Patient Selection and Characteristics
The 24 selected counties had provided ART to 5,256 patients of whom 926 (18%) were excluded because they received ART for less than 6 months, 23 (0.4%) because they were below 15 years of age, and 3 (<0.1%) because they were not receiving ART for free. Of 4,304 patients remaining, 237 (6%) patients were known to have died, 52 (1%) terminated ART, 84 (2%) were lost-to-follow-up, and 37 (1%) moved away.
From the remaining 3,894 eligible patients, 1,170 (30%) were selected using stratified random sampling as described above. Data from 1,153 patients were available for analysis. Baseline demographic characteristics at the time of ART initiation are shown in . The median age was 39 years (interquartile range [IQR] 34–46) and 56% of patients were male. Reported transmission routes were 59% history of plasma donation, 8% injection drug use, 25% sexual transmission, and 8% other.
Demographic and clinical characteristics of patients by treatment duration group among a stratified random sample of patients in 8 provinces in China, 2007
Treatment Conditions and Outcomes
Median time on ART was 22.8 months (IQR 14.2–35.6) with a median duration of 9.1 months in the 6–12 month group, 18.9 months in the 12–24 month group, and 35.9 months in the ≥24 month group. Initial regimens consisted of NVP or EFV, plus AZT or d4T, plus ddI or 3TC, with 3% on other regimens. In total, 51% of patients were started on 3TC-based regimens and 46% on ddI-based regimens with 3TC use increasing in those initiating ART more recently. In the ≥24 month group, ddI was used in 91%, 82%, 51%, and 15% of regimens at village clinics, township health centers, public health clinics, and county-level or above hospitals, respectively (p<0.001). In the 6–11 month strata, ddI use continued in village and township health centers with 61% and 56% patients starting a ddI-based regimen but declined at public health clinics and county-level hospitals to 4% and 1%, respectively.
Median baseline CD4 cell count was 119 (interquartile range [IQR] 40–199) cells per mm3 with data missing for 21% of participants. Treatment locations were divided between 23% village clinics, 27% township health centers, 29% public health clinics, and 21% county level hospitals. Self-reported 100% adherence in the previous week was 92%.
Viral suppression (HIV viral load < 400 copies/mL) stratified by duration on ART was 82%, 73%, and 67% for participants on ART for 6–11 months, 12–23 months, and ≥24 months, respectively. Probable virologic failure (HIV viral load 1,001–10,000 copies/mL) was found in 6%, 8%, and 12% of patients and definite virologic failure (HIV viral load > 10,000 copies/mL) was found in 10%, 14%, and 18% of patients, among the same 3 treatment duration strata. For 4% of patients, viral load was indeterminate (400–1,000 copies/mL). Virologic outcomes by time on first line ART regimen and treatment facility are shown in for patients receiving 3TC-based regimens, for patients receiving ddI-based regimens, and for the whole study population. At the time of viral load testing, 18%, 13%, and 14% of subjects in the 6–11 month, 12–23 month, and ≥24 month groups, respectively, demonstrated immunologic failure. Discordance between immunologic and virologic responses is shown in . Combined, 66% of patients achieved both a virologic and immunologic response, 6% virologic alone, 20% immunologic alone, and 9% no response.
Figure 1 Virologic response divided by treatment duration and facility type for patients with initial regimens containing lamivudine (A), for patients with initial regimens containing didanosine (B), and for the whole study population (C) among a stratified random (more ...)
Discordance of immunologic and virologic responses to ART by treatment duration group among a stratified random sample of patients in 8 provinces in China, 2007
Predictors of Viral Suppression
Five treatment characteristics were included in the age and gender adjusted multivariate modeling: ART duration, ART regimen, site type, regimen switch, and missed pills in the last 7 days. Adjusted odds ratios for failure to reach viral suppression are shown in . Care received at county level hospitals was superior to that received elsewhere with adjusted odds ratios for failure to reach viral suppression of 5.4 (95% CI, 2.9–10.1) for village clinics, 3.1 (1.7–5.6) for public health clinics, and 3.1 (CI, 1.7–5.6) for township health centers. Patients receiving ddI-based regimens experienced inadequate virologic response more frequently than those receiving 3TC-based regimens (AOR, 3.9; CI, 2.7–5.7). Missing pills in the last 7 days was also associated with inadequate virologic response (AOR, 1.9; CI, 1.2–3.0). Patients in longer treatment duration groups showed an insignificant trend towards higher failure.
Association of clinical and treatment characteristics with inadequate viral response for subjects in four separate logistic regression models for each treatment duration group among a stratified random sample of patients in 8 provinces in China, 2007
When results were stratified into individual models for each treatment duration group, compared to treatment at county level hospitals, treatment at all other facilities in the 12–23 month group and villages in the ≥ 24 month group was associated with failure to reach viral suppression. Compared to 3TC-based regimens, ddI-based regimens showed significantly increased AORs for failure to achieve viral suppression in all three treatment duration groups. Less than 100% adherence was associated with failure to achieve viral suppression only for the 6–11 month treatment group. Having changed regimens was associated with failure in the ≥24 month group.
Further stratification by drug regimen continued to demonstrate association of treatment facility and virologic success. Compared to patients receiving care in county level hospitals, adjusted odds ratios for failure to reach viral suppression, among patients receiving 3TC-based regimens, were 5.3 (CI, 2.1–13.3) for village clinics, 3.4 (1.6–6.9) for public health clinics, and 4.3 (CI, 2–9.4) for township health centers, and among those receiving ddI-based regimens, were 8.3 (CI, 2.1–33.5) for village clinics, 4.4 (CI 1.1–18.0) for public health clinics, and 3.9 (CI, 0.97–15.5) for township health centers.
The above associations continued to show significant relationships upon additional multivariate modeling adjusting for baseline CD4 count for the subset of patients with existing data, with the exception of regimen switch in the ≥24 month model. Upon sensitivity analysis, when indeterminate results were removed, all above findings maintained significant relationships except for facility type and treatment outcome association in the ≥24 month model and public health clinic inferiority in the ddI model.