Abstract
Objective:
This paper examines the scientific bases for the recommendations on readiness, trust, and adherence that were contained in 5 HIV treatment guidelines.
Methods:
The authors reviewed the treatment guidelines of 5 internationally recognized expert panels, and 5 review articles on readiness, trust, and adherence to determine the evidence for the guidelines' recommendations. These review articles were assessed to determine whether they contained: 1) a definition of the concept, 2) a measure that predicted outcomes, or 3) evidence that a replicable intervention altered outcomes.
Results:
There is no good definition, measure, or evidence that readiness predicts adherence. There were no interventions that improved readiness in a clinically useful way. There was no good definition or measure of trust, nor was there evidence that patient trust predicts adherence or that interventions improve trust. It is not clear whether trust is a cause or an effect of clinical outcomes. There is agreement that adherence is defined as taking medications as prescribed. Six measures of adherence were found, and several studies showed that they correlated poorly when applied to the same sample of patients. Adherence interventions have had some success but are poorly described and idiosyncratic such that it is difficult to replicate them in other settings.
Conclusion:
Research on readiness and trust are unlikely to yield useful results. Systematic, long term research on clinician-based adherence interventions must be done. A life-long process such as adherence to antiretroviral treatment is unlikely to be understood without having a long-term study of a cohort of patients taking these medications.
Introduction
Five expert panels have prepared guidelines on treating HIV-infected patients. Four of these have based their recommendations on the strength of scientific evidence. The US Department of Health and Human Services (DHHS) guidelines rank their recommendations as strong, moderate, and optional and give 3 levels of scientific support for the recommendations. 1 The International AIDS Society–USA (IAS-USA), 2 the British HIV Association 3 (BHIVA), and the World Health Organization 4 (WHO) use tiered levels of recommendations and ratings for the strength of evidence. The fifth organization that issues guidelines is the European AIDS Clinical Society (EACS). 5
All these guidelines also provide evidence-based principles for treatment of HIV-infected patients who are pregnant or who have HIV-associated conditions such as nephropathy, HBV coinfection, hepatitis C coinfection, tuberculosis, cardiovascular disease, etc. They provide evidence-based, detailed suggestions on when to treat or give prophylaxis for a variety of acute opportunistic infections. The DHHS, IAS-USA, and WHO guidelines support their recommendation with strength of evidence metrics. The BHIVA guidelines for treating associated conditions are heavily referenced. The EACS guidelines also make use of the literature in supporting their recommendations. All these treatment guidelines are scholarly and current. They are developed by the leading clinicians and researcher in their areas of practice. Clinicians who follow these guidelines are assured that they are making treatment decisions that are based on the best scientific evidence.
These 5 panels acknowledge the crucial nature of patient attitudes and behaviors in the successful treatment of HIV and make recommendations for considering them in treating patients. However, unlike the treatment recommendations, there are no strength-of-evidence ratings for attitudes and behaviors. In addition, none of the guidelines contain evidence-based mechanisms for the clinician to implement the behavioral recommendations. The guidelines, to varying degrees, specify behavioral and attitudinal recommendations for 3 important concepts—readiness, trust, and adherence. This article examines recently published reviews for evidence to support the inclusion of the recommendations in science-based, clinical guidelines.
Methods
First, the recommendations of the 5 expert groups were reviewed to ascertain the importance that these panels assigned to the 3 concepts. Then, reviews of these concepts published in 2010 were examined. 6 –8 The abstracts of all the publications that were referenced in these 3 reviews were examined. Any articles that appeared to meet any one of the 3 inclusion criteria were then examined. The criteria for inclusion were that the article had to contain either (1) a clear definition of the concept, (2) a measure of the concept that had been shown to predict clinical outcomes, and/or (3) research-based evidence that a replicable intervention would change treatment outcomes. All these articles were then read to assess the strength of the evidence and to determine whether the evidence justifies inclusion in the guidelines. The primary criterion for whether the article provided evidence that would support clinical guidelines was that the study contained a sensitive and specific measurement of patient behaviors or attitudes that would guide a clinician’s treatment decisions.
Results on Readiness
Guidelines
The DHHS guidelines recommend that clinicians assess readiness before writing the first prescription for antiretroviral therapy (ART). These guidelines do not specify how to assess readiness. This absence of direction can be seen in table 12 of the guidelines where the column that should provide examples of establishing readiness to start ART is blank.1 The recommendation section of IAS-USA in its abstract begins with “Patient readiness for treatment is a key consideration when deciding when to initiate ART.”2 However, there is no additional discussion or guidance for assessing readiness anywhere else in these guidelines. The BHIVA guidelines state that patients who have CD4 counts between 200 and 350 cells/mm3 should only be started if they are “ready.” These guidelines also fail to provide any evidence-based means of determining readiness. The EACS guidelines recommend examining potential barriers to adherence and provide a series of questions to ask to determine patient readiness. 5 They provide no evidence that the questions can sort the “ready” from the “unready.” The EACS guidelines also suggest measuring readiness on a visual analog scale. The guidelines do not suggest a cutoff area on the scale where drugs should be prescribed and/or withheld. Also, they do not provide any evidence that the visual analog scale measuring readiness will predict future adherence. The WHO has a document (not part of its guidelines) that lists questions for patients that purport to assist the clinician to assess readiness. However, the document provides no guidance in how to use the patient’s answers so as to determine readiness nor is there evidence of the questions’ utility in predicting adherence. 9
Evidence
A 2010 review of 25 HIV/readiness studies found 6 different definitions and measurements of readiness. 6 These included (1) 2 questions to determine where the patients were in the first steps of the transtheoretical model; (2) a 5-item scale based on the wellness motivation theory; (3) a 30-item readiness scale; (4) knowledge of ART and its efficacy; (5) a visual analog scale that measured patients’ perception of their readiness; and (6) a Likert-type scale that asked patients how ready they were to start therapy. Nine of the 25 studies examined the relationship of a measure of readiness with adherence to ART. Two measures (a 2-item Stage of Change Scale and a Medication Readiness Scale) demonstrated a modest relationship with adherence. Neither would be applicable in the clinical setting because of poor sensitivity and specificity, which would lead to misclassification of patients with regard to their receiving or not receiving ART. Also, one of the studies only followed patients for 1 month so its long-term predictive ability is unknown. 6 Contrary to the EACS recommendation, the use of visual analog scale to measure readiness did not predict adherence in 1 study. 10
Conclusion
Emphasizing the need for readiness might be important but is insufficient for the clinician who has to decide whether a particular patient should or should not be started on ART. The clinician needs a tool to ascertain the patient’s state of readiness that will predict future adherence. The assessment has to be highly sensitive and specific in categorizing the ready from the unready. Failing to start ART in a patient who is wrongly deemed to be unready can have severe consequences for that patient. Starting a patient who was wrongly deemed ready is likely to result in nonadherence and reduce future treatment options. In addition, clinicians also want to know how to improve readiness if it is lacking.
The authors of the review on readiness also concluded that the conceptualization of readiness as a steady state that will not change is flawed. Patients who are on life-long therapy must be ready daily to take their pills, be ready monthly to renew their prescriptions, and be ready quarterly to have blood drawn for laboratory tests and to attend clinic. It is unlikely that any instrument will capture all these aspects of readiness and do it for the foreseeable future. As the review points out “Believing that a readiness measure that makes a prediction for a person starting HAART in 2010 will be able to predict the patient’s adherence behavior in 2012 or 2020 seems like a demand for a crystal ball rather than a psychometric instrument.”6
In summary, at the time of this review, no clinically useful measure of readiness had been published and there was no clinically useful evidence that readiness, however, measured, is linked to future adherence and/or patient outcomes. As a result, interventions to improve readiness have not been demonstrated.
Results on Trust
Guidelines
The DHHS guidelines advocate establishing a trusting relationship with patients as a means of improving and maintaining adherence. However, these guidelines do not provide any means of measuring trust or ways to establish a trusting relationship. This is evidenced by the fact that, in the guideline’s table 12, there is a blank space where there should be examples of how to assess and improve trust.1 The EACS guidelines discuss establishing trust to determine readiness to start therapy. Both of these guidelines fail to provide evidence that trust results in a better patient outcomes. Nor do they provide any method for developing trust should it be lacking.
Evidence
Multiple studies have employed various scales to measure patient trust in physicians, 11 –16 pharmacists, 17 and nurses. 18 There are also scales to measure distrust in the health care system. 19,20 Most of these studies examined what factors predicted trust but did not investigate the scale’s ability to predict patient outcomes. A review of studies of trust found no agreed upon definition or measure of trust. The review did find modest evidence that patients who trusted their physicians and/or the health care system (however, measured) were somewhat more likely to access health care, remain in care, and report better health. 20 –23
There is one study that suggests that low levels of trust may impact outcomes of HIV-infected persons. Patients who were being seen at a clinic were periodically asked whether they trusted their physicians. The researchers found that patients who lost trust during the course of their treatment were more likely to require changes in their medication regimens and to have adverse events related to their HIV or their treatment. However, it is not clear whether the loss of trust is the cause of or the result of these regimen changes or adverse events. Patients may lose trust because their regimen did not work, had side effects, and so on. However, it is noteworthy that some patients in this study who lost trust in their providers regained it later, suggesting that trust, like readiness, is not a steady state of mind. 24 Another study examined whether patients with higher levels of trust in physicians or the health care system initiated HIV care with higher CD4 counts and would be more likely to remain in care. Lack of trust in either physicians or the health care system was not associated with CD4 counts at entry to care. A higher level of trust in physicians, however, was associated with remaining in care. 25,26
Conclusion
While it may be important for the guidelines to recommend the establishment of a trusting relationship with patients, the clinician needs a tool for deciding whether a patient’s failure to follow advice or to take medications as prescribed is the result of a lack of trust or because of other reasons. Clinicians also need a means of developing a trust relationship should it be determined that it is lacking. At the time the trust review 7 was published, there was no agreed upon definition or measure of trust. There was no evidence that trust is linked to following clinical advice or adhering to antiretroviral treatments (ART) prescriptions. There is only modest evidence that patient trust is associated with patient outcomes in HIV-infected patients although the direction of the causal relationship is not clear. This review did not find evidence that any intervention changed levels of trust nor altered any patient outcomes. 7 This finding was consistent with a Cochrane review from 2006 which concluded that “Overall there remains insufficient evidence to conclude that any intervention may increase or decrease trust in doctors.” 27 So, there is no measure of trust that is sensitive or specific enough to provide assistance to a clinician who is trying to assess a patient.
Results on Adherence Measurement
Guidelines
All the guidelines encourage clinicians to assess adherence to ART and to intervene when adherence is poor. The BHVA guidelines discuss using self-report as a means of assessing adherence but also report that it is likely to overstate true adherence and do point out that it is not well correlated with clinical outcomes. There is also a comment that multiple questions might obtain a better measure of adherence. However, the reference on which that statement is based does not support the guidelines’ comment. The EACS guidelines suggest asking 2 questions about the frequency and pattern of missing doses in the past 28 days to assess adherence. It is not clear whether an improper answer to one or both questions identifies a patient as nonadherent. The questions are based on a Swiss study where answers to either of the questions were associated with a higher likelihood of viral rebound in HIV-infected patients. Their combined effect was not addressed in the article. While there was evidence of a relationship between the questions and a biological outcome, neither the sensitivity nor the specificity of the questions with regard to adherence or viral load change was published by the authors. 28
The EACS guidelines also suggest using therapeutic drug monitoring as a means of measuring adherence. However, its accuracy in measuring adherence is restricted to a short period before blood is drawn. The efficacy of a blood test that was conducted at one point in time to predict future adherence has not been established. Having a good measure of adherence is very important to clinicians who are often in the position of having to decide whether failure to suppress HIV is the result of drug failure or poor adherence. This can be particularly important in resource-limited settings where the cost of second-line regimens is much higher than initial regimens. Once again the practicing clinician needs a sensitive and specific measure of adherence to improve his or her ability to assess and counsel their patients.
Evidence
There is agreement on the definition of adherence to ART. It is taking one’s medications as prescribed. However, the literature on how to measure adherence is complex and confusing. In 2010, Simoni et al updated earlier reviews by this group on adherence.
8,29,30
In these reviews, they found that adherence had been measured by self-report, pill counts, electronic monitoring devices, insurance claim records, provider estimates, and pharmacy refill records. This would not be a problem if these methods are measuring the same thing. However, this does not seem to be the case. Holzemer et al administered 5 measures of adherence to the same patients—the AIDS Clinical Trial Group–revised reasons for missing medications, self-report of medication nonadherence, pill count, electronic recording devices, and pharmacy refills. The average correlation between the 5 adherence measures was
According to the Simoni et al reviews, 8,29,30 the most commonly used measure was self-report. A problem with the use of self-report as a measure is that there is no agreed upon interval over which a patient is to report his or her adherence. A review of 77 self-report studies found that the time intervals for patient recall in the studies were 1, 2, 3, 4, 7, 14, 28, and 30 days as well as 1, 3, and 6 months. The patients’ adherence reports were sometimes treated as dichotomous variables (usually <95% and ≥ 95%) and sometimes as continuous variables. 29 It is difficult to believe that these various self-report-based studies are measuring the same thing.
Another unexplored area of adherence is the pattern of nonadherence. For example, if one were to suggest that a patient needed to achieve 95% adherence on a once-per-day regimen, it would mean that the patient could not miss more than 18 doses in a year. Those episodes of nonadherence could be 1 or 2 per month, 18 straight days, every other day for 36 days, and so on. And, each of the possible patterns could have an impact on viral loads depending on the nature of a patient’s virus, the medication regimen that the patient is taking, and the particular resistance profile of that patient’s virus. 36 A further complication is that patients may adhere to some drugs in their regimens but not others. In 2 studies, 15% and 29% of patients took some but not all their medications at different times. These patterns were shown to impact virologic failure. 37,38
The adherence literature has also been bedeviled by the uncertainty of what level of adherence is necessary to achieve viral suppression. The guidelines are not specific about the level of adherence that is necessary. However, the optimum level of adherence is often stated as taking >95% of one’s prescribed ART drugs. This level is based on a small number of studies of a small number of patients. The most referenced of these studies (>600 Medline citations) is the Paterson et al study published in 2000. 39 This study began with 99 patients but only 81 provided data that could be evaluated. Patients were followed up for 3 to 15 months (median 6 months). Their adherence was measured by electronic recording devices on only one of the drugs in the regimen. The study showed that patients who had greater than 95% adherence were more likely to have undetectable viral loads. Interestingly, 22% of the patients who had >95% adherence failed to reach undetectable levels (<400 copies), while 20% of those who had <80% adherence were found to be undetectable. The phenomenon of some very good adherers not achieving undetectability while some poor adherers were undetectable was also found in a study of patients under treatment in 1998 to 1999. This study found that 20% to 25% of low adherers had undetectable viral loads and 20% to 25% of those who had >95% adherence were detectable. 40 (Viral suppression levels varied according to whether the patients were taking a protease inhibitor or a nonnucleoside reverse transcriptase inhibitor [NNRTI].) The current relevance of either of these studies is questionable, given that the outcomes were established using medications that, in the main, clinicians stopped using several years ago.
Another highly referenced adherence study (94 citations) examined 76 patients in 2000 and found that low adherence, measured by unannounced pill counts, was associated with progression to AIDS. 41 A more recent study of 903 patients in Vancouver who were on therapy during 2000 to 2004 demonstrated that patients with <95% adherence, measured by prescription renewals, were more than 3 times more likely to die. 42 A Swiss study of patients being treated between 2003 and 2007 also showed that patients who had <95% adherence to their regimens, as measured by self-report of missed doses over the previous 28 days, were more than 3 time more likely to experience viral rebound. 28
Conclusions
There are a wide variety of ways to measure adherence, but none has been shown to be a sensitive and/or specific predictor of an individual patient’s biologic outcomes. There has been little research on the patterns of nonadherence as opposed to the rate of nonadherence and, as a result, little is known about the impact of various patterns on viral suppression. While there was once a general agreement that >95% adherence is necessary to achieve durable viral suppression, there is little evidence that supports that level of adherence in patients taking modern therapies. There is even contrary evidence that more current therapies are more forgiving of less than perfect adherence. A review by Gardner et al pointed out that newer drugs have longer half-lives so that bouts of poor adherence can be tolerated because drug levels are less likely to fall to subtherapeutic levels. 43 As Bangsberg et al showed with data from 1998 to 1999, a regimen with 2 nucleoside reverse transcriptase inhibitors (NRTIs), with shorter half-lives than currently recommended regimens together with an NNRTI, led to viral suppression even when patients were less than 95% adherent. 40 The 2011 DHHS guidelines questioned the use of 95% as a hard-and-fast rule of adherence.1 An unfortunate side effect of the 95% adherence rule is that many studies were conducted with adherence being treated as a dichotomous variable. So patients with 92% adherence were lumped in with those who had 70% adherence and with those who never took a pill.
While it is clear that taking a high percentage of one’s prescribed regimen is better than taking fewer doses, the level of necessary adherence is unclear. There is no current relevance of studies from 1998 to 2000, given that the DHHSs preferred therapies in that era included indinavir (IDV), ritonavir ([RTV]; as a single protease inhibitor), nelfinavir (NFV), saquinavir ([SQV]; unboosted), and zalcitabine (ddC). 44 Neither the Vancouver nor the Swiss studies were restricted to patients taking the long-acting NRTIs that are the backbone of all the current regimens that are highly rated by the 5 guidelines. There are no studies of the level of adherence necessary for viral suppression in patients using integrase inhibitors or entry inhibitors. So the applicability of requiring extremely high levels of adherence to modern therapy is questionable. It should also be noted that all these studies show that a population of good adherers has better outcomes than a population of poor adherers.
However, applying these population-based findings to the individual patient is not of much assistance to the clinician who must decide whether a poor clinical response in a specific patient is a medication problem or an adherence problem. While there is evidence that self-reports of poor adherence are associated with poor clinical outcomes, these population-based data are of little utility when applied to the individual patient. If a clinician asks a patient about adherence over the kind of short time period where recall is likely good (eg, a few days), some patients with poor adherence will not be identified. As the length of time for recall increases, most nonadherers will be identified. However, as the time interval increases, recall becomes less accurate. Also, as the number of days increases, an increasing number of good adherers will have missed a dose and be identified as nonadherent. So while a self-report of poor adherence should alert the clinician, he or she still must decide for an individual patient whether to change regimens, order resistance testing, work with patients on adherence issues, refer to an adherence counselor, and so on. This is further complicated by the evidence that moderate adherence is more likely to lead to resistance than either excellent or very bad adherence. 45 Clinicians must also consider that there is evidence that very good adherers may have failing regimens and poor adherers continue to suppress their virus. So truthful reports of good adherence may not always lead to desired clinical outcomes, and truthful reports of poor adherence do not always lead to loss of virologic control.
Results on Adherence Interventions
Guidelines
If adherence could be measured in some meaningful way and it was low, then some kind of an intervention would be needed to improve it. This is the stated or implied goal in the guidelines for assessing adherence. The DHHS guidelines provide a table for clinicians that will assist their patients in adhering. However, the guidelines do not offer any evidence-based suggestions on how to intervene if a patient is not adhering. The EACS guidelines suggest that if a patient is not adhering to his or her current regimen, the clinician should change to a regimen that the patient will find easier to tolerate. No guidance is given on how to determine that the regimen is the cause of nonadherence nor are there recommendations on how to determine whether another regimen will prove superior. The IAS-USA guidelines also suggest that regimen switching is recommended in the event that the patients experience toxicities that may cause them to stop adhering. These guidelines do not provide any other suggestions for intervening with the nonadherent patient. The BHIVA guidelines contain an extensive review (with references) of the literature on adherence interventions. However, the studies are population based and do not give a clinician any guidance about which intervention to use with any given patient who is having difficulty adhering. The WHO guidelines suggest that changing from stavudine (d4T) to zidovudine (ZDV) may improve adherence by reducing toxicity. This publication also mentions intervening when there is poor adherence but does not provide any guidance on how to carry out an intervention.
Evidence
There have been several reviews of the impact of adherence interventions. 8,30,46,47 The reviews revealed an extraordinary diversity of methods, follow-up periods, measure of adherence, characteristics of the individuals conducting the interventions, types of patients, intensity of the intervention, and sample sizes. Follow-up periods ranged from the end of the intervention to 15 months. Types of interventions included cognitive behavioral therapy, motivational interviewing, medication management strategies, and group and individual counseling sessions. Individuals delivering the interventions included lay individuals, health advocates, social workers, psychologists, nurses, pharmacists, and physicians. The interventions were directed toward ART-naive patients, patients who had previous difficulty adhering, women, minorities, and ART-experienced patients. The interventions ranged from single sessions to over 50 sessions and were as short as an hour and as long as 15 months. Individual sessions lasted from 20 minutes to 2.5 hours while sample sizes ranged from as few as 10 to over 7000. Most studies had fewer than 100 patients. Adherence was measured by electronic monitoring, pill counts, medication diaries, patient self-report, provider report, viral loads, and clinic or pharmacy records.
The effectiveness of the adherence interventions has been variable. The 2010 review by Simoni et al 8 reviewed 13 adherence interventions that were conducted between 2007 and 2009. The personnel doing the intervention included nurses, social workers, peers, case managers, pharmacists, therapists, psychologists, and counselors. Sample sizes ranged from 20 to 7018. These studies were conducted in the United States, Canada, Mozambique, Brazil, and China. The studies described a wide variety of interventions, including motivational interviewing, education by a pharmacist when patients picked up their medications, determining barriers to adherence and developing strategies to overcome them, counseling followed by phone calls, adding medication adherence education to other self-improvement activities such as safer sex, substance abuse, and so on and a variety of other methods. The number of session ranged from 6 to 26 and they varied in duration from 20 minutes to 2 hours. 8 In spite of this heterogeneity, Simoni et al found evidence that 11 of these studies showed improvements in adherence following the intervention.8 This might lead one to conclude that almost any intervention would work. However, Rueda et al summarized 43 interventions to improve adherence. They found that patient populations ranged in size from 22 to 966. In only 53% of the studies did the intervention have a positive effect on adherence. Twenty of the studies that showed a positive impact on adherence also measured virologic or immunologic outcomes. Only 7 (35%) improved the biological outcome. 48
Similarly, the Centers for Disease Control and Prevention (CDC) has developed a Web site for what it terms evidence-based interventions to improve medication adherence in HIV-infected persons. 49 The CDC compendium highlights 8 studies that they classify as “good” interventions. None of them met the CDCs criteria of best evidence because of flaws in the study design or conduct.
Five of the studies were based on education/counseling of patients. One study involved 24 home visits by a nurse and a community worker over a 12-month period that resulted in a modest improvement in adherence in the experimental group over the control group, but it did not have statistically significant intervention effect on HIV-RNA level or CD4 count. A second study used two 2- to 3-hour sessions just prior to beginning medication; three 1.5-hour sessions (range, 45 minutes to 2 hours) at weeks 2, 4, and 8 following medication initiation, with 5 support phone calls between sessions (weeks 1, 6, and 10 and months 4 and 5 after), and a 1.5-hour booster session at 6 months. The content of any of these sessions was not specified. The study was assessed by the percentage of patients who achieved 90% adherence by an electronic recording device. In the experimental group, 46% achieved this level of adherence versus 28% in the control group. The third study involved a 6-session individual/dyadic and group-level intervention. The social support/problem-solving intervention delivered two 2- to 3-hour sessions just before the initiation of therapy and three 1.5-hour sessions in the first 2 months after starting ART. A follow-up session was held 6 months afterward. After 3 months, the intervention group was more likely to achieve >90% adherence, but > 90% adherence declined “steadily and significantly over time.”
The fourth study in the CDC compendium utilized posters in the clinic, brochures, flyers, and the presentation of adherence messages by the medical provider during the medical examination. At each clinic visit, the provider delivered a brief counseling session (3-5 minutes) which discussed pill scheduling and adherence goals. At 11 to 18 months post-initiation of the intervention, the percentage of participants reporting
Two other studies from the CDC compendium involved directly observed therapy (DOT). One of these involved DOT delivered from a mobile van for one of an individual’s doses Monday through Friday with pills provided for other daily doses and the weekends. The DOT group had a 16% higher rate of meeting the success criteria of having either an undetectable viral load or having a >1 log drop in viral load than the control group. Even with DOT, 30% of the experimental group failed to meet the criteria after 6 months. The other DOT study was performed with observations being done at a methadone maintenance clinic where one of the daily ART doses was administered with the methadone dosing. Take-home doses were provided for second daily doses and for days when the clinic was closed. Those in the DOT arm of the study had much better rates of undetectable viral loads than the control group at the end of 6 months (71% versus 46%).
The eighth study that CDC designated as a “good” evidence-based medication adherence intervention involved the use of a pager to remind patients to take their medications and to provide educational messages to patients. Patients receiving the pagers were more likely to suppress their viral load to <1000 copies/mL than the control group and showed a higher rate of attaining a CD4 count of >350 cells/mm3.
Conclusions
It is difficult to evaluate the adherence intervention literature. There is no agreed upon measure of adherence and, as has been previously discussed, the measures that are used have not correlated with one another in several studies. Therefore, it is difficult to determine and compare the effectiveness of these interventions. Seemingly effective interventions might have been deemed unsuccessful if some other measure of adherence had been used. Interventions judged to be failures may have had poor results because of the adherence measure and not because of the intervention. The CDC compendium of what they term “good” studies showed that participants in the intervention groups typically showed rates of adherence and/or viral suppression that were 15% to 25% higher than the control groups. However, all but one of the education/counseling interventions were labor intensive and would have been prohibitively expensive for a typical HIV clinic to use. In addition, all the studies were population based and would provide little guidance to the clinician on how to deal with a specific patient who was not adhering. It is interesting that CDC sees these 2 DOT studies as evidence-based interventions which are apparently worthy of emulation, while the International Association of Physicians in AIDS Care’s recent guidelines strongly recommend against the use of DOT to improve adherence. 50
There is also a potential for publication bias. Unsuccessful interventions may not be submitted to journals or may be unfavorably reviewed. Additionally, there is significant potential for bias in subject selection. A review article reported that patients are more likely to attend intervention sessions that fit into their schedules and to attend and remain in interventions where it is possible to develop a strong relationship with the intervener. Antiretroviral-naive patients are more likely to stay with an intervention than the ART experienced, and patients are more likely to remain in programs conducted separately from regular clinic operations. 51 The other potential explanation for the apparent effectiveness of interventions may be a medical “Hawthorne effect.” Patients are responding to the attention and/or oversight that they receive from participating in the studies and are not benefiting from the actual intervention.
There is some evidence that some interventions to improve adherence worked in some people but there is absolutely no evidence that an intervention that was successful in one setting could be transferred to another place. As a matter of fact, the published interventions are usually so poorly described and/or idiosyncratic that it would be impossible for them to be replicated in another location.
So while the clinician is expected to assess adherence and to intervene when it is lacking, there is no agreed upon measure of adherence and no guidance on what intervention will work with any individual patient. It is also important to note that all these population-based interventions had short-term follow-up. All were less than 2 years, with a majority being 6 months or less. This is a tentative basis on which to base clinical decisions that are supposed to maximize lifetime benefits to HIV-infected persons.
Overall Adherence Summary
The recent guidelines that were issued by the International Association of Physicians in AIDS Care review much of this literature. 50 This panel made 19 recommendations with regard to medication adherence. These recommendations were graded on a scale of I, II, and III, with level I indicating a recommendation with strong research support based on a randomized clinical trial (RCT). Level II indicated that there was research support based on an RCT but that it had limitations. The third level was for recommendations that were based on an RCT with critical limitations or an observational study without important limitations. The group also used an A, B, and C rating scheme for the strength of the recommendation. An A level strength meant that almost all patients should receive the action. B level recommendations indicated that most patients should receive the recommendation but other choices were appropriate. The C level meant that the recommendation could be considered for some patients but not to be followed routinely. Of the 19 recommendations, 8 were in level III for the quality of the evidence and 4 of these were category C for strength of recommendation. Seven of the recommendations were category II for quality of the research. Only 4 of the 19 recommendations reached level I. Two of the level I recommendations were of actions that should not be done—directly administered therapy and routine use of electronic monitors. These, of course, will not assist a clinician who is trying to improve an individual patient’s adherence. Another level I recommendation with a B level of strength advocates the use of reminder devices of various kinds including those that allow interactions between the reminder and those persons being reminded. The fourth level I recommendation had a strength of recommendation level A. It states that “Education and Counseling using specific adherence-related tools is recommended.” However, the recommendation provides no details on the content of the counseling and education sessions, length of sessions, frequency of sessions, and/or timing of the sessions in relationship to the initiation of therapy. The underlying research for this recommendation is 7 studies, 6 of which examined biologic outcomes and only 3 showed improvement in the biologic measures.
Discussion
The authors of this article set out to review whether there was sufficient scientifically sound research to provide strength of evidence ratings for readiness, trust, and adherence in the guidelines. They chose to do so by reviewing recent reviews and looking at the 5 major guidelines for HIV care under the assumption that it would have been highly unlikely that all of these sources would have failed to notice a sensitive and specific measure of these concepts or missed a highly successful intervention. Unfortunately, based on the reviews published in 2010 and the most recent guidelines, these highly important behavioral and attitudinal concepts are not measurable in a way that can be used by clinicians. Additionally, if they could be measured in some meaningful way, there is little or no guidance on how one could alter undesirable measurements.
This raises the question of, “Why are these 3, immeasurable, nondata-based concepts for which there are no useful interventions included in guidelines that are so heavily based on scientific evidence?” The primary reason is that they are important concepts. It is logical to assume that they are necessary adjuncts to successful treatment of HIV infection. So their inclusion makes sense. The fact that there is so little evidence to support the implementation of the concepts is not the fault of the guideline writers but rather is the lack of good science that underlies the concepts. The guidelines are based on the best available science, which is inadequate with respect to readiness, trust, and adherence.
The really important question is “Why are 3 clinically crucial concepts so inadequately researched 16 years into the era of ART?” There are at least 3 potential explanations for this. First, behavioral scientists have been more concerned with finding statistically significant relationships than with the clinical utility of their findings. Unfortunately, the clinicians who have done research in this area have also fallen into the trap of preferring statistical significance over clinical significance. The clinician’s world consists of patients who admit that he or she has “not taken my pills as well as I should have” and whose viral load has gone from undetectable to 2300 copies/mL. There is nothing in the reams of adherence intervention research that would assist a clinician in helping this particular patient return to undetectable. Population-based data showing modest, but statistically significant, improvements in adherence are of little utility to the practicing clinician who is faced with a single, apparently nonadhering patient.
Second, both behaviorists and clinicians have failed to utilize the expertise of basic scientists. The authors were unable to find any adherence studies that examined the nature of the patients’ virus or the patients’ immune system beyond merely counting CD4 and quantifying HIV. The immune system is made up of a wide variety of interacting cells, each of which has an important role in suppressing disease-causing organisms. The HIV utilizes multiple CD4 receptors and it is unknown how they may impact adherence outcomes. To assume that poor adherence only affects one generic type of immune cells seems to be incomplete thinking. To only examine the count of HIV and not its resistance pattern, fitness or clade seems to be equally inadequate.
The third reason that there is so little clinically useful research in these areas is that research tends to follow incentives and there have been few rewards to conduct the long-term research that is necessary to explore these concepts. In the United States, there is no Federal agency that is designated to coordinate and stimulate research on patient behaviors. Some behavioral research is funded in each of the institutes of the National Institutes of Health. However, it is focused toward the aims of each individual institute. For example, the National Institute of Drug Abuse sponsors a great deal of behavioral research, but it is for a specific population—users of addicting substances. The CDC funds behavioral research and behavioral interventions, but their portfolio is directed toward disease prevention. The Food and Drug Administration is highly concerned with the safety and effectiveness of medication but not with the behavioral factors that lead to the failure of medications. Nowhere in Europe or North America is there an agency or organization that funds or coordinates research efforts on examining the patient behaviors that lead to medical failures.
Directions for Future Research
The authors would like to suggest some directions for future research. First, some investigations need not be done. As was stated in the conclusion of the readiness section of this article, it seems impossible to devise an instrument that measures readiness that predicts future adherence for the rest of a patient’s life for even a small part of the future of a patient. In the absence of such an impossible-to-construct instrument, it seems unlikely that evaluable interventions can be created to improve readiness. Similarly, it seems unlikely that trust-based research will ever yield any useful output. The basis of doing trust research is to determine which patients do not trust their clinician or the system from which they are receiving care. Once that is determined, presumably an intervention would occur to increase trust. However, this relies on the patient who does not trust the provider or the health care system to reveal to the untrustworthy that they are not trusted. It does not seem likely that an untrusting person would reveal to those who he or she does not trust that they are untrustworthy. Further research on readiness or trust is not recommended. It is also recommended that guideline writers cease including these undefined, not-possible-to-measure concepts for which there are no feasible interventions in their recommended practices.
No further RCTs of adherence interventions should be conducted, which do not involve patients and their interactions with their first-line clinicians. Large-scale RCTs of various interventions have, in general, suggested interventions that are impractical to implement in typical clinical settings which are unlikely to have the expertise or resources to implement them. More importantly, they have had modest effects on adherence even using very intensive and expensive methods. The effect of these interventions on clinical markers is even more modest.
Further research is not needed to find a measure of adherence that predicts viral loads and/or CD4 counts. If these biological measures are the gold standard, they should be used to evaluate the outcome of adherence interventions.
The authors have 2 recommendations with regard to changing the direction of adherence research. The first is that research should start with the clinician–patient interaction. Every HIV clinician has to deal with nonadhering patients on a daily basis. These clinicians certainly perform some interventions in order to improve their patients’ adherence. Some of these are successful and some of them fail. Research that examined these interactions and tried to determine which approaches worked in which patients would be more likely to provide useful interventions than one more randomized, impractical, clinical trial. There is evidence to this effect from the study of Milam et al which showed the clinicians who provided 3- to 5-minute counseling sessions for their patients at regular clinic visits had outcomes similar to the RCTs that provided many hours of counseling and education sessions. 52
Our second recommendation is that there is a need to create a prospective adherence cohort study. Adherence is a long-term, multiyear or even multi-decade problem. Such a problem can only be understood by long-term studies of patients. Following many patients over the course of their treatment will provide useful insight into the economic, social, psychological, health care delivery system, and so on problems that impact adherence. One more 6- or 12-month intervention will not solve a multi-decade problem. One can only regret that a cohort study was not begun in 1996 when ART began. Our understanding of nonadherence over the long term would certainly be better understood.
Final Thoughts
The HIV adherence researchers have led us to better understand the complexities of patient adherence. Until these researchers attacked the problem of adherence in HIV-infected persons, there were few studies on patients’ adherence to any medical regimen. So not much is known about why patients do not engage in prescribed behaviors in all areas of medical care. The HIV epidemic and its need for lifelong therapy has only made us realize that there is such a major void in our understanding of how to care for patients.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This publication was made possible with help from the Baylor-UT Houston Center for AIDS Research (CFAR), a National Institutes of Health funded program (AI036211).