shows selected baseline characteristics of the 1422 HIV-infected subjects included in this study and the cross-sectional associations of IGF-I and IGFPB-3 levels with these characteristics. A low IGF-I level was significantly associated with a low CD4+ T cell count and a high HIV RNA level, albeit not with clinical AIDS. A number of other factors were also associated with IGF-I levels, and, in the final multivariate model, the level of IGF-I was associated with the CD4+T cell count, HIV RNA level, age, race/ethnicity, HCV serostatus, BMI, and the level of IGFBP-3. For IGFBP-3 levels, we observed cross-sectional associations with the level of HIV RNA but not with CD4+ T cell count or clinical AIDS, and, in the final multivariate model, IGFBP-3 levels were significantly associated with HIV RNA levels, age, race/ethnicity, HCV serostatus, and IGF-I levels.
Selected baseline characteristics and their cross-sectional associations with insulin-like growth factor (IGF)–I and IGF-binding protein (IGFBP)–3 levels.
Incident clinical AIDS occurred in 101 of the 1010 women without AIDS at baseline (before April 1996). Cox proportional hazards models were used to assess the associations of circulating IGF-I and IGFBP-3 levels with the risk (hazard ratio [HR]) of AIDS as well as to identify relevant cofactors. Preliminary Cox models that adjusted only for age are shown in . Having a low CD4+T cell count, high HIV RNA level, use of azidothymidine, low BMI, low hemoglobin level, HCV seropositivity, and high AST level were each significant predictors of incident clinical AIDS. IGF-I and IGFBP-3 levels were not significantly associated with the incidence of clinical AIDS in these initial models.
Statistical associations of incident clinical AIDS with patient characteristics and serum insulin-like growth factor (IGF)–I and IGF-binding protein (IGFBP)–3 levels in age-adjusted Cox regression models.
However, in multivariable Cox analysis, the level of IGFBP-3 was highly significant and, as hypothesized, was positively associated with the risk of incident AIDS (). In our a priori model, the hazard ratio contrasting the highest and lowest IGFBP-3 quartiles [HRq4-q1] was 2.65 (95% confidence interval [CI], 1.30–5.42) with adjustment for age, CD4+ T cell count, HCV seropositivity, and the level of IGF-I. Examination of other models containing the level of IGFBP-3 provided similar inferences. For example, IGFBP-3 levels remained significantly associated with the risk of incident clinical AIDS (HRq4-q1, 2.50 [95% CI, 1.20–5.22]) in a model that adjusted for the CD4+ T cell count, HIV RNA level, HCV seropositivity, smoking status, hemoglobin level, age, and the level of IGF-I. A low BMI (<18.5 kg/m2) and abnormal liver function test results were also highly significant predictors of incident clinical AIDS in multivariable models, but inclusion of these predictors did not meaningfully alter the relationship between IGFBP-3 levels and the end point or its statistical significance. The level of IGF-I was not significantly associated with AIDS (although the relationship was in the predicted direction), nor was age, but both were retained in the final model because of their perceived biological relevance to the study of IGFBP-3 and AIDS. The trend in HRs across IGFBP-3 quartiles showed that there was a gradient of increasing risk of incident clinical AIDS with increasing IGFBP-3 level (Ptrend = .02).
Final multivariable Cox model of the associations of insulin-like growth factor (IGF)–I and IGF-binding protein (IGFBP)–3 levels with the risk of AIDS.
shows the associations IGF-I and IGFBP-3 levels with risk of a rapid and persistent decline in CD4+T cell count. There were 83 women who had complete data for the first 3 visits and met our strict case definition—that is, a CD4+ T cell count that (1) at the second and third visits fell into a lower CD4+ T cell stratum than at baseline or (2) remained in the same stratum but was ≥50% lower. The comparison group was 190 women with complete data and a stable or increasing CD4+ T cell count. In multivariable logistic regression, we observed contrasting associations of IGF-I level (odds ratio [OR]q4-q1, 0.32 [95% CI, 0.12–0.83]; Ptrend = .02) and IGFBP-3 levels (ORq4-q1, 2.40 [95% CI, 0.95–6.09]; Ptrend = .02) with the risk of rapid and persistent CD4+ T cell count decline, after adjusting for the starting CD4+ T cell stratum, age, and HCV seropositivity. A similar but non-significant association with IGF-I levels (ORq4-q1, 0.55 [95% CI, 0.25–1.18]; Ptrend = .13) was observed using a more moderate decline of ≥20% (instead of ≥50%) in the case definition (n = 124 cases), whereas the association with IGFBP-3 levels remained significant (ORq4-q1, 2.12 [95% CI, 0.97–5.01]; Ptrend = .01). The findings for a rapid and persisting elevation in HIV RNA level were nonsignificant but showed a similar pattern; that is, the risk of rapid and persisting elevation in HIV RNA levels had a nonsignificant inverse association with the level of IGF-I (ORq4-q1, 0.53 [95% CI, 0.23–1.23]) and a nonsignificant positive association with IGFBP-3 levels (ORq4-q1, 1.72 [95% CI, 0.74–4.04]), after adjusting for starting CD4+ T cell stratum, age, and HCV seropositivity. Inclusion of additional covariates in our multivariable models of CD4+ T cell count or HIV RNA level did not meaningfully alter the above findings (data not shown).
Table 4 Multivariable logistic regression model of the associations of insulin-like growth factor (IGF)–I and IGF-binding protein (IGFBP)–3 levels with a rapid and persistent CD4+ T cell decline (n = 83) vs. a stable or increasing CD4+ T cell (more ...)