Over a nine year period of observation, the median follow-up duration for each inmate was 31 months (inter-quartile range: 6 months - 9.5 years). The 512 individuals had a median of 5 jail stays (range 2–20), and 36% had more than five visits in jail. The average jail stay length was 104 days (3.5 months) The median age of the inmates at study entry was 36 years (inter quartile range 19–66 years) (). A majority of the participants were African Americans (51%). Men accounted for 86% of the cohort. Inmates in the continuous ART group were the oldest. In all three groups of inmates, CD4 and VL were considered at baseline in jail. More than three-quarters (76%) of the inmates took intermittent ART; 9.0% refused ART and the remaining 15% took continuous ART throughout the study period. Using chi-square tests, we tested for differences amongst three ART groups (). At baseline, factors that were significant between the three groups are age, baseline VL (p<0.05). Some other significant factors from are follow up time in jail, exit CD4, exit VL (p<0.05) Over time, based on results of final models (, ), the factors that were associated with category of ART included age, gender, ethnicity, time on treatment, baseline CD4 and baseline VL.
Characteristics of study participants by treatment pattern: continuous, Intermittent and never on treatment.
Multivariate associations between demographic and clinical characteristics and CD4 cell count levels. Continuous treatments at study entry are the referent state. Model 1: CD4.
Model 2: VL. Multivariate associations between demographic and clinical characteristics and viral load(VL). Continuous treatments at study entry are the referent state.
Change in CD4 cell counts over time across treatment groups
There were strong associations between ART groups with CD4 cell counts and HIV VL over time consistent with the use of guidelines to offer ART for lower CD4 cell counts and higher VL. The interaction terms for ART groups and time were significant for intermittent ART and never on ART groups. ()
On an average, continuously treated inmates gained an average adjusted of 0.67 CD4 cells per month. The difference (95% CI) in the adjusted rate of change in the intermittently treated inmates as compared to continuously treated inmates was −1.60 (−2.13, −1.06), resulting in the intermittent ART group inmates lost CD4 cells at an average adjusted rate of 0.93 cells per month. The difference (95% CI) in the adjusted rate of change in the never on ART inmates as compared to continuously treated inmates was −1.97 (−3.00, −0.96), so that the never-treated inmates lost CD4 cells at an average adjusted rate of 1.29 cells per month. However, this difference was not statistically significant. (p
0.33). At baseline, the differences in CD4 counts and VL among the three ART groups were statistically significant (p<0.001; )
As observed in , significant differences in the rate of change of treatment groups were observed over time (variables intermittent* time, never on treatment*time; as indicated by p values<0.001). Variable time (months) was significant (p value <0.007). Further interpreting other significant variables like age in the models, if we were to compare two inmates who were the same with respect to treatment group, time since baseline, ethnicity, and gender but differed in age by 1 year, we would expect the CD4 cell counts to differ by 0.05 (−1.29, 1.38) cells. Furthermore, if we were to compare two inmates who were the same with respect to treatment group, time since baseline, ethnicity, and age but of different gender ( female vs. male), then, we would expect CD4 cell count to differ by 2.73 (−27.77, 33,23) cells.
and illustrates the un-adjusted and adjusted changes in CD4 cell counts for the three ART groups. In , the CD4 cell counts increase in the continuous ART group and fall most rapidly in the never on ART inmates. In , a descriptive summary of the data is presented and the VL is not held constant (unadjusted).
Local average of CD4 counts over time by treatment pattern: intermittent, continuous, and never-on treatment.
Expected CD4 over time by for a 30 year-old Caucasian male, as predicted from a linear mixed effects model.
In , we have attempted to more clearly illustrate the effect of each ART group. VL was held constant in deriving these predicted trajectories (adjusted). The baseline CD4 and the VL were selected to reflect typical (baseline) values for each of the ART groups. The estimated course of CD4 cell count over time predicted by the model for a 30 year old male with a constant VL is shown. Note the trajectories in each of the ART groups in are broadly similar to the observed trajectories in . We would not expect these to coincide exactly since VL was held constant over time in , which is not the case in the unadjusted, descriptive summary of the data shown in . These results suggest that compared to inmates that were on never on ART, the inmates in the continuous ART group documented the best response overall in terms of gain in CD4 cells over time, followed by the inmates in the Intermittent ART group with a slower rate of loss of CD4 cells over time.
Changes in HIV VL over time across treatment groups
Overall the HIV VL decreased by 6% per year in each of the three treatment groups. From the model, () we can infer that the VL of intermittent ART inmates was approximately one and a half times and of never-on ART inmates three times greater than those on continuous ART. After adjusting for the different covariate distributions, although there was significant differences in the change in HIV VL over time between Never on treatment group and referent continuous treatment group, (2.74 (1.50, 5.00); p value <0.001), these differences between intermittent treatment and continuous treatment group were non significant (1.34 (0.92–1.96); p value <0.133).Furthermore, Baseline CD4, Baseline VL, and CD4 over time, were found to significantly differ between treatment groups (p<0.05) and a non- significant effect of time was also observed (0.99(0.98,1.00) p value<0.151).
In , as in , the un-adjusted local average observed trajectories of VL in three treatment groups have been illustrated.
Local average of viral load (VL) over time by treatment pattern: intermittent, continuous, and never-on treatment.