E-mail Abstract Author Session Search Abstracts Program


Session 63 Poster Presentations
Relationships between Drug Levels and Their Effects
Session Day and Time: Wednesday 1:30 - 3:30 pm
Room: Hall A


529
Are Random Antiretroviral Drug Levels Associated with Objectively Measured Adherence Behavior?
C. Liechty*1, C. Alexander2, R. Harrigan2, D. Guzman1, E. D. Charlebois1, A. R. Moss1, D. R. Bangsberg1
1Univ of California at San Francisco and 2British Columbia Ctr for Excellence in HIV/AIDS, Vancouver, Canada

Background: Currently available approaches to measure adherence have notable limitations. Several studies have employed random antiretroviral drug levels as a candidate measure of adherence. We conducted a cross-sectional study to test the hypothesis that random measurements of antiretroviral drug levels correlate with level of adherence.

Methods: Antiretroviral drug levels were compared with objectively measured adherence in a prospective cohort of low-income HIV-positive adults (The REACH Cohort). Adherence was measured by unannounced pill counts. Plasma specimens were drawn within 1 month of the most recent adherence assessment. Unannounced pill counts are closely associated with electronic medication monitors (r = 0.91, p < 0.00001, AIDS Beh 2001), and with viral load (r = 0.67, p < 0.001 AIDS 2000) and do not preclude the use of pill box organizers. Random drug levels for nelfinavir, indinavir, efavirenz, and nevirapine were measured using HPLC MS-MS and categorized a priori as abnormally low if below the published reference Cmin – 1 SD. The relationship between normal drug level and adherence above or below the 90% adherence threshold was examined by the Fisher’s Exact test.

Results: Table 1 summarizes our results from 88 individuals. Drug level was significantly associated with adherence dichotomized at 90% (p < 0.05). Except for indinavir, detectable drug level was 100% sensitive for adherence ³ 90%. Specificity was uniformly poor. Overall sensitivity was 91.7% (p < 0.05). Similarly, normal concentrations of all drugs except indinavir had negative predictive values (NPV) of 100% (abnormally low drug level predicting < 90% adherence), but poor positive predictive values. Overall NPV was 82.4% (p < 0.05). Areas under ROC curves ranged from 0.295–0.753.

Discussion: We found that a single abnormally low random drug level can predict adherence below 90%, but normal levels do not reliably indicate adherence above this threshold. This relationship is strong for all drugs studied except indinavir.

 

Table 1. Detectable drug level predicting adherence ³ 90%

 

n

Sensitivity

Specificity

NPV

PPV

AUC

Indinavir

17

57.1%

20%

40.0%

33.3

0.686

Nelfinavir

29

100.0

21.4

100.0

57.7

0.587

Efavirenz

14

100.0

27.3

100.0

27.3

0.295

Nevirapine

28

100.0

35.3

100.0

50.0

0.753

All drugs

88

91.7

26.9

82.4

46.5

0.546

NPV=negative predictive value, PPV=positive predictive value, AUC=area under the ROC curve to predict adherence.