Until it is possible for individuals with HIV to safely stop taking antiretrovirals, those who have started treatment must continue for the foreseeable future. Many factors contribute to the success or failure of a long-term treatment regimen. Among the most basic of these are the health care provider's ability to prescribe drugs correctly and the individual's ability to adhere to the prescribed regimen.

Buyer Beware

This June, a national survey by Johns Hopkins University and the University of California at San Francisco of 476 physicians found that between January and March of 1998, 25% of treatment-naïve patients were started on therapy inconsistent with the Public Health Service Guidelines for the Use of Antiretroviral Agents (for more on the guidelines, see Treatment Issues, Winter 1997/98). Doctors with the most experience in treating HIV more consistently prescribed according to the guidelines. Close to 90% of the experienced physicians prescribed three or more medications for their treatment-naïve patients, compared to 60% of physicians with the least experience, who prescribed an average of two medications.

Another concern was highlighted in a study presented in September at the Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC) by Bonnie Purdy, Pharm.D., and colleagues, of the Albany Medical Center (abstract I-169). The researchers found that health care providers frequently prescribed medications incorrectly. One hundred prescribing errors in 81 patients were documented over a 31-month period. The most common mistakes were over- or under-dosing. Dr. Purdy cited examples such as nelfinavir prescribed at 250 mg three times a day instead of 750 mg three times a day and ddI at 400 mg every four hours instead of 400 mg total per day.

Seventy-three percent of errors were considered serious or severe and 21% were noted as clinically significant. The mistakes appeared to result from the increasing complexity of treatment regimens and error rates increased from 2% in 1996 to 7% in 1997 to 14% in the first seven months of 1998. Dr. Purdy stated that interns (first-year residents) wrote most of the prescriptions. This suggests that not only do less experienced health care providers prescribe treatments that are not consistent with the recommended guidelines, but when writing up the prescription orders, there is a potential for further errors in dosing or administration.

Measuring Adherence

Another factor that affects the outcome of a treatment regimen is the patient's own capacity for adherence. At ICAAC, Margaret Chesney, Ph.D., from the University of California at San Francisco, reported on several different methods employed to track adherence (mini-lecture I-168). Dr. Chesney pointed out that there is no gold standard and each method has flaws.

Patient self-report is perhaps the simplest technique. It does overestimate adherence. However, it tends to be 99% accurate when a patient reports nonadherence and should be considered seriously. Since patients may experience memory problems, it is most effective when focused on the recent past. Computer-conducted interviews seem to generate more accurate answers, especially in response to sensitive questions.

Pill counts and assays of plasma drug levels are not particularly useful measures of adherence. Pill counts usually underestimate nonadherence. Plasma drug levels generally only measure the last dose taken, especially if the drug has a short half-life. (A novel approach for monitoring long-term drug levels in hair was presented by a French group. There was a correlation between high indinavir levels in hair samples and complete virologic response -- abstract A-71).

Dr. Chesney discussed what is perhaps the most effective method of measuring adherence last. This technique uses an electronic monitoring system or MEMS cap. A computer chip is placed in the cap of the medication bottle. The time and date is recorded each time the bottle is opened and is counted as one dose taken. The major drawback with MEMS is that more than one dose may be taken out of the bottle at a time to be used throughout the day. This will only be recorded as one dose. The system is also expensive and requires specific software for interpretation. A useful feature of MEMS is that the computer chip generates a chart of each patient's adherence by day and time that can then be plotted against viral load fluctuations.

A study presented by Kathleen Melbourne, Pharm.D., of the University of Rhode Island, compared monitoring adherence with MEMS to patient self-report in 50 participants (abstract I-175). Both methods showed a high level of patient adherence (over 90%) with prescribed medication doses over a three-month period. However, patient self-report of adherence was consistently higher than that measured by MEMS. Twenty-four percent of study subjects overestimated by 10% how many doses they took. In addition, MEMS showed that many participants did not take their doses exactly on time -- over 40% were two or more hours off their reported dosing times. Dr. Melbourne concluded that self-report alone was not a reliable method of accurately measuring the study population's overall medication adherence.

Factors Affecting Adherence

Several studies at ICAAC attempted to unravel the factors that affect an individual's ability to consistently comply with rigid medication schedules. One such study, presented by David Paterson, M.D., of the Veterans Administration (VA) Medical Center in Pittsburgh, PA, assessed adherence to a protease inhibitor-containing regimen by MEMS caps in 45 participants (abstract I-172). At the start of the study, physicians were asked to predict the adherence of the participants. The physicians proved to be poor judges, predicting that 33% of those who turned out to be nonadherent would be adherent and that 36% of those with better than 95% adherence would be nonadherent.

In this group, active depression was associated with poor adherence. Another VA Medical Center in Columbia, South Carolina, reported similar findings in a six-month prospective study conducted in 48 participants (abstract I-215). Higher depression scores correlated with lower adherence and higher viral loads.

Interestingly, the total number of pills per day that a regimen requires was not associated with adherence. Median adherence was equivalent in the twice-a-day and three-times-a-day dosing groups in Dr. Paterson's study. An English study corroborated these results (abstract I-171). An association between difficult regimens and nonadherence may be lacking because more proactive, motivated patients are on the more complex regimens. It also may be a reflection that most of the regimens that have been available up until now are cumbersome and there is no real difference between them. A truly simple regimen involving just a few pills once or twice a day might do significantly better. (Literature on adherence from other diseases demonstrates that once-daily regimens are associated with 90% compliance, twice daily with 80% and three or more times daily with 65%.1)

The English study found that the impact of drug and alcohol use was great, with 29% of the cohort reporting missing one-fifth of their doses because they were "too stoned or drunk to take their medications."

The researchers also reported a connection between increasing time on treatment and decreasing adherence. Only half of participants who were 100% compliant for less then two years, remained so after two years. This May, a national survey conducted on 665 HIV-positive individuals documented that duration of "drug holidays" (periods when individuals stop taking their medications completely) increased from an average of 6.2 days for patients on therapy for two to 12 months, to an average of 14.4 days for those on therapy for more than 25 months.2 Difficulty of coping with food requirements increased from 13% for those on therapy for less than 12 months to 28% for those on therapy for more than 25 months.

Correlation between Adherence and Viral Load

Emerging data are confirming an association between adherent behavior and enhanced viral suppression. (Although it is unclear if patients with detectable viral loads are less adherent because they are discouraged by a poor virologic response and unmotivated to continue rigid dosing schedules or whether poor adherence on their part led to the detectable viral loads.) Successful adherence has been traditionally defined as 80% of doses taken. This may not be sufficient for protease inhibitor-containing regimes, as demonstrated by the results of Dr. Paterson's study (see table).

Median baseline CD4 count was 296 and 31% of the 45 study participants had viral loads below 400 copies/ml. Participants with greater than 95% adherence (defined as missing 1 dose in 20), had the best chance of achieving a viral load below 400 copies/ml at three months follow-up. The chances of virologic failure increased as more doses were missed. The differences between the categories were highly statistically significant.

Mark Shelton, of the State University of NY at Buffalo, reported a similar trend in a larger group of 295 patients who self-reported adherence during clinic visits (abstract I-170). Viral load was significantly lower among those with greater adherence with a direct relationship between number of missed doses and viral load. In addition, the proportion of patients with undetectable viral load was highest among those who reported no missed doses (45% of the group reported perfect adherence).

A study presented at the Retrovirus Conference in Chicago in February by Pablo Tebas, M.D., from Washington University in St. Louis, Missouri, determined that among 66 participants studied for at least six months, 75% were adherent to 85% or more of the drug regimen (abstract 149). Adherence to one drug correlated well with adherence to other drugs in the regimen and with the likelihood of a patient having a complete virologic response. In the study group, response to the regimen was similar across all stages of HIV disease. Patients with more advanced disease were as likely as patients with early infection to be adherent to their treatment regimens.

Finally, in a late breaker ICAAC presentation by Julio Montaner, M.D., of St. Pauls's Hospital, Vancouver, patient data from three clinical trials were analyzed to assess the effect of adherence on the duration of virologic suppression (abstract LB 10). Study subjects were classified as nonadherent if they took less than 75% of their medications prior to one or more visits. All participants had viral loads below 1,000 copies/ml at the start of the analysis period. After 48 weeks of treatment, over 80% of the adherent participants remained below 1,000 copies/ml, compared to about 50% of the nonadherent participants. This difference was statistically significant. In this population, adherent behavior increased the likelihood of maintaining virologic suppression.

Improving the Odds

It is difficult to determine average adherence to combination therapy in a "real world" setting because of the relatively inexact measuring techniques utilized and the lack of standardized classifications of levels of adherence. However, it is apparent that the more adherent a person is to his or her regimen, the more likely he or she is to experience durable and effective viral load suppression (90% adherence seems to be the emerging gold standard).

Poor adherence appears to be more closely associated with factors such as depression and active alcohol and drug use than with number of pills or frequency of doses. These conditions should be diagnosed and treated in conjunction with, or prior to, the initiation of antiretroviral therapy. Many of those who are expected to adhere to treatment do not, so physicians should negotiate a treatment plan to which their patient commits, provide specific education and monitor their patients' on-going experiences. Finally, interventions to assist in the long-term maintenance of adherence are crucial. Even a viral load below the limit of detection may not be enough incentive to keep taking pills ad infinitum. It would be worthwhile for HIV-positive individuals in all phases of treatment to have adherence counseling as a routine part of their medical care.


1. Sackett DL and Snow JS. Compliance in Health Care. Baltimore: Johns Hopkins University Press, 1979, page 18.

2. Gallant J and Block D. Journal of the International Association of Physicians in AIDS Care. May 1998; 4(5):32-5.

Treatment Failure Increases as Adherence Decreases in a Cohort of Patients Taking Protease Inhibitors
(ICAAC abstract I-172)
Adherence Category
Number of Doses Missed Per Week
(of a Thrice-daily Regimen)
Percent Above 400 copies/ml at Most Recent Viral Load Test
above 95%
90% to 95%
80% to 89.9%
three to four
70% to 79.9%
five to six
less than 70%
greater than six
Back to GMHC Treatment Issues October 1998 contents page.