The detection of non-adherence by self-administered questionnaires can be optimized by protease inhibitor plasma concentration determination

Self-reported adherence and plasma protease inhibitor concentration (PPIC) were assessed in 642 HIV-infected patients at month 4 of a single protease inhibitor-containing regimen. PPIC was below the limit of quantification (LOQ) in 32% of non-adherent patients and in 8% of adherent patients. The relationship between non-adherence and a detectable HIV viral load was enhanced when adherent patients with PPIC below LOQ were considered to be non-adherent. PPIC combined with self-report may more reliably detect non-adherence.

Adherence to highly active antiretroviral therapy (HAART) is a crucial determinant for reducing HIV replication [¹]. A dose-response relationship between levels of adherence and undetectable viral loads has been documented [²]. Measuring a patient's adherence remains a difficult task, and several approaches to measure adherence have been evaluated and compared showing specific limitations [^3-5]. To improve adherence assessment, one possible approach has been to combine several measures of adherence [^6,7]. Patients with low plasma protease inhibitor concentrations (PPIC) are more likely to have reported non-adherence behaviours [^8-10]. In this paper, using cross-sectional data at the 4-month visit of the French APROCO cohort, we tested whether a combined measure of adherence based on self-report and the detection of PPIC could better identify individuals with non-adherence behaviours.

The French APROCO study was set up in 1997 to study clinical and sociobehavioural characteristics, including an adherence assessment in HIV-1-infected patients started on a protease inhibitor (PI)-containing regimen. To measure adherence, we used an indicator combining five questions in a self-administered questionnaire to distinguish three groups of patients: highly, moderately and non-adherent. Patients who reported taking 100% of their prescribed antiretroviral doses in the 4 days before the visit were classified as highly adherent. These patients were reclassified as moderately adherent if they had skipped a dose during the past weekend or had 'almost totally' followed their HAART regimen, or had modified the prescribed scheduling several times or had taken all their medication at one time. Patients were considered moderately adherent if they reported missing no more than 20% of their treatment. Patients taking less than 80% of their pills or reporting to be 'partially' or 'not at all' adherent to their HAART prescriptions were considered non-adherent.

Viral load titres were considered undetectable if they were lower than the threshold value specific to each centre where the assay was performed. PPIC was determined using a high-performance liquid chromatography assay. The limits of quantification (LOQ) of PI in the plasma for indinavir, ritonavir, nelfinavir and its active metabolite M8 [nelfinavir (+M8)], and saquinavir were 5, 30, 30 (+30), and 9 ng/ml, respectively. In order to improve the detection of non-adherent patients, patients classified as moderately or highly adherent on the basis of self-report only were reclassified as non-adherent if their PPIC was lower than the LOQ. In addition, we alternatively used the lower limit of the minimal plasma concentration interval to reclassify patients with suboptimal or normal PPIC, taking into account the wide interindividual variability and the lack of information about the time of intake (the values used for indinavir, ritonavir, nelfinavir (+M8), and saquinavir were 50, 900, 600, and 12 ng/ml, respectively) [^11-13].

The APROCO cohort enrolled 1281 patients between May 1997 and June 1999 in 47 French hospitals. Among the 861 patients treated with a single PI-containing regimen, 642 patients with complete data about viral load, PPIC and adherence were considered for this analysis.

Among the 642 patients, 305, 85, 173, and 79 were respectively treated with indinavir, ritonavir, nelfinavir, and saquinavir, respectively. Of these, 62 were classified as non-adherent, 236 as moderately adherent and 344 as highly adherent (Table 1). The proportion of patients with PPIC below the LOQ was 32.3% of the 62 non-adherent patients, whereas it was only 7.6% among the 580 highly or moderately adherent patients. Of the 642 patients, only 64 (10.0%) had PPIC below the LOQ: 20 patients (31.2%) were classified as non-adherent, whereas 19 (29.7%) and 25 (39.1%), respectively, were classified as highly or moderately adherent. As these latter 44 patients with PPIC below the LOQ were likely to have been non-adherent in the past 24 h, they were reclassified as non-adherent in the combined definition (Table 1). When calculating the association between adherence (as expressed by self-report only or by the combined definition) and detectable viral load, the odds ratio estimation for non-adherence increased to 0.7, showing that there should be a portion of non-adherent patients who are captured by the absence of PI in plasma (Table 1). When using as a threshold for PPIC the minimal plasma concentration, 101 patients (15.7%) had low PPIC: 34 (33.7%) and 42 (41.6%), respectively, were classified as highly and moderately adherent. When reclassifying these 76 patients as non-adherent for constructing an alternative combined definition of adherence, odds ratio estimates representing the association between non-adherence and a detectable viral load were similar to those corresponding to self-report only (Table 1).

Table 1

Self-administered questionnaires have been shown not to be sensitive enough to capture all dimensions of non-adherence [⁵]. However, as already reported, self-reported adherence is significantly associated with virological success. A slight but not significant increase in the strength of the association between self-reported non-adherence and detectable viral load was obtained when patients' self-report was combined with PPIC below the LOQ. According to the combined definition, non-adherent patients have three times the risk of having detectable levels of plasma viral load at month 4, whereas such a risk was 2.3 for non-adherent patients, as defined by self-reports only.

Two-thirds of non-adherent patients had PPIC greater or equal to the LOQ, showing that using PPIC only in terms of the presence or absence of PI is not sensitive enough to detect non-adherence behaviour. An alternative cut-off for PI interpretation (minimal plasma concentration) to reclassify non-adherent patients did not improve the association between non-adherence and a detectable viral load. As our data refer to the first generation of PI without ritonavir boosting (i.e. PI with short half-lives), this result could be expected, also considering that detectable PPIC lower than the minimal plasma concentration may occur in patients who did not take the last dose at the scheduled time. These minor deviations may have more negligible consequences on virological outcome than skipping an entire dose.

We would like to emphasize that using PPIC as a measure of adherence is reductive. The interest in PPIC mainly relies on the detection and management of pharmacological interactions, adverse events and therapeutic drug monitoring.

Because there is still no gold standard to assess adherence to HAART, implementing patients self-report by drug plasma concentrations may provide a useful tool to identify non-adherence behaviour and to adapt therapeutic strategies for HIV care.