Table 5.
Viral Load Monitoring After Antiretroviral Therapy Initiation, Before and After Treat-All Introduction in High/Upper-Middle-Income Countries
| Patient Group | |||
|---|---|---|---|
| Adults (>19 Years)a | Adolescents (10–19 Years)b | Children (<10 Years)c | |
| n | 185 158 | 6373 | 9069 |
| Viral load monitoring after ART initiation, n (%) | 103 273 (55.8) | 3144 (49.3) | 4773 (52.6) |
| Before Treat-All adoption | 83 411 (55.2) | 2447 (49.3) | 4391 (52.6) |
| After Treat-All adoption | 19 862 (58.1) | 697 (49.5) | 382 (52.8) |
| Risk difference at the Treat-All adoption threshold, PP (95% CI)d | 2.9 (.5, 5.4) | −5.0 (−13.5, 3.4) | 7.7 (−3.7, 19.2) |
| P value | .020 | .241 | .186 |
| Imbens-Kalyanaraman bandwidth, days | 275 | 596 | 472 |
| No. within bandwidth | 23 939 | 2141 | 1172 |
| TED (95% CI) | −.018 (−.034, −.002) | −.019 (−.044, .005) | −.014 (−.057, .028) |
| TED P value | .026 | .123 | .515 |
| Predicted outcomes at the Treat-All threshold,d (95% CI) | |||
| Enrollment just before Treat-All adoption (%) | 58.2 (56.4, 60.0) | 54.6 (48.4, 60.8) | 50.7 (42.6, 58.9) |
| Enrollment just after Treat-All adoption (%) | 61.1 (59.5, 62.8) | 49.6 (43.9, 55.3) | 58.5 (50.4, 66.5) |
| Relative change at threshold (%) | 5.0 | −9.2 | 15.2 |
| Slopes before and after Treat-All adoptione | |||
| Percentage point change per year before Treat-All | 3.4 (2.4, 4.3) | 6.3 (1.4, 11.2) | −3.7 (−9.1, 1.7) |
| Percentage point change per year after Treat-All | −4.2 (−5.2, −3.2) | −3.0 (−8.1, 2.0) | −4.7 (−11.7, 2.4) |
| P value for interaction term (difference in slopes) | <.0001 | .009 | .834 |
Abbreviations: ART, antiretroviral therapy; PP, Percentage point; CI, confidence interval; TED, treatment effect derivative.
aCanada, China including Hong Kong SAR, South Africa, South Korea, United States.
bCanada, Malaysia, South Africa, South Korea, Thailand, United States.
cMalaysia, South Africa, Thailand.
dRisk difference and predicted outcomes at the Treat-All threshold are from regression discontinuity analysis estimating the difference in local linear predictions at the threshold (ie, in the limit, as the threshold is approached from above and below). The Imbens-Kalyanaraman bandwidth defines the region in which the relationship between enrollment timing and testing outcome is assumed to be linear in local linear regression models.
eSlope comparison is from separate linear regression models comparing the 2 years before Treat-All adoption and after adoption.