART initiation, baseline ART regimen, HIV transmission risk factor, presence of ADE at or prior to ART initiation, baseline CD4+ count, and baseline HIV-1 RNA. Results are reported with stabilized weights [33] truncated at the 1st and 99th percentiles. The median of the untruncated stabilized weights was 0.98 (IQR 0.90 to 1.02) for the model including allADEs as the exposure; weights for models corresponding to specific ADE exposures were similar (results not shown).

In a sensitivity analysis, we fitted severalalternative models of the association between ADEs and non-AIDS mortality, including an unadjusted model that did not adjust for any covariates and a model that only adjusted for baseline covari-ates. For both models, we did not include any IPW and thus did not account for time-dependent covariates. We also refitted the marginal structural model after excluding observations from sites with >50% missing codes of death, after excluding those with AIDS at or prior to baseline, and after limiting the cohort to only patients with a baseline CD4+ count above 500 cells/mm3. Allstatisticaltests were two-sided. Statistical analyses were performed using R version 3.2. Analysis code is available at http://biostat.mc.vanderbilt.edu/ ArchivedAnalyses.

3 | RESULTS

There were 124,587 patients included and followed up for a totalof 770,259 p-y (median 5.18 years) in this study. The median age at ART initiation was 38 years, 76% were male, and 12% reported IDU as an HIV transmission factor. The median baseline CD4+ count was 244 cells/mm3 (IQR 117 to 369 cells/mm3) and the median baseline HIV-1 RNA was 66,580 copies/ml(IQR: 14,000 to 200,000 copies/ml) (Table 1). There were 14,245 patients (11%) who developed at least one ADE after ART initiation. Of these, 2174 (15%) had TB, 1864 (13%) had PJP, and 939 (7%) had NHL (patients could contribute to >1 ADE category). Among patients with

Table 1. Demographic characteristics of the study population, ART-CC

Overall (N=124,587)

Age (years)-median (IQR) 38 (32 to 46)

Male sex 95,001 (76%)

Year of ART initiation-median (IQR) 2005 (2001 to 2009)

Baseline ART regimen

NNRTI 49,936 (40%)

PI 64,306 (52°%)

Other 10,345 (8%)

ADE at or prior to baseline 24,849 (20%)

Transmission risk

Heterosexual 43,644 (35%)

MSM 43,387 (35%)

IDU 15,347 (12%)

Blood transfusion 980 (1%)

Other/unknown 21,229 (17%)

Baseline CD4+ count (cells/ml), 244 (117 to 369)

median (IQR)

Baseline HIV-1 RNA (copies/ml), 66,580 (14,000 to 200,000)

median (IQR)

Years of follow-up, median (IQR) 5.18 (2.28 to 9.42)

ADE, AIDS-Defining Event; IQR, interquartile range; ART, antiretroviral therapy; PI, protease inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor; HIV, human immunodeficiency virus; MSM, male-male sexual contact; IDU, injection drug use; ml, millilitre.

an ADE, 6588 (46.2%) were on a PI-based regimen, 3955 (27.8%) were on an NNRTI-based regimen, 1114 (7.8%) were another regimen, and 2588 (18.2%) were not on ART at the time of the ADE. The proportion of patients with an ADE after ART initiation was higher among patients with a baseline a CD4+ count less than 500 cells/ml3 (13,324/110,412; 12.1%) compared to those with a baseline CD4+ count greater than 500 cells/ml3 (921/14,175; 6.5%).

There were 11,280 deaths during the study period; 2661 (24%) were AIDS-related deaths, 4051 (36%) were non-AIDS related deaths, and 4568 (40%) were unknown types of deaths. Among non-AIDS related deaths, 956 (24%) were due to a non-AIDS malignancy, 673 (17%) were due to accident/ suicide/overdose, 649 (16%) were due to cardiovascular disease, 598 (15%) were non-AIDS infection, 529 (13%) were hepatic, 121 (3%) were respiratory, 68 (2%) were renal, 41 (1%) were metabolic, 57 (1%) were centralnervous system (CNS) and 359 (9%) were due to other causes (Table 2). The proportion of missing death codes varied by cohort (10% to 92%) (Table S1). Among those who died, death codes were missing for 41% of persons without an ADE after ART initiation compared to 39% of persons with an ADE after ART initiation (p=0.03).

The crude overallnon-AIDS mortality rate was 5.26/ 1000 p-y (95% confidence interval(CI) 5.09 to 5.42). Persons with any ADE after ART initiation had a higher crude overall non-AIDS mortality rate (14.90/1000 p-y (95% CI 13.99 to 15.79)) compared to persons without any ADE after ART initiation (4.28/1000 p-y (95% CI 4.12 to 4.43)). This finding of a

Table 2. Classification of deaths, ART-CC

CoDe* codes Number (%)

AIDS coded deaths 2661 (24%)

AIDS infection 01.1 1036

AIDS malignancy 01.2 635

AIDS unspecified 01 990

NADE coded deaths 4051 (36%)

Non-AIDS malignancy 04 956

Accident/suicide/overdose 16, 17, 19 673

Cardiovascular 08, 09, 12, 24 649

Non-AIDS infection 02 598

Hepatic 03, 14 529

Respiratory 13, 25 121

Renal 15 68

Central nervous 23 57

Metabolic 05, 06,07 41

Other 10, 11, 18, 20, 22, 359

26, 27, 28, 29, 90

Unknown death 4568 (40%)

Coded, unknown 91, 92 1253

Uncoded (no information) 3315

Total deaths 11,280

*Coding of Death in HIV (CoDe) project (http://www.cphiv.dk/Tools-Standards/CoDe/About). AIDS, acquired immunodeficiency syndrome; NADE, non-AIDS-defining event.

higher crude overallnon-AIDS mortality rate held true for each of the specific ADEs evaluated (Table 3). The median time from diagnosis of any ADE after ART initiation to NADE mortality was 702 days (IQR: 147 to 1661 days). This time from ADE to NADE mortality for the three specific ADEs of interest was 902 days for TB (IQR: 377 to 1912 days), 568 days for PJP (IQR 170 to 1413 days) and 225 days for NHL (IQR 89 to 1534 days).

Table 3. Crude non-AIDS related mortality rate per 1000 person-years, ART-CC

NADE mortality

NADE Follow-up rate per 1000

deaths (years) person-years (95% CI)

Overall 4051 770,259 5.26 (5.09 to 5.42)

Persons with ADE 1057 70,917 14.88 (13.99 to 15.79)

Persons without ADE 2994 699,342 4.28 (4.12 to 4.43)

Persons with TB 131 11,587 11.17 (9.28 to 13.07)

Persons without TB 3920 758,573 5.17 (5.00 to 5.32)

Persons with PJP 129 8247 15.62 (13.05 to 18.44)

Persons without PJP 3922 762,013 5.15 (4.98 to 5.30)

Persons with NHL 44 2886 15.12 (10.88 to 20.02)

Persons without NHL 4007 767,373 5.22 (5.05 to 5.38)

NADE, non-AIDS defining event; ADE, AIDS-defining event; CI, confidence interval; TB, tuberculosis; PJP, Pneumocystis jiroveci pneumonia; NHL, non-Hodgkin's lymphoma.

In the marginal structural model which adjusted for time-updated covariates, the adjusted hazard of death due to any NADE was significantly higher among patients with any ADE after ART initiation (adjusted hazard ratio (aHR) 2.21, 95% CI 2.00 to 2.43) compared to patients without any ADE after ART initiation. This finding was true for overallNADE deaths and for each NADE death category except for metabolic NADE deaths. The association of cause-specific NADE mortality differed depending on the specific ADE of interest. TB was associated with a higher risk of cardiovascular, metabolic, and non-AIDS infection NADE mortality. PJP was associated with a higher risk of non-AIDS infection, respiratory and accident/suicide/overdose NADE deaths. NHL was associated with a higher risk of renalNADE deaths. Allthree ADEs (TB, PJP and NHL) were associated with a higher risk of death due to non-AIDS malignancies (Table 4).

The results from the marginal structural model were somewhat attenuated although largely unchanged when compared to the results of the unadjusted model with a hazard ratio of 3.50 (95% CI 3.26 to 3.75) for death due to any NADE among patients with any ADE after ART initiation compared to patients without any ADE after ART initiation (Table S2). Similarly, in the model adjusted only for baseline covariates the adjusted ratio for death due to any NADE was 2.68 (95% CI 2.47 to 2.90) among patients with any ADE after ART initiation compared to patients without any ADE after ART initiation (Table S3).

The results of sensitivity analyses with alternative exclusion criteria as described in the methods were largely unchanged as well. In the modelexcluding allsites with >50% missing codes of death (Table 5), the aHR for death due to any NADE was 2.23 (95% CI 1.97 to 2.53). In the modelexcluding those individuals with ADEs at or prior to baseline (Table S4), the aHR for death due to any NADE was 2.37 (95% CI 2.10 to 2.67). Notably, among 14,175 (11%) patients in this cohort with a baseline CD4 cellcount >500 cells/ml3, the aHR for overall NADE mortality following any ADE was 2.66 (95% CI 1.84 to 3.87).

4 | DISCUSSION

In this large, multi-centre cohort collaboration with standardized assignment of causes of death, the risk of death due to NADEs overall was over two times higher for patients with any ADE after ART initiation compared to without an ADE following ART initiation. This finding held true for each of the three selected individualADEs (TB, PJP and NHL). However, there was heterogeneity in the risk of cause-specific NADE mortality across the individual ADEs evaluated, consistent with previous studies of all-cause mortality. These findings suggest that prevention of ADEs, perhaps with more frequent HIV testing and early treatment with ART as recommended by current guidelines, may decrease subsequent NADE mortality.

Overall, the crude mortality rate per 1000 p-y due to NADEs (5.26 (95% CI 5.09 to 5.42)) was slightly higher than that reported from other cohorts in high-income countries, including South Korea (3.71, 95% CI 2.52 to 5.48) [35], Spain (3.75, 95% CI 2.84 to 4.94) [36], and from the Data collection on Adverse Events of anti-HIV Drugs (D:A:D) cohort in

Table 4. Adjusted marginal structural models for hazard ratios of cause-specific mortality

N=6712 aHR (95% CI) aHR (95% CI) aHR (95% CI) aHR (95% CI)

Type of death coded deaths Tuberculosis Pneumocystis Non-Hodgkin Lymphoma All ADEs

ADEs 2661 2.65 (2.20 to 3.19) 5.10 (4.36 to 5.97) 29.49 (25.48 to 34.13) 7.51 (6.64 to 8.49)

NADEs 4051 1.68 (1.38 to 2.04) 2.21 (1.78 to 2.73) 2.95 (2.12 to 4.10) 2.21 (2.00 to 2.43)

NADE malignancy 956 1.84 (1.23 to 2.76) 2.12 (1.30 to 3.47) 5.63 (3.46 to 9.16) 2.59 (2.13 to 3.14)

Accident, suicide, overdose 673 1.39 (0.81 to 2.38) 2.60 (1.54 to 4.38) 2.37 (0.90 to 6.26) 1.37 (1.05 to 1.79)

Cardiovascular 649 1.88 (1.15 to 3.08) 1.74 (0.96 to 3.15) 0.95 (0.20 to 4.46) 1.95 (1.54 to 2.48)

NADE infection 598 1.63 (1.05 to 2.54) 2.66 (1.74 to 4.06) 1.57 (0.45 to 5.47) 2.17 (1.68 to 2.81)

Hepatic 529 1.36 (0.77 to 2.40) 1.06 (0.52 to 2.19) 2.48 (0.79 to 7.76) 2.09 (1.61 to 2.72)

Respiratory 121 0.85 (0.24 to 2.92) 6.76 (3.03 to 15.09) 1.78 (0.44 to 7.16) 4.28 (2.67 to 6.88)

Renal 68 3.13 (0.88 to 11.19) 0.95 (0.22 to 4.05) 9.31 (2.25 to 38.45) 5.81 (2.69 to 12.56)

Central nervous 57 0.74 (0.10 to 5.47) 0.66 (0.09 to 5.04) c 2.53 (1.18 to 5.44)

Metabolic 41 4.30 (1.35 to 13.69) £ S 1.53 (0.67 to 3.47)

NADE, other 359 1.77 (0.86 to 3.66) 3.20 (1.40 to 7.30) 2.45 (0.96 to 6.25) 2.98 (2.14 to 4.16)

aDenotes that no deaths occurred in these categories.

+Models were adjusted for baseline CD4+ count, baseline HIV-1 RNA, sex, HIV transmission risk group, age, calendar year of ART initiation, baseline ART regimen (PI/NNRTI/other), ADE at or prior to the time of enrolment, and ART-CC cohort.

Table 5. Adjusted marginal structural models for hazard ratios of cause-specific mortality (excluding all sites with >50% missing codes of death)

N=4,664 aHR (95% CI) aHR (95% CI) aHR (95% CI) aHR (95% CI)

Type of death coded deaths tuberculosis pneumocystis non-Hodgkin lymphoma All ADEs

ADEs 1841 3.34 (2.64 to 4.23) 7.48 (6.06 to 9.23) 31.23 (26.83 to 36.36) 9.24 (8.08 to 10.58)

All NADEs 2823 1.82 (1.41 to 2.36) 1.58 (1.14 to 2.18) 2.48 (1.74 to 3.53) 2.23 (1.97 to 2.53)

NADE malignancy 674 1.86 (1.10 to 3.15) 0.70 (0.22 to 2.18) 3.68 (2.10 to 6.46) 2.14 (1.65 to 2.77)

Accident, Suicide, Overdose 491 0.87 (0.33 to 2.31) 1.18 (0.50 to 2.81) 2.57 (0.98 to 6.72) 1.14 (0.78 to 1.67)

Cardiovascular 413 2.90 (1.57 to 5.36) 0.99 (0.41 to 2.40) 1.02 (0.23 to 4.64) 2.11 (1.50 to 2.96)

Hepatic 384 1.06 (0.52 to 2.19) 1.08 (0.45 to 2.63) 1.70 (0.42 to 6.83) 1.83 (1.31 to 2.57)

NADE infection 352 2.03 (1.09 to 3.77) 3.01 (1.69 to 5.37) 1.70 (0.49 to 5.98) 2.50 (1.76 to 3.54)

Respiratory 98 1.70 (0.50 to 5.80) 4.81 (2.02 to 11.43) 2.02 (0.50 to 8.08) 4.72 (2.86 to 7.80)

Renal 52 5.03 (1.41 to 17.94) 1.23 (0.16 to 9.28) 10.00 (2.51 to 39.88) 9.19 (4.03 to 20.97)

Central nervous 47 1.11 (0.15 to 8.12) 1.31 (0.17 to 10.20) £ 2.52 (1.15 to 5.50)

Metabolic 28 1.57 (0.20 to 12.36) s £ 1.47 (0.54 to 3.99)

NADE, other 284 2.09 (0.93 to 4.71) 3.11 (1.16 to 8.32) 2.75 (1.07 to 7.07) 3.16 (2.18 to 4.57)

aDenotes that no deaths occurred in these categories.

+Models were adjusted for baseline CD4+ count, baseline HIV-1 RNA, sex, HIV transmission risk group, age, calendar year of ART initiation, baseline ART regimen (PI/NNRTI/other), ADE at or prior to the time of enrolment, and ART-CC cohort.

Australia, Europe, and the United States (4.28, 95% CI 4.06 to 4.53) [12,37]-the latter two have some overlap with our dataset. It is possible this finding is related to differences in baseline CD4+ count (342 cells/mm3 (IQR 163 to 546) in Spain, 400 (IQR 242 to 590) in D:A:D, and 244 (IQR 116 to 369) for ART-CC, respectively) as several previous studies have described lower CD4+ counts as risk factors for NADE mortality [13,14]. In addition, there may be background differences in NADE prevalence and NADE risk factors within the study populations as wellas changes in these populations over time.

Over 60% of the coded causes of death in ART-CC were due to NADEs. This is consistent with the findings from a

recent systematic review and meta-analysis which estimated the pooled proportion of patients with death due to NADEs in high-income countries during the ART era was 54% (95% CI 46% to 62%) [37]. The most frequent cause of NADE death in our study was non-AIDS malignancies (24%) followed by accidents/suicides/overdoses (17%), cardiovascular deaths (16%), non-AIDS infections (15%) and hepatic deaths (13%). These cause-specific findings are also consistent with this recent systematic review and meta-analysis (non-AIDS malignancies 28% (95% CI 23% to 33%), cardiovascular deaths 15% (95% CI 13% to 17%), non-AIDS infections 10% (95% CI 6% to 14%) and hepatic deaths 14% (95% CI 10% to 19%)) [37]. Pooled findings for accidents (n=187; 28%), suicides (n=180; 27%)

and overdoses (n=306; 45%) were not reported although these types of deaths were the second most common NADE death in our population.

A previous study conducted within the ART-CC found that overallmortality rates (deaths due to both ADEs and NADEs) subsequent to an ADE were dependent on the specific ADE diagnosis; NHL was classified as severe and TB and PJP were classified as mild [29]. In this study, while the associations between specific ADEs and overall NADE mortality were all statistically significant, we note that the strength of that association varied. The point estimate for the risk of NADE mortality overallwas highest for NHL (2.95, 95% CI 2.12 to 4.10), previously designated as a severe ADE [29].

However, the risk of cause-specific NADE mortality differed between the three ADEs evaluated (TB, PJP, and NHL). For example, only TB was associated with cardiovascular NADE deaths. Previous epidemiologic studies have described an association between TB and cardiovascular disease. A large retrospective study in Taiwan reported a 40% increased risk of cardiovascular events among patients with TB compared to patients without TB after controlling for important co-morbidities [38]. Several possible mechanisms of the association include direct mycobacterial tissue invasion, auto-immunity mediated by antibodies against mycobacteria, as well as increased inflammation and immune activation, [39], although prospective studies are needed.

Only PJP was associated with subsequent death due to respiratory events and accident/suicide/overdoses. In the Pulmonary Complications of HIV Study, PJP was associated with declines in lung function that persisted even after the acute infection resolved [40]. However, the association of PJP with accidents/ suicides/overdoses suggests that not allof these associations may be mechanistic and that unmeasured confounders such as mentalillness and alcohol/drug use are likely important.

There are some limitations to this study. First, there are likely unmeasured confounders not accounted for in these analyses. The ART-CC does not collect information on other important lifestyle factors such as tobacco use, alcoholuse, active IDU and body mass index (BMI). PLWH who smoke have higher rates of cardiovascular disease, respiratory disease and malignancies than people without HIV infection who smoke. A recent modelling study showed that PLWH in the US who smoke lose as much or more life expectancy from smoking than from HIV infection itself. Moreover, this study showed that smoking cessation would result in a greater gain in life expectancy than early HIV testing and treatment or improved ART adherence [41]. It is unclear if the associations we found would remain after controlling for such potentially important unmeasured variables.

Second, it is possible that some causes of death may be coded inaccurately as the coding was performed retrospectively without access to complete medicalhistories. It is also possible that we categorized diseases of different aetiologies in the same cause-specific categories. For example, it is possible that a dysrhythmia due to ischemic heart disease and a dysrhythmia due to drug overdose may have both been categorized as a cardiovascular cause of death. However, given the rigorous standardized process used for coding of deaths it is likely that the majority were coded correctly. Moreover, some deaths could not be coded and therefore some deaths due to NADEs may have been missed.

Thirdly, cause of death was not available for 40% of deaths within the cohort over the study period. Therefore, we performed a sensitivity analysis in which we excluded allsites with >50% missing codes of death. Compared to the main results, the results of the sensitivity analysis were largely unchanged. However, causes of death in the main analysis were differentially missing between those with and without ADEs after ART initiation, which may have led to biased results.

5 j CONCLUSIONS

In conclusion, in this cohort of persons initiating ART, NADE mortality rates were higher among those with an ADE after ART initiation compared to those without an ADE after ART initiation. Consistent with previous studies of overallmortality, NADE mortality rates after an ADE depended on the specific ADE diagnosed. Although there may be unmeasured con-founders and associations may not be mechanistic, these findings suggest that a common pathway may be independently driving both ADEs and NADE mortality. ADE prevention, perhaps by more frequent HIV testing and initiating treatment at higher CD4+ counts, as wellas the continued modification of risk factors such as tobacco use, may reduce subsequent NADE mortality. In addition, these findings underscore the need for future studies to elucidate a potential mechanism for this association, including that of chronic inflammation and immune activation due to ADEs. In these future studies it will be important to address potentialconfounders including important modifiable lifestyle factors such as tobacco use.

AUTHORS' AFFILIATIONS

1Division of Infectious Diseases, Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, USA; 2Department of Biostatistics, Vanderbilt University MedicalCenter, Nashville, TN, USA; 3BristolMedicalSchool, University of Bristol, Bristol, UK; 4Division of Infectious Diseases, University of Calgary, Calgary, Canada; 5Department of Internal Medicine, University of Cologne, Cologne, Germany; 6Sorbonne Universites, UPMC Univ Paris 06, UMR_S 1136, Institut Pierre Louis d'Epidemiologie et de Sante Publique, Paris, France; 7INSERM, UMR_S 1136, Institut Pierre Louis d'Epidemiologie et de Sante Publique, Paris, France; 8Division of Infectious Disease, Department of Medicine, University of Alabama, Birmingham, AL, USA; 9National Epidemiology Center, Carlos III Health Institute, Madrid, Spain; 10Yale University Schoolof Medicine, New Haven, CT, USA; 11VA Connecticut Healthcare System, West Haven, CT, USA; 12HospitalClinic- Institut d'Investigacions Biomediques Pi i Sunyer (IDI-BAPS), University of Barcelona, Barcelona, Spain; 13Service of Infectious Diseases, Lausanne University Hospitaland University of Lausanne, Lausanne, Switzerland; 14INSERM U.1218 Bordeaux Population Health, ISPED, Bordeaux University, Bordeaux, France; 15Clinic of Infectious Diseases & TropicalMedi-cine, San Paolo Hospital, University of Milan, Milan, Italy; 16Stichting HIV Monitoring, Division of Infectious Diseases, Department of Global Health, Academic MedicalCenter, University of Amsterdam, Amsterdam Institute for Global Health and Development, Amsterdam, The Netherlands; 17HIV Atlanta VA Cohort Study (HAVACS), Atlanta Veterans Affairs MedicalCenter, Decatur, GA, USA; 18Division of Epidemiology and Population Health, British Columbia Centre for Excellence in HIV/AIDS, Vancouver, Canada; 19Faculty of Medicine, University of British Columbia, Vancouver, Canada; 20Research Department of Infection and Population Health, University College London, London, UK; 21Department of Infectious Diseases, Copenhagen University Hospital, Copenhagen, Denmark; 22Center for AIDS Research, University of Washington, Seattle, WA, USA; 23Research Department of Infection and Population Health, UCL, London, UK; 24Unit of Infectious Diseases, HospitalSierrallana, Torrelavega, Spain; 25Innsbruck MedicalUniversity, Innsbruck, Austria

COMPETING INTERESTS

Allauthors have no competing interests to declare.

AUTHORS' CONTRIBUTIONS

Study conception and design: AP, MG, SI, MM, BS, JS, TS. Acquisition of data: SI, MM, MG, GF, SA, MS, JD, JM, MC, FD, AM, PR, JG, DM, LS, NO, HC, CS, RT, RZ, JS, TS. Data analysis and interpretation: AP, MG, SI, MM, BS, JS, TS. Drafting of manuscript: AP, MG. Revising manuscript for important intellectualcontent: AP, MG, SI, MM, BS, MG, GF, SA, MS, JD, JM, MC, FD, AM, PR, JG, DM, LS, NO, HC, CS, RT, RZ, JS,TS. Allauthors have read and approved the finalmanuscript.

ACKNOWLEDGEMENTS

We would like to thank the following cohorts for participating in this study; the AIDS Therapy Evaluation Project Netherlands (ATHENA); ANRS CO4 French HospitalDatabase on HIV (FHDH); ANRS CO3 Aquitaine Cohort, France; Departments of InternalMedicine at University of Cologne and Bonn, Germany; Köln/Bonn Cohort, Germany; Italian Cohort of Antiretroviral-Naïve Patients (ICONA); CORIS, Spain; Proyecto para la Informatización delSeguimiento Clínico-epidemiológico de la Infección por HIV y SIDA (PISCIS), Spain; Royal Free HospitalCohort, London UK; Swiss HIV Cohort Study (SHCS); The Multicenter Study Group on EuroSIDA; Southern Alberta Clinic, Canada; HIV Atlanta Veterans affairs Cohort Study (HAVACS), USA; UAB 1917 Clinic Cohort, Birmingham, Alabama, USA; Veterans Ageing Cohort Study (VACS8),USA; Vanderbilt-Meharry and Tennessee Center for AIDS Research, Nashville, Tennessee, USA; University of Washington Harborview MedicalCenter, USA; Österreichische HIV-Kohortenstudie (OEHIVKOS), Austria; Danish HIV Cohort Study, Denmark; VACH, (Spain); and HAART ObservationalMedicalEvaluation and Research (HOMER), British Columbia, Canada. We would also like to thank the ART-CC Steering group; Andrew Boulle (IeDEA Southern Africa), Christoph Stephan (Frankfurt), Jose M. Miro (PISCIS), Matthias Cavassini (SHCS), Geneviève Chêne (Aquitaine), Dominique Costagliola (FHDH), Francois Dabis (Aquitaine), Antonella D'Arminio Monforte (ICONA), Julia DelAmo (CoRIS), Ard Van Sighem (ATHENA), Jörg Janne Vehreschild (Koln/Bonn), John Gill (South Alberta Clinic), Jodie Guest (HAVACS), David Hans-Ulrich Haerry (EATG), Robert Hogg (HOMER), Amy Justice (VACS), Leah Shepherd (EuroSIDA), Neils Obel (Denmark), Heidi Crane (Washington), Colette Smith (Royal Free), Peter Reiss (ATHENA), Michael Saag (Alabama), Tim Sterling (Vanderbilt-Meharry), Ramon Teira (VACH), Matthew Williams (UK-CAB), Robert Zangerle (Austria) and the ART-CC Co-ordinating team; Jonathan Sterne and Margaret May (Principal Investigators), Suzanne Ingle, Adam Trickey (statisticians).

FUNDING

This work was jointly funded by the UK MedicalResearch Council(MRC) (grant number MR/J002380/1) and the UK Department for International Development (DFID) under the MRC/DFID Concordat agreement and is also part of the EDCTP2 programme supported by the European Union. Sources of funding for April Pettit: National Institutes of Health K08 AI104352. Vanderbilt University MedicalCenter: Vanderbilt-Meharry CFAR (NIH P30 AI54999), Tennessee CFAR (NIH P30 AI110527), Sterling K24 (NIH K 24 AI065298). PISCIS was supported in part by the Ministerio de Sanidad y Consumo, Instituto de Salud Carlos III, Madrid (Spain), and Spanish Network for AIDS Research (RIS; ISCIII-RETIC RD06/006). COHERE Acknowledgements: The COHERE study group has received unrestricted funding from: Agence Nationale de Recherches sur le SIDA et les Hepatites Virales (ANRS), France; HIV Monitoring Foundation, the Netherlands; and the Augustinus Foundation, Denmark. COHERE receives funding from the European Union Seventh Framework Programme (FP7/2007-2013) under EuroCoord grant agreement no 260694. A list of the funders of the participating cohorts can be found on the Regional Coordinating Centre websites at http://www.cphiv.dk/COHERE/tabid/295/Default.aspx and http:// etudes.isped.u-bordeaux2.fr/cohere. Jose M. Miró received a personalintensifi-cation research grant #INT15/00168 during 2016 from Instituto de Salud Carlos III, Madrid, Spain. Jonathan Sterne is funded by National Institute for Health Research Senior Investigator award NF-SI-0611-10168.

REFERENCES

1. Antiretroviral Therapy Cohort Collaboration. Causes of death in HIV-1-infected patients treated with antiretroviraltherapy, 1996-2006: collaborative analysis of 13 HIV cohort studies. Clin Infect Dis. 2010;50(10):1387-96.

2. Lima VD, Lourenco L, Yip B, Hogg RS, Phillips P, Montaner JS. AIDS incidence and AIDS-related mortality in British Columbia, Canada, between 1981 and 2013: a retrospective study Lancet HIV. 2015;2(3):e92-7.

3. Granich R, Gupta S, Hersh B, Williams B, Montaner J, Young B, et al. Trends in AIDS deaths, new infections and ART coverage in the top 30 countries with

the highest AIDS mortality Burden; 1990-2013. PLoS One. 2015;10(7): e0131353.

4. Siddiqi AE, Hall HI, Hu X, Song R. Population-based estimates of life expectancy after HIV diagnosis. United States 2008-2011. J Acquir Immune Defic Syndr. 2016;72:230-6.

5. AntiretroviralTherapy Cohort Collaboration. Life expectancy of individuals on combination antiretroviraltherapy in high-income countries: a collaborative analysis of 14 cohort studies. Lancet. 2008;372(9635):293-9.

6. Johnson LF, Mossong J, Dorrington RE, Schomaker M, Hoffmann CJ, Keiser O, et al. Life expectancies of South African adults starting antiretroviraltreat-ment: collaborative analysis of cohort studies. PLoS Med. 2013;10(4):e1001418.

7. van Sighem AI, Gras LA, Reiss P, Brinkman K, de Wolf F; ATHENA national observationalcohort study. Life expectancy of recently diagnosed asymptomatic HIV-infected patients approaches that of uninfected individuals. AIDS. 2010;24 (10):1527-35.

8. Lewden C, Salmon D, Morlat P, Bevilacqua S, Jougla E, Bonnet F, et al. Causes of death among human immunodeficiency virus (HIV)-infected adults in the era of potent antiretroviral therapy: emerging role of hepatitis and cancers, persistent role of AIDS. Int J Epidemiol. 2005;34(1):121-30.

9. Louie JK, Hsu LC, Osmond DH, Katz MH, Schwarcz SK. Trends in causes of death among persons with acquired immunodeficiency syndrome in the era of highly active antiretroviraltherapy, San Francisco, 1994-1998. J Infect Dis. 2002;186(7):1023-7.

10. Neuhaus J, Angus B, Kowalska JD, La Rosa A, Sampson J, Wentworth D, et al. Risk of all-cause mortality associated with nonfatalAIDS and serious nonAIDS events among adults infected with HIV. AIDS. 2010;24(5):697-706.

11. Ingle SM, May MT, Gill MJ, Mugavero MJ, Lewden C, Abgrall S, et al. Impact of risk factors for specific causes of death in the first and subsequent years of antiretroviraltherapy among HIV-infected patients. Clin Infect Dis. 2014;59(2):287-97.

12. Smith CJ, Ryom L, Weber R, Morlat P, Pradier C, Reiss P, et al. Trends in underlying causes of death in people with HIV from 1999 to 2011 (D:A:D): a multicohort collaboration. Lancet. 2014;384(9939):241-8.

13. Monforte A, Abrams D, Pradier C, Weber R, Reiss P, Bonnet F, et al. HIV-induced immunodeficiency and mortality from AIDS-defining and non-AIDS-defining malignancies. AIDS. 2008;22(16):2143-53.

14. Weber R, Sabin CA, Friis-Moller N, Reiss P, El-Sadr WM, Kirk O, et al. Liver-related deaths in persons infected with the human immunodeficiency virus: the D:A: D study. Arch Intern Med. 2006;166(15):1632-41.

15. Data collection on Adverse events of Anti-HIV Drugs Study Group, Friis-Mol-ler N, Reiss P, Sabin CA, Weber R, Monforte A, et al. Class of antiretroviraldrugs and the risk of myocardialinfarction. N EnglJ Med. 2007;356(17):1723-35.

16. ObelN, Thomsen HF, Kronborg G, Larsen CS, Hildebrandt PR, Sorensen HT, et al. Ischemic heart disease in HIV-infected and HIV-uninfected individuals: a population-based cohort study. Clin Infect Dis. 2007;44(12):1625-31.

17. Crum-Cianflone N, Hullsiek KH, Marconi V, Weintrob A, Ganesan A, BarthelRV, et al. Trends in the incidence of cancers among HIV-infected persons and the impact of antiretroviraltherapy: a 20-year cohort study. AIDS. 2009;23(1):41-50.

18. Mocroft A, Reiss P, Gasiorowski J, Ledergerber B, Kowalska J, Chiesi A, et al. Serious fataland nonfatalnon-AIDS-defining illnesses in Europe. J Acquir Immune Defic Syndr. 2010;55(2):262-70.

19. Hasse B, Ledergerber B, Furrer H, Battegay M, Hirschel B, Cavassini M, et al. Morbidity and aging in HIV-infected persons: the Swiss HIV cohort study. Clin Infect Dis. 2011;53(11):1130-9.

20. Marin B, Thiebaut R, Bucher HC, Rondeau V, Costagliola D, Dorrucci M, et al. Non-AIDS-defining deaths and immunodeficiency in the era of combination antiretroviraltherapy. AIDS. 2009;23(13):1743-53.

21. Moore RD, Gebo KA, Lucas GM, Keruly JC. Rate of comorbidities not related to HIV infection or AIDS among HIV-infected patients, by CD4 cell count and HAART use status. Clin Infect Dis. 2008;47(8):1102-4.

22. Deeks SG, Phillips AN. HIV infection, antiretroviraltreatment, ageing, and non-AIDS related morbidity. BMJ. 2009;338:a3172.

23. Krishnan S, Schouten JT, Jacobson DL, Benson CA, Collier AC, Koletar SL, et al. Incidence of non-AIDS-defining cancer in antiretroviral treatment-naive subjects after antiretroviral treatment initiation: an ACTG longitudinal linked randomized trials analysis. Oncology. 2011;80(1-2):42-9.

24. Ross AC, Rizk N, O'Riordan MA, Dogra V, El-Bejjani D, Storer N, et al. Relationship between inflammatory markers, endothelial activation markers, and carotid intima-media thickness in HIV-infected patients receiving antiretroviral therapy. Clin Infect Dis. 2009;49(7):1119-27.

25. Triant VA, Meigs JB, Grinspoon SK. Association of C-reactive protein and HIV infection with acute myocardialinfarction. J Acquir Immune Defic Syndr. 2009;51(3):268-73.

26. Baker JV, Peng G, Rapkin J, Abrams DI, Silverberg MJ, MacArthur RD, et al. CD4+ count and risk of non-AIDS diseases following initialtreatment for HIV infection. AIDS. 2008;22(7):841-8.

27. Phillips AN, Neaton J, Lundgren JD. The role of HIV in serious diseases other than AIDS. AIDS. 2008;22(18):2409-18.

28. Data collection on Adverse events of Anti-HIV Drugs Study Group, Smith C, Sabin CA, Lundgren JD, Thiebaut R, Weber R, et al. Factors associated with specific causes of death amongst HIV-positive individuals in the D:A:D Study. AIDS. 2010;24(10):1537-48.

29. Antiretroviral Therapy Cohort Collaboration, Mocroft A, Sterne JA, Egger M, May M, Grabar S, et al. Variable impact on mortality of AIDS-defining events diagnosed during combination antiretroviraltherapy: not allAIDS-defining conditions are created equal. Clin Infect Dis. 2009;48(8):1138-51.

30. May MT, Ingle SM, Costagliola D, Justice AC, de Wolf F, Cavassini M, et al. Cohort profile: AntiretroviralTherapy Cohort Collaboration (ART-CC). Int J Epidemiol. 2014;43(3):691-702.

31. The CoDe ("Coding of Death in HIV") Project. Available from: http://www.c phiv.dk/Tools-Standards/CoDe/About. Accessed 8 January 2016

32. Centers for Disease Controland Prevention. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Recomm Rep. 1992;41(RR-17):1-19.

33. Robins JM, Hernan MA, Brumback B. Marginal structural models and cau-salinference in epidemiology. Epidemiology. 2000;11(5):550-60.

34. D'Agostino RB, Lee ML, Belanger AJ, Cupples LA, Anderson K, KannelWB. Relation of pooled logistic regression to time dependent Cox regression analysis: the Framingham Heart Study. Stat Med. 1990;9(12):1501-15.

35. Lee SH, Kim KH, Lee SG, Cho H, Chen DH, Chung JS, et al. Causes of death and risk factors for mortality among HIV-infected patients receiving antiretroviral therapy in Korea. J Korean Med Sci. 2013;28(7):990-7.

36. Masia M, Padilla S, Alvarez D, Lopez JC, Santos I, Soriano V, et al. Risk, predictors, and mortality associated with non-AIDS events in newly diagnosed HIV-infected patients: role of antiretroviraltherapy. AIDS. 2013;27(2):181-9.

37. Farahani M, Mulinder H, Farahani A, Marlink R. Prevalence and distribution of non-AIDS causes of death among HIV-infected individuals receiving

antiretroviraltherapy: a systematic review and meta-analysis. Int J STD AIDS. 2017;28:636-50.

38. Chung WS, Lin CL, Hung CT, Chu YH, Sung FC, Kao CH, et al. Tuberculosis increases the subsequent risk of acute coronary syndrome: a nationwide population-based cohort study. Int J Tuberc Lung Dis. 2014;18(1):79-83.

39. Huaman MA, Henson D, Ticona E, Sterling TR, Garvy BA. Tuberculosis and cardiovascular disease: linking the epidemics. Trop Dis TravelMed Vaccines. 2015;1:10.

40. Morris AM, Huang L, Bacchetti P, Turner J, HopewellPC, Wallace JM, et al. Permanent declines in pulmonary function following pneumonia in human immunodeficiency virus-infected persons. The Pulmonary Complications of HIV Infection Study Group. Am J Respir Crit Care Med. 2000;162(2 Pt 1):612-6.

41. Reddy KP, Parker RA, Losina E, Baggett TP, Paltiel AD, Rigotti NA, et al. Impact of cigarette smoking and smoking cessation on life expectancy among people with HIV: a US-based modeling study. J Infect Dis. 2016;214(11):1672-81.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article:

Table S1. Missing cause of death codes by cohort Table S2. Unadjusted models for hazard ratios of cause-specific mortality

Table S3. Adjusted models for hazard ratios of cause-specific mortality (non-marginalstructuralmodels)+ Table S4. Adjusted marginalstructuralmodels for hazard ratios of cause-specific mortality+ (excluding those with ADE at or prior to baseline)