210 likes | 317 Views
CD4 criteria improves the sensitivity of a clinical algorithm developed to identify viral failure in HIV-positive patients on first-line antiretroviral therapy. Denise Evans, Mhairi Maskew, Lynne McNamara, Patrick MacPhail, Christopher Mathews, Ian Sanne, Matthew Fox.
E N D
CD4 criteria improves the sensitivity of a clinical algorithm developed to identify viral failure in HIV-positive patients on first-line antiretroviral therapy Denise Evans, Mhairi Maskew, Lynne McNamara, Patrick MacPhail, Christopher Mathews, Ian Sanne, Matthew Fox This research has been supported by the President's Emergency Plan for AIDS Relief (PEPFAR) through South Africa Mission of the US Agency for International Development (USAID) under the terms of grant number 674-A-00-08-00007-00
Background • Routine viral load monitoring is neither affordable nor available in most resource-limited settings • Previous attempts to identify low-cost surrogate markers of virologic failure have shown conflicting results(Lynen et al., 2009; Robbins et al., 2010; Meya et al., Johannessen et al., 2008; Lawn et al., 2006) • Need to develop simpler, cheaper monitoring strategies that can be administered by minimally trained clinic personnel • Used data from a large South African clinic to try to develop an algorithm using accessible, inexpensive, routinely collected markers to identify virological failure
Methods • Study site and subjects • Themba Lethu Clinic, Johannesburg, South Africa, >20,000 patients on ART • Included non-pregnant, HIV+ ART naïve adults (≥18) • Initiated on public sector first-line regimen, April 2004 – February 2010 • Limited to those with ≥2 viral load and >6 months on ART • Study data • Included inexpensive, routinely collected baseline and follow-up markers • e.g. BMI, blood pressure, WHO staging, hemoglobin, albumin, mean cell volume, new conditions, CD4 count etc. • WHO failure criteria: WHO stage III/IV, CD4 < baseline or < 100 cells/mm3 • Matched virologic failures to comparison group (1:3) on person-time • Outcomes • Virological failure: 2 HIV-RNA ≥400 copies/ml after suppression Human Research Ethics Committee of the University of the Witwatersrand (HREC M060623/M110140/M10418)
Statistical analysis • Cox PH model to identify important predictors of viral failure to create risk score • Univariate p < 0.1 included (age and gender) • Risk score calculated by assigning rounded aHR for each predictor* • E.g. if low BMI has aHR 1.75 = +2 • For each subject, total score is sum of risk scores • Low risk (0-4), Medium/High risk (≥ 4)** • Assessed diagnostic accuracy of risk groups with virologic failure as gold standard • Se, Sp, PPV, NPV • Model with and without CD4 criteria Themba Lethu Clinic, Johannesburg *Spiegelhalter and Knill-Jones method; Berkley et al., 2003; Lynen et al., 2009 **Robbins et al., 2010
Multivariate risks scores (HRs) Baseline
Multivariate risks scores (HRs) Baseline 6-12 months before failure
Virologic failure by risk score √ WHO stage III/IV, CD4 < baseline or CD4 < 100cells/mm3 *Lynen et al., 2009; Meya et al., 2009
Virologic failure by risk score √ WHO stage III/IV, CD4 < baseline or CD4 < 100cells/mm3 *Lynen et al., 2009; Meya et al., 2009
Virologic failure by risk score √ WHO stage III/IV, CD4 < baseline or CD4 < 100cells/mm3 *Lynen et al., 2009; Meya et al., 2009
Virologic failure by risk score √ WHO stage III/IV, CD4 < baseline or CD4 < 100cells/mm3 *Lynen et al., 2009; Meya et al., 2009
Conclusions • Average sensitivity and poor specificity of virologic failure • Similar to previous scoring systems • Total score without CD4 criteria improved Se over WHO criteria or WHO stage III/IV • CD4 criteria further improves Sensitivity • Limitations • Missing values (20-25%) • Data from a single site (validate) • Could only evaluate what actually happened, not whether method could predict future events • Looking at further refinements of the model • Could be useful to screen for risk of virologic failure • Absence of routine viral load testing, targeted laboratory testing
Acknowledgements NIH and USAID HE2RO/CHRU – WITS Health Consortium Hazel Molefe Daphne Radebe Bontle Mahlatsi Desiree Louw Keagile Komane Frank Phakathi University of California – San Diego Right to Care Patients and staff at Themba Lethu Clinic – Helen Joseph Hospital This research has been supported by the President's Emergency Plan for AIDS Relief (PEPFAR) through South Africa Mission of the US Agency for International Development (USAID) under the terms of grant number 674-A-00-08-00007-00
Se/Sp of each score excluding variables with < 80% of the data