Accounting for the dead in the longitudinal analysis of income-related health inequalities

This paper develops an accounting framework to consider the effect of deaths on the longitudinal analysis of income-related health inequalities. Ignoring deaths or using inverse probability weights (IPWs) to re-weight the sample for mortality-related attrition can produce misleading results, since to do so would be to disregard the most extreme of all health outcomes. Incorporating deaths into the longitudinal analysis of income-related health inequalities provides a more complete picture in terms of the evaluation of health changes in respect to socioeconomic status. We illustrate our work by investigating health mobility in Quality Adjusted Life Years (QALYs) as measured by the SF6D from 1999 till 2004 using the British Household Panel Survey (BHPS). We show that for Scottish males explicitly accounting for the dead, rather than using IPWs to account for mortality-related attrition, changes the direction of the relationship between relative health changes and initial income position, while for other population groups it increases the strength of this relationship by up to 14 times. When deaths are explicitly incorporated into the analysis it is found that over this five year period for both Scotland and England & Wales the relative health changes were significantly regressive such that the poor experienced a larger share of the health losses relative to their initial share of health and a large amount of this was related to mortality.

HIV-1 coreceptor usage and CXCR4-specific viral load predict clinical disease progression during combination antiretroviral therapy

Background: Although combination antiretroviral therapy (cART) dramatically reduces rates of AIDS and death, a minority of patients experience clinical disease progression during treatment. <p>Objective: To investigate whether detection of CXCR4(X4)-specific strains or quantification of X4-specific HIV-1 load predict clinical outcome. Methods: From the Swiss HIV Cohort Study, 96 participants who initiated cART yet subsequently progressed to AIDS or death were compared with 84 contemporaneous, treated nonprogressors. A sensitive heteroduplex tracking assay was developed to quantify plasma X4 and CCR5 variants and resolve HIV-1 load into coreceptor-specific components. Measurements were analyzed as cofactors of progression in multivariable Cox models adjusted for concurrent CD4 cell count and total viral load, applying inverse probability weights to adjust for sampling bias. Results: Patients with X4 variants at baseline displayed reduced CD4 cell responses compared with those without X4 strains (40 versus 82 cells/mu l; P= 0.012). The adjusted multivariable hazard ratio (HR) for clinical progression was 4.8 [95% confidence interval (Cl) 2.3-10.0] for those demonstrating X4 strains at baseline. The X4-specific HIV-1 load was a similarly independent predictor, with HR values of 3.7(95%Cl, 1.2-11.3) and 5.9 (95% Cl, 2.2-15.0) for baseline loads of 2.2-4.3 and > 4.3 log(10)copies/ml, respectively, compared with < 2.2 log(10)copies/ml. Conclusions: HIV-1 coreceptor usage and X4-specific viral loads strongly predicted disease progression during cART, independent of and in addition to CD4 cell count or total viral load. Detection and quantification of X4 strains promise to be clinically useful biomarkers to guide patient management and study HIV-1 pathogenesis.

A sequential Cox approach for estimating the causal effect of treatment in the presence of time-dependent confounding applied to data from the Swiss HIV Cohort Study

When estimating the effect of treatment on HIV using data from observational studies, standard methods may produce biased estimates due to the presence of time-dependent confounders. Such confounding can be present when a covariate, affected by past exposure, is both a predictor of the future exposure and the outcome. One example is the CD4 cell count, being a marker for disease progression for HIV patients, but also a marker for treatment initiation and influenced by treatment. Fitting a marginal structural model (MSM) using inverse probability weights is one way to give appropriate adjustment for this type of confounding. In this paper we study a simple and intuitive approach to estimate similar treatment effects, using observational data to mimic several randomized controlled trials. Each 'trial' is constructed based on individuals starting treatment in a certain time interval. An overall effect estimate for all such trials is found using composite likelihood inference. The method offers an alternative to the use of inverse probability of treatment weights, which is unstable in certain situations. The estimated parameter is not identical to the one of an MSM, it is conditioned on covariate values at the start of each mimicked trial. This allows the study of questions that are not that easily addressed fitting an MSM. The analysis can be performed as a stratified weighted Cox analysis on the joint data set of all the constructed trials, where each trial is one stratum. The model is applied to data from the Swiss HIV cohort study.

HIV-1 coreceptor usage and CXCR4-specific viral load predict clinical disease progression during combination antiretroviral therapy

BACKGROUND: Although combination antiretroviral therapy (cART) dramatically reduces rates of AIDS and death, a minority of patients experience clinical disease progression during treatment. OBJECTIVE: To investigate whether detection of CXCR4(X4)-specific strains or quantification of X4-specific HIV-1 load predict clinical outcome. METHODS: From the Swiss HIV Cohort Study, 96 participants who initiated cART yet subsequently progressed to AIDS or death were compared with 84 contemporaneous, treated nonprogressors. A sensitive heteroduplex tracking assay was developed to quantify plasma X4 and CCR5 variants and resolve HIV-1 load into coreceptor-specific components. Measurements were analyzed as cofactors of progression in multivariable Cox models adjusted for concurrent CD4 cell count and total viral load, applying inverse probability weights to adjust for sampling bias. RESULTS: Patients with X4 variants at baseline displayed reduced CD4 cell responses compared with those without X4 strains (40 versus 82 cells/microl; P = 0.012). The adjusted multivariable hazard ratio (HR) for clinical progression was 4.8 [95% confidence interval (CI) 2.3-10.0] for those demonstrating X4 strains at baseline. The X4-specific HIV-1 load was a similarly independent predictor, with HR values of 3.7 (95% CI, 1.2-11.3) and 5.9 (95% CI, 2.2-15.0) for baseline loads of 2.2-4.3 and > 4.3 log10 copies/ml, respectively, compared with < 2.2 log10 copies/ml. CONCLUSIONS: HIV-1 coreceptor usage and X4-specific viral loads strongly predicted disease progression during cART, independent of and in addition to CD4 cell count or total viral load. Detection and quantification of X4 strains promise to be clinically useful biomarkers to guide patient management and study HIV-1 pathogenesis.

Non-ignorable loss to follow-up: correcting mortality estimates based on additional outcome ascertainment

Loss to follow-up (LTFU) is a common problem in many epidemiological studies. In antiretroviral treatment (ART) programs for patients with human immunodeficiency virus (HIV), mortality estimates can be biased if the LTFU mechanism is non-ignorable, that is, mortality differs between lost and retained patients. In this setting, routine procedures for handling missing data may lead to biased estimates. To appropriately deal with non-ignorable LTFU, explicit modeling of the missing data mechanism is needed. This can be based on additional outcome ascertainment for a sample of patients LTFU, for example, through linkage to national registries or through survey-based methods. In this paper, we demonstrate how this additional information can be used to construct estimators based on inverse probability weights (IPW) or multiple imputation. We use simulations to contrast the performance of the proposed estimators with methods widely used in HIV cohort research for dealing with missing data. The practical implications of our approach are illustrated using South African ART data, which are partially linkable to South African national vital registration data. Our results demonstrate that while IPWs and proper imputation procedures can be easily constructed from additional outcome ascertainment to obtain valid overall estimates, neglecting non-ignorable LTFU can result in substantial bias. We believe the proposed estimators are readily applicable to a growing number of studies where LTFU is appreciable, but additional outcome data are available through linkage or surveys of patients LTFU. Copyright © 2013 John Wiley & Sons, Ltd.

Trends in Treatment Patterns and Outcomes for Ductal Carcinoma In Situ

BACKGROUND Impact of contemporary treatment of pre-invasive breast cancer (ductal carcinoma in situ [DCIS]) on long-term outcomes remains poorly defined. We aimed to evaluate national treatment trends for DCIS and to determine their impact on disease-specific (DSS) and overall survival (OS). METHODS The Surveillance, Epidemiology, and End Results (SEER) registry was queried for patients diagnosed with DCIS from 1991 to 2010. Treatment pattern trends were analyzed using Cochran-Armitage trend test. Survival analyses were performed using inverse probability weights (IPW)-adjusted competing risk analyses for DSS and Cox proportional hazard regression for OS. All tests performed were two-sided. RESULTS One hundred twenty-one thousand and eighty DCIS patients were identified. The greatest proportion of patients was treated with lumpectomy and radiation therapy (43.0%), followed by lumpectomy alone (26.5%) and unilateral (23.8%) or bilateral mastectomy (4.5%) with significant shifts over time. The rate of sentinel lymph node biopsy increased from 9.7% to 67.1% for mastectomy and from 1.4% to 17.8% for lumpectomy. Compared with mastectomy, OS was higher for lumpectomy with radiation (hazard ratio [HR] = 0.79, 95% confidence interval [CI] = 0.76 to 0.83, P < .001) and lower for lumpectomy alone (HR = 1.17, 95% CI = 1.13 to 1.23, P < .001). IPW-adjusted ten-year DSS was highest in lumpectomy with XRT (98.9%), followed by mastectomy (98.5%), and lumpectomy alone (98.4%). CONCLUSIONS We identified substantial shifts in treatment patterns for DCIS from 1991 to 2010. When outcomes between locoregional treatment options were compared, we observed greater differences in OS than DSS, likely reflecting both a prevailing patient selection bias as well as clinically negligible differences in breast cancer outcomes between groups.

Prevalence of mixed neuropathologies in autopsied older adults and associations with clinical disease progression

Thesis (Ph.D.)--University of Washington, 2016-06

Statistical analysis of survey-based event history data with application to modeling of unemployment duration

Longitudinal surveys are increasingly used to collect event history data on person-specific processes such as transitions between labour market states. Surveybased event history data pose a number of challenges for statistical analysis. These challenges include survey errors due to sampling, non-response, attrition and measurement. This study deals with non-response, attrition and measurement errors in event history data and the bias caused by them in event history analysis. The study also discusses some choices faced by a researcher using longitudinal survey data for event history analysis and demonstrates their effects. These choices include, whether a design-based or a model-based approach is taken, which subset of data to use and, if a design-based approach is taken, which weights to use. The study takes advantage of the possibility to use combined longitudinal survey register data. The Finnish subset of European Community Household Panel (FI ECHP) survey for waves 1–5 were linked at person-level with longitudinal register data. Unemployment spells were used as study variables of interest. Lastly, a simulation study was conducted in order to assess the statistical properties of the Inverse Probability of Censoring Weighting (IPCW) method in a survey data context. The study shows how combined longitudinal survey register data can be used to analyse and compare the non-response and attrition processes, test the missingness mechanism type and estimate the size of bias due to non-response and attrition. In our empirical analysis, initial non-response turned out to be a more important source of bias than attrition. Reported unemployment spells were subject to seam effects, omissions, and, to a lesser extent, overreporting. The use of proxy interviews tended to cause spell omissions. An often-ignored phenomenon classification error in reported spell outcomes, was also found in the data. Neither the Missing At Random (MAR) assumption about non-response and attrition mechanisms, nor the classical assumptions about measurement errors, turned out to be valid. Both measurement errors in spell durations and spell outcomes were found to cause bias in estimates from event history models. Low measurement accuracy affected the estimates of baseline hazard most. The design-based estimates based on data from respondents to all waves of interest and weighted by the last wave weights displayed the largest bias. Using all the available data, including the spells by attriters until the time of attrition, helped to reduce attrition bias. Lastly, the simulation study showed that the IPCW correction to design weights reduces bias due to dependent censoring in design-based Kaplan-Meier and Cox proportional hazard model estimators. The study discusses implications of the results for survey organisations collecting event history data, researchers using surveys for event history analysis, and researchers who develop methods to correct for non-sampling biases in event history data.

A comparison of patient characteristics and 30-day mortality outcomes after transcatheter aortic valve implantation and surgical aortic valve replacement for the treatment of aortic stenosis: a two-centre study

AIMS: It is unclear whether transcatheter aortic valve implantation (TAVI) addresses an unmet clinical need for those currently rejected for surgical aortic valve replacement (SAVR) and whether there is a subgroup of high-risk patients benefiting more from TAVI compared to SAVR. In this two-centre, prospective cohort study, we compared baseline characteristics and 30-day mortality between TAVI and SAVR in consecutive patients undergoing invasive treatment for aortic stenosis. METHODS AND RESULTS: We pre-specified different adjustment methods to examine the effect of TAVI as compared with SAVR on overall 30-day mortality: crude univariable logistic regression analysis, multivariable analysis adjusted for baseline characteristics, analysis adjusted for propensity scores, propensity score matched analysis, and weighted analysis using the inverse probability of treatment (IPT) as weights. A total of 1,122 patients were included in the study: 114 undergoing TAVI and 1,008 patients undergoing SAVR. The crude mortality rate was greater in the TAVI group (9.6% vs. 2.3%) yielding an odds ratio [OR] of 4.57 (95%-CI 2.17-9.65). Compared to patients undergoing SAVR, patients with TAVI were older, more likely to be in NYHA class III and IV, and had a considerably higher logistic EuroSCORE and more comorbid conditions. Adjusted OR depended on the method used to control for confounding and ranged from 0.60 (0.11-3.36) to 7.57 (0.91-63.0). We examined the distribution of propensity scores and found scores to overlap sufficiently only in a narrow range. In patients with sufficient overlap of propensity scores, adjusted OR ranged from 0.35 (0.04-2.72) to 3.17 (0.31 to 31.9). In patients with insufficient overlap, we consistently found increased odds of death associated with TAVI compared with SAVR irrespective of the method used to control confounding, with adjusted OR ranging from 5.88 (0.67-51.8) to 25.7 (0.88-750). Approximately one third of patients undergoing TAVI were found to be potentially eligible for a randomised comparison of TAVI versus SAVR. CONCLUSIONS: Both measured and unmeasured confounding limit the conclusions that can be drawn from observational comparisons of TAVI versus SAVR. Our study indicates that TAVI could be associated with either substantial benefits or harms. Randomised comparisons of TAVI versus SAVR are warranted.

Tenofovir or zidovudine in second-line antiretroviral therapy after stavudine failure in southern Africa

BACKGROUND There is debate over using tenofovir or zidovudine alongside lamivudine in second-line antiretroviral therapy (ART) following stavudine failure. We analyzed outcomes in cohorts from South Africa, Zambia and Zimbabwe METHODS: Patients aged ≥16 years who switched from a first-line regimen including stavudine to a ritonavir-boosted lopinavir-based second-line regimen with lamivudine or emtricitabine and zidovudine or tenofovir in seven ART programs in southern Africa were included. We estimated the causal effect of receiving tenofovir or zidovudine on mortality and virological failure using Cox proportional hazards marginal structural models. Its parameters were estimated using inverse probability of treatment weights. Baseline characteristics were age, sex, calendar year and country. CD4 cell count, creatinine and hemoglobin levels were included as time-dependent confounders. RESULTS 1,256 patients on second-line ART, including 958 on tenofovir, were analyzed. Patients on tenofovir were more likely to have switched to second-line ART in recent years, spent more time on first-line ART (33 vs. 24 months) and had lower CD4 cell counts (172 vs. 341 cells/μl) at initiation of second-line ART. The adjusted hazard ratio comparing tenofovir with zidovudine was 1.00 (95% confidence interval 0.59-1.68) for virologic failure and 1.40 (0.57-3.41) for death. CONCLUSIONS We did not find any difference in treatment outcomes between patients on tenofovir or zidovudine; however, the precision of our estimates was limited. There is an urgent need for randomized trials to inform second-line ART strategies in resource-limited settings.

Assessment of letrozole and tamoxifen alone and in sequence for postmenopausal women with steroid hormone receptor-positive breast cancer: the BIG 1-98 randomised clinical trial at 8·1 years median follow-up.

BACKGROUND: Postmenopausal women with hormone receptor-positive early breast cancer have persistent, long-term risk of breast-cancer recurrence and death. Therefore, trials assessing endocrine therapies for this patient population need extended follow-up. We present an update of efficacy outcomes in the Breast International Group (BIG) 1-98 study at 8·1 years median follow-up. METHODS: BIG 1-98 is a randomised, phase 3, double-blind trial of postmenopausal women with hormone receptor-positive early breast cancer that compares 5 years of tamoxifen or letrozole monotherapy, or sequential treatment with 2 years of one of these drugs followed by 3 years of the other. Randomisation was done with permuted blocks, and stratified according to the two-arm or four-arm randomisation option, participating institution, and chemotherapy use. Patients, investigators, data managers, and medical reviewers were masked. The primary efficacy endpoint was disease-free survival (events were invasive breast cancer relapse, second primaries [contralateral breast and non-breast], or death without previous cancer event). Secondary endpoints were overall survival, distant recurrence-free interval (DRFI), and breast cancer-free interval (BCFI). The monotherapy comparison included patients randomly assigned to tamoxifen or letrozole for 5 years. In 2005, after a significant disease-free survival benefit was reported for letrozole as compared with tamoxifen, a protocol amendment facilitated the crossover to letrozole of patients who were still receiving tamoxifen alone; Cox models and Kaplan-Meier estimates with inverse probability of censoring weighting (IPCW) are used to account for selective crossover to letrozole of patients (n=619) in the tamoxifen arm. Comparison of sequential treatments to letrozole monotherapy included patients enrolled and randomly assigned to letrozole for 5 years, letrozole for 2 years followed by tamoxifen for 3 years, or tamoxifen for 2 years followed by letrozole for 3 years. Treatment has ended for all patients and detailed safety results for adverse events that occurred during the 5 years of treatment have been reported elsewhere. Follow-up is continuing for those enrolled in the four-arm option. BIG 1-98 is registered at clinicaltrials.govNCT00004205. FINDINGS: 8010 patients were included in the trial, with a median follow-up of 8·1 years (range 0-12·4). 2459 were randomly assigned to monotherapy with tamoxifen for 5 years and 2463 to monotherapy with letrozole for 5 years. In the four-arm option of the trial, 1546 were randomly assigned to letrozole for 5 years, 1548 to tamoxifen for 5 years, 1540 to letrozole for 2 years followed by tamoxifen for 3 years, and 1548 to tamoxifen for 2 years followed by letrozole for 3 years. At a median follow-up of 8·7 years from randomisation (range 0-12·4), letrozole monotherapy was significantly better than tamoxifen, whether by IPCW or intention-to-treat analysis (IPCW disease-free survival HR 0·82 [95% CI 0·74-0·92], overall survival HR 0·79 [0·69-0·90], DRFI HR 0·79 [0·68-0·92], BCFI HR 0·80 [0·70-0·92]; intention-to-treat disease-free survival HR 0·86 [0·78-0·96], overall survival HR 0·87 [0·77-0·999], DRFI HR 0·86 [0·74-0·998], BCFI HR 0·86 [0·76-0·98]). At a median follow-up of 8·0 years from randomisation (range 0-11·2) for the comparison of the sequential groups with letrozole monotherapy, there were no statistically significant differences in any of the four endpoints for either sequence. 8-year intention-to-treat estimates (each with SE ≤1·1%) for letrozole monotherapy, letrozole followed by tamoxifen, and tamoxifen followed by letrozole were 78·6%, 77·8%, 77·3% for disease-free survival; 87·5%, 87·7%, 85·9% for overall survival; 89·9%, 88·7%, 88·1% for DRFI; and 86·1%, 85·3%, 84·3% for BCFI. INTERPRETATION: For postmenopausal women with endocrine-responsive early breast cancer, a reduction in breast cancer recurrence and mortality is obtained by letrozole monotherapy when compared with tamoxifen montherapy. Sequential treatments involving tamoxifen and letrozole do not improve outcome compared with letrozole monotherapy, but might be useful strategies when considering an individual patient's risk of recurrence and treatment tolerability. FUNDING: Novartis, United States National Cancer Institute, International Breast Cancer Study Group.

The effect of combined antiretroviral therapy on the overall mortality of HIV-infected individuals.

OBJECTIVE: To estimate the effect of combined antiretroviral therapy (cART) on mortality among HIV-infected individuals after appropriate adjustment for time-varying confounding by indication. DESIGN: A collaboration of 12 prospective cohort studies from Europe and the United States (the HIV-CAUSAL Collaboration) that includes 62 760 HIV-infected, therapy-naive individuals followed for an average of 3.3 years. Inverse probability weighting of marginal structural models was used to adjust for measured confounding by indication. RESULTS: Two thousand and thirty-nine individuals died during the follow-up. The mortality hazard ratio was 0.48 (95% confidence interval 0.41-0.57) for cART initiation versus no initiation. In analyses stratified by CD4 cell count at baseline, the corresponding hazard ratios were 0.29 (0.22-0.37) for less than 100 cells/microl, 0.33 (0.25-0.44) for 100 to less than 200 cells/microl, 0.38 (0.28-0.52) for 200 to less than 350 cells/microl, 0.55 (0.41-0.74) for 350 to less than 500 cells/microl, and 0.77 (0.58-1.01) for 500 cells/microl or more. The estimated hazard ratio varied with years since initiation of cART from 0.57 (0.49-0.67) for less than 1 year since initiation to 0.21 (0.14-0.31) for 5 years or more (P value for trend <0.001). CONCLUSION: We estimated that cART halved the average mortality rate in HIV-infected individuals. The mortality reduction was greater in those with worse prognosis at the start of follow-up.

In-hospital and 1-year mortality in patients undergoing early surgery for prosthetic valve endocarditis.

IMPORTANCE: There are limited prospective, controlled data evaluating survival in patients receiving early surgery vs medical therapy for prosthetic valve endocarditis (PVE). OBJECTIVE: To determine the in-hospital and 1-year mortality in patients with PVE who undergo valve replacement during index hospitalization compared with patients who receive medical therapy alone, after controlling for survival and treatment selection bias. DESIGN, SETTING, AND PARTICIPANTS: Participants were enrolled between June 2000 and December 2006 in the International Collaboration on Endocarditis-Prospective Cohort Study (ICE-PCS), a prospective, multinational, observational cohort of patients with infective endocarditis. Patients hospitalized with definite right- or left-sided PVE were included in the analysis. We evaluated the effect of treatment assignment on mortality, after adjusting for biases using a Cox proportional hazards model that included inverse probability of treatment weighting and surgery as a time-dependent covariate. The cohort was stratified by probability (propensity) for surgery, and outcomes were compared between the treatment groups within each stratum. INTERVENTIONS: Valve replacement during index hospitalization (early surgery) vs medical therapy. MAIN OUTCOMES AND MEASURES: In-hospital and 1-year mortality. RESULTS: Of the 1025 patients with PVE, 490 patients (47.8%) underwent early surgery and 535 individuals (52.2%) received medical therapy alone. Compared with medical therapy, early surgery was associated with lower in-hospital mortality in the unadjusted analysis and after controlling for treatment selection bias (in-hospital mortality: hazard ratio [HR], 0.44 [95% CI, 0.38-0.52] and lower 1-year mortality: HR, 0.57 [95% CI, 0.49-0.67]). The lower mortality associated with surgery did not persist after adjustment for survivor bias (in-hospital mortality: HR, 0.90 [95% CI, 0.76-1.07] and 1-year mortality: HR, 1.04 [95% CI, 0.89-1.23]). Subgroup analysis indicated a lower in-hospital mortality with early surgery in the highest surgical propensity quintile (21.2% vs 37.5%; P = .03). At 1-year follow-up, the reduced mortality with surgery was observed in the fourth (24.8% vs 42.9%; P = .007) and fifth (27.9% vs 50.0%; P = .007) quintiles of surgical propensity. CONCLUSIONS AND RELEVANCE: Prosthetic valve endocarditis remains associated with a high 1-year mortality rate. After adjustment for differences in clinical characteristics and survival bias, early valve replacement was not associated with lower mortality compared with medical therapy in the overall cohort. Further studies are needed to define the effect and timing of surgery in patients with PVE who have indications for surgery.

Self-reported nonadherence to antiretroviral therapy as a predictor of viral failure and mortality

OBJECTIVE: To determine the effect of nonadherence to antiretroviral therapy (ART) on virologic failure and mortality in naive individuals starting ART. DESIGN: Prospective observational cohort study. METHODS: Eligible individuals enrolled in the Swiss HIV Cohort Study, started ART between 2003 and 2012, and provided adherence data on at least one biannual clinical visit. Adherence was defined as missed doses (none, one, two, or more than two) and percentage adherence (>95, 90-95, and <90) in the previous 4 weeks. Inverse probability weighting of marginal structural models was used to estimate the effect of nonadherence on viral failure (HIV-1 viral load >500 copies/ml) and mortality. RESULTS: Of 3150 individuals followed for a median 4.7 years, 480 (15.2%) experienced viral failure and 104 (3.3%) died, 1155 (36.6%) reported missing one dose, 414 (13.1%) two doses and, 333 (10.6%) more than two doses of ART. The risk of viral failure increased with each missed dose (one dose: hazard ratio [HR] 1.15, 95% confidence interval 0.79-1.67; two doses: 2.15, 1.31-3.53; more than two doses: 5.21, 2.96-9.18). The risk of death increased with more than two missed doses (HR 4.87, 2.21-10.73). Missing one to two doses of ART increased the risk of viral failure in those starting once-daily (HR 1.67, 1.11-2.50) compared with those starting twice-daily regimens (HR 0.99, 0.64-1.54, interaction P = 0.09). Consistent results were found for percentage adherence. CONCLUSION: Self-report of two or more missed doses of ART is associated with an increased risk of both viral failure and death. A simple adherence question helps identify patients at risk for negative clinical outcomes and offers opportunities for intervention.