S a hemoglobin concentration of 85 g/L for both sexes [22].Statistical MethodsBaseline hypoalbuminemia was defined as a serum albumin concentration of 35 g/L, which is the generally used clinical definition that was also used in previous studies of HIVinfected individuals [23]. We examined risk factors for baseline hypoalbuminemia with log-binomial regression models to obtain risk ratio estimates [24, 25]. In a few instances, the models did not converge, and log-Poisson models, which provide consistent but not fully efficient estimates of the relative risk and its confidence intervals (CIs), were used [26]. Variables included in this analysis were sex and baseline age (30, 309, 409, and 50 years), BMI (16.0 [severely underweight], 16.08.4 [underweight], 18.55.0 [normal], and 25.0 [overweight]), WHO HIV disease stage (I/II, III, and IV), CD4+ T-cell count (50, 509, 10099, and 200 cells/ L), alanine transaminase (ALT) concentration (40 and 40 IU/L), and hemoglobin concentration (8.5, 8.51, and 11 g/dL). Effect modification was assessed with the likelihood ratio test, using variables included in the risk factor analyses listed above. Proportional hazard models were used to investigate the relationship between hypoalbuminemia at baseline and death, the incidence of WHO HIV stage IV disease or death, and the first occurrence of morbidity and nutritional outcomes, including pneumonia, oral thrush, pulmonary tuberculosis, chronic diarrhea, Kaposi sarcoma, EP tuberculosis, severe anemia, anemia, wasting, and 10 weight loss [27]. We excluded individuals who received a diagnosis of the event of interest at baseline. Individuals without events were censored at the date of last follow-up. The results of these analyses are reported as adjusted hazard ratios (HRs) and 95 CIs. In secondary analyses, we conducted continuous analyses of baseline serum albumin concentration to examine whether thegenerally used definition of hypoalbuminemia captured mortality and morbidity relationships for HIV-infected individuals, since the cutoff may differ by disease.Conivaptan hydrochloride The possible nonlinear relationship between serum albumin concentration and death and morbidity outcomes was examined nonparametrically with restricted cubic splines [28, 29].1,2-Dioleoyl-sn-glycero-3-phosphoethanolamine Tests for nonlinearity used the likelihood ratio test, comparing the model with only the linear term to the model with the linear and the cubic spline terms.PMID:24275718 We also present categorical analyses, using cutoffs suggested by continuous analyses. Associations between serum albumin concentration at ART initiation and CD4+ T-cell counts were analyzed with generalized estimating equations (GEEs). Change in CD4+ T-cell count between consecutive visits was treated as a longitudinal continuous outcome, and baseline serum albumin concentration time since ART initiation, and baseline covariates (excluding baseline CD4+ T-cell count) were treated as explanatory variables. We used an m-dependent working correlation matrix (m = 1), which assumes the correlation coefficients of adjacent observations are nonzero and equal. Robust estimators of the variances, which are consistent estimators even if the working correlation matrix is misspecified, were used to construct CIs. The potential nonlinear relationship of the change in T-cell counts between consecutive visits over time was examined nonparametrically with restricted cubic splines for time since ART initiation [28, 29]. If a nonlinear relationship was found, we added the selected cubic splin.