‘Drug holidays’ best predictor of treatment failure, adherence studies find

Gus Cairns
Published: 23 November 2009

Failure to take every dose on time is not a good guide to whether patients will fail their therapy, according to studies from several countries presented at the European AIDS Conference.

A better guide, and one that correlated with the number of pills taken, was whether patients had taken ‘drug holidays’ defined as at least two days of missed therapy.

A study from Italy (Ammassari) asked all 584 patients on HAART at a Rome clinic whether they had taken all doses in the last week, been more than two hours late with a dose, had a complete treatment interruption in the last three months, or taken a treatment break of at least two days in the last month.

The proportions answering ‘yes’ to these questions were:

  • Missed dose, last week: 19%
  • Late dose, last week: 35%
  • Interruption in last three months: 11%
  • ‘Drug holiday’ in last month: 9%

Using a sensitive viral load assay, it stratified patients into those with viral loads over 40 copies/ml, those with under 40 copies/ml but ‘detectable’ (nowhere defined on the poster, but reported in another study (Loftis) to be above ten copies/ml with this assay), and completely undetectable.

The point of doing this is that most patients have some degree of clinically-irrelevant residual viremia of the order of 2-3 copies/ml, but anything over ten copies/ml may have clinical relevance.

Fifty-five per cent of patients had viral loads under 10 copies/ml, 22% had viral loads between 10 and 40 copies/ml, and 22% had viral loads over 40 copies/ml.

A high proportion of patients (63%) had some level of poor adherence by this study’s measures, even 60% of those with viral loads under 10 copies/ml, indicating that strictly-defined ‘poor adherence’ may not be a reliable guide to viral control. Women were the best adherers, with injecting drug users twice as likely to report poor adherence and, in this study at least, gay men three times more likely.

There were only two independent predictors of having a viral load over 10 copies per ml in this study. Drug holidays of more than two days in the last month were associated with a fivefold greater risk of detectability (odds ratio 0.18), and an NNRTI-based regimen, as opposed to a protease inhibitor-based one, was associated with twice the likelihood of undetectability (the proportion of patients on boosted PIs was not stated).

“It is likely that for deep HIV replication control, drug holidays, in contrast to other non-adherence behaviours, might determine higher virological rebound with more pronounced impact on virological control,” the authors state.

Another study by the same team, restricted to patients who had taken HAART for more than four years (Trotta), confirmed that although each indicator of poor adherence produced a 60% greater likelihood of a detectable viral load (over 50 copies/ml in this study), the only independent predictor of viral load failure was a drug holiday of at least two days in the last month. This was associated with a more than sixfold decline in the chance of virological success (OR 0.16).

Studies from resource-poor settings have shown that timely adherence counselling can improve patients’ difficulties and almost eliminate virological failure, but what are the most reliable indicators of poor adherence? Doctors may miss poor adherers due to their own assumptions and patients’ tendency to overestimate their own adherence when asked.

Researchers from Brazil (de Souza) gave a short questionnaire (average time to complete, three minutes) to 441 patients asking them to rate their own adherence on a visual analogue scale in which they indicated it by putting a mark on a line stretching from zero to ten. It also asked whether they had missed any dose or been more than two hours late with a dose in the last three days (a deliberately short timespan was used to increase the reliability of patients’ memories), or had missed doses last weekend. It correlated the answers with viral load and with refill reports from the pharmacy.

Eighty per cent of the patient group had an undetectable viral load. In multivariate analysis the visual analogue scale, associated with a fourfold greater risk of failure, was the only independent predictor of failure in the questionnaire (odds ratio 4.07).

This correlated with having refills of drugs more than 40 days apart (>33% longer than the usual refill interval of 30 days – odds ratio 2.32).

The researchers concluded that self-reported adherence could be a reliable guide to poor adherence as long as it was indicated by a visual analogue scale and correlated with pharmacy returns.

An interesting study from Spain (Quintana Basterra) found evidence that drug regimens were growing more ‘forgiving’ of relatively poor adherence. It studied more than 1500 patients starting therapy (all within the last four months) in the four years between 2005 and 2008, with roughly 380 patients studied per year. It determined adherence by pharmacy returns.

In 2005 only two-thirds of patients with better than 90% adherence had viral loads below 50 copies/ml. In 2006 and 2007 this rose to 82%, and in 2008 it was 87.5%.

Patients who took 80-90% of their doses were, between 2005 and 2007, roughly 15% less likely to achieve an undetectable viral load: roughly 50% in 2005 and 60% in 2006-7 had a viral load under 50 copies/ml. But in 2008 the proportion of patients in this adherence stratum who were undetectable was 91%: at least as good as patients with over 90% adherence.

However less than 80% adherence was still associated with poor suppression, with only 52% of these patients undetectable in 2008.

Do patients find a one-pill, once-a-day regimen easier than taking two pills once a day? Oddly, despite the first once-a-day triple tenofovir/FTC/efavirenz pill , Atripla, being introduced two years ago, a direct comparison has not been made between taking the individual drugs and the singe pill.

A study of 441 patients who were switched from efavirenz plus tenofovir plus 3TC or FTC, or efavirenz plus Truvada, to Atripla, found that there was a trend towards adherence improvement which was sustained for six months, with the average rate of adherence improving from 93.8% to 96.2%.

There was a just-significant improvement in quality of life as assessed by the standard SF-36 questionnaire (proportion rating QoL as ‘good’ or ‘excellent’ rising from 68.8% to 72.7%, p=0.042).

The number of reported side effects and adverse events decreased significantly over time. Just 10% had no adverse events at baseline, 18% reported none in month 1, 23% reported none in month 3, and 30% reported none by month 6.

Patients` perceptions of tolerability (p=0.003), convenience (p=0.05), simplicity (p=0.033) and potency (p=<0.0001) improved after the first month, but only continued to improve thereafter in terms of tolerability and potency.

Finally, a study from Tanzania (Kimambo) looked both at patients’ willingness to take ARVs and their adherence once started, and the way perceptions of body shape change this.

It compared 52 people who had started HAART with 53 who had not. The patients on HAART were slightly older (median age, 41 versus 38), somewhat more likely to be men (25% versus 19%) and not unexpectedly had higher CD4 counts (median: 296 cell/mm3 versus 160 cells/mm3).

Patients on ART had significant fat gain, especially women. Three-quarters of patients on ART said they had perceived body shape changes, with equal numbers citing peripheral fat loss (lipoatrophy: 48%) and central fat gain (lipohypertrophy: 50%). Twenty-one per cent of patients not on ART also perceived body fat changes.

Only six per cent of the patients on ART said body shape change had affected their willingness to take ART but 36% of patients not on ART said that concern about body shape change made them reluctant to take ART.

A quarter of patients with perceived body shape changes (whether taking ART or not) said it affected their willingness to take ART ‘a lot’ and another 25% said it affected their willingness ‘a bit’.


De Souza CAP et al. A simple and reliable assessment of adherence to antiretroviral therapy in clinical practice. 12th European AIDS Conference, Cologne. Abstract PE10.1/5. 2009.

Ammassari A et al. Impact of different types of adherence behaviours and of cART characteristics on plasma HIV-1 RNA detection under the lower limit of quantification at real-time assay. 12th European AIDS Conference, Cologne. Abstract PE10.1/11. 2009.

Loftis A et al. Optimization of Abbott m2000 Real Time HIV-1 Viral Load Assay on Breast Milk, Dried Blood Spots, Seminal Plasma, Cerebrospinal Fluid, Urine, and Cervical Swabs. Fifth IAS conference on HIV Pathogenesis, Treatment and Prevention, Capt Town. Abstract MOPDB103. 2009.

Trotta MP et al. Detrimental effect of current patient-reported drug holidays on long-term sub-optimal treatment outcome in advanced naive subjects initiating cART. 12th European AIDS Conference, Cologne. Abstract PE10.1/10. 2009.

Quintana Basterra A et al. Association of effectiveness and adherence of antiretroviral treatment. 12th European AIDS Conference, Cologne. Abstract PE10.1/6. 2009.

Airoldi M et al. Adherence to and quality of life of a one-pill once-a-day HAART: results of the ADONE study. 12th European AIDS Conference, Cologne. Abstract PE10.1/3. 2009.

Kimambo SJ et al. Acceptance of and adherence to antiretroviral therapy in Tanzania: the influence of lipodystrophy and of traditional medicine. 12th European AIDS Conference, Cologne. Abstract PE9.3/1. 2009.

Community Consensus Statement on Access to HIV Treatment and its Use for Prevention

Together, we can make it happen

We can end HIV soon if people have equal access to HIV drugs as treatment and as PrEP, and have free choice over whether to take them.

Launched today, the Community Consensus Statement is a basic set of principles aimed at making sure that happens.

The Community Consensus Statement is a joint initiative of AVAC, EATG, MSMGF, GNP+, HIV i-Base, the International HIV/AIDS Alliance, ITPC and NAM/aidsmap