National Academies Press: OpenBook

Methodological Challenges in Biomedical HIV Prevention Trials (2008)

Chapter: 5 Design Considerations: Adherence

« Previous: 4 Design Considerations: Pregnancy
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 119
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 120
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 121
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 122
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 123
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 124
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 125
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 126
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 127
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 128
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 129
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 130
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 131
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 132
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 133
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 134
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 135
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 136
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 137
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 138
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 139
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 140
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 141
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 142
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 143
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 144
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 145
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 146
Suggested Citation:"5 Design Considerations: Adherence." Institute of Medicine. 2008. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: The National Academies Press. doi: 10.17226/12056.
×
Page 147

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

5 Design Considerations: Adherence F or a new biological HIV prevention product to be effective, users must adhere to the prescribed regimen. During a clinical trial, imper- fect adherence reduces the product’s effectiveness, and also makes it difficult for investigators to assess efficacy. If a trial shows that a product provides an overall benefit, relating the level of protection to the level of adherence is valuable for interpreting the results. Understanding who was protected during the trial and under what circumstances also has important implications for predicting how effective the product will be in various real-world settings. To properly interpret the results of a clinical trial that failed to show a protective effect, investigators need to distinguish the extent to which the product was not biologically efficacious, participants did not use it as directed, or they engaged in more risky behavior because they thought the product was protecting them. Regardless of whether a trial demonstrates an effect, understanding when and why participants did not adhere to the product regimen can provide valuable insights into the design and delivery of future HIV prevention interventions. While interventions given once or a very few times, such as vaccines or circumcision, usually do not entail adherence challenges, other exist- ing and new biomedical HIV interventions such as condoms, PrEP, and microbicides require longer-term administration. Adherence is more than “simply remembering medications, but rather, a complex issue involving social, cultural, economic, and personal factors” (Chesney, 2006). In the antiretroviral treatment (ART) field, there has been insufficient progress in understanding the correlates of adherence and strategies to increase 119

120 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS adherence (Sankar et al., 2006). Adherence to biomedical HIV prevention interventions is even less well understood. This underscores the importance of understanding the context and factors that affect individuals’ adherence and risk behavior. Researchers should not assume that definitions or models of adherence (or risk reduction) developed in one sociocultural context are equally as valid in another (Ware et al., 2006). They should be evaluated and adapted to new settings as appropriate (see Ware et al., 2006, for one such adaptation model). Studies that disregard the sociocultural context of adherence, or that rely on models of adherence that are not appropriate to the cultural context, are more likely to find their efforts to understand and improve adherence ineffective or irrelevant (Sankar et al., 2006; Ware et al., 2006). Although underutilized in the HIV field, qualitative research methods can be particularly useful to understanding the multiple factors that influence adherence and risk behavior patterns and developing suitable adherence measurements and improvement strategies (Friedland, 2006; Sankar et al., 2006). This chapter examines four important aspects of adherence: defining it, measuring it, improving it, and analyzing data on it. The committee makes recommendations for future practice and research in each area. The chapter also highlights the need for multidisciplinary teams to col- laborate in addressing challenges to adherence in trials of HIV prevention interventions. Defining Adherence Although researchers agree on the importance of product adherence in both research and real-world settings, there is less agreement on how to define it. Despite the complexity of adherence, clinical trials often report it as a simple number, such as the percentage of coital acts in which par- ticipants use a gel, or the percentage of pills they take over a given time period (Chesney, 2006). The use of a single number to define adherence may mask crucial insights into adherence problems, product acceptability, and potential areas for intervention (Kerr et al., 2005; Berg and Arnsten, 2006). This number can also reflect variability stemming from adherence behaviors that the measure is not intended to address (“construct-irrelevant variance”) (Kerr et al., 2005). Consider a trial to investigate whether suppression of HSV-2, the her- pes simplex virus, prevents HIV infection. The perfectly adherent patient would take one dose in the morning and one in the evening at the same times each day (Vrijens et al., 2006). Suppose the study identifies four participants with imperfect adherence who take 79 percent of prescribed doses during an observation interval. That simple percentage can mask highly disparate adherence patterns (see Figure 5-1): The first patient was

V VVVVVVVVVVVVVV VVVVVVVVVV V V V V 03:00 03:00 24:00 24:00 18:00 18:00 12:00 12:00 Dosing Time Dosing Time 06:00 06:00 03:00 03:00 Comp Comp Start V1 V2 V3 End Start V1 V2 V3 End fdur fdur Non Pers. Pers. Daily Intake 0 20 40 60 80 100 120 0 20 40 60 80 100 120 Dosing Date Dosing Date 2 doses V VVVV VVVVV 1 dose V 03:00 03:00 24:00 0 doses 24:00 18:00 18:00 12:00 12:00 Dosing Time Dosing Time 06:00 06:00 03:00 03:00 Comp Comp Start V1 V2 V3 End Start V1 V2 V3 End fdur fdur Pers. Pers. 0 20 40 60 80 100 120 0 20 40 60 80 100 120 Dosing Date Dosing Date alternate fig 5-1--grabbed from Methods for Compliance Measurement, B. Vjirens and J. Urquhart, 2007 Figure 5-1  Differing patterns of product use among four patients—all of whom are 79 percent adherent. Source: Vrijens, 2007. 121

122 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS adherent but discontinued the prescribed regimen early. The second patient took the morning doses consistently but missed several evening doses. The third patient took the dosages erratically and stopped altogether for a period of time. The fourth patient decided to take only the morning dose and stopped taking the evening dose. Even though each patient was “79 percent adherent,” the clinical consequences of those four patterns could be very different, and the actions needed to improve adherence would differ for the four participants. As Figure 5-1 shows, product adherence is a complex concept. It involves three major components: acceptance of the prescribed regimen, execution of the regimen, and discontinuation of the regimen (Vrijens and Urquhart, 2005; Friedland, 2006) While acceptance and discontinuation tend to happen at a single point in time, quality of execution varies over time. Two aspects of adherence are particularly important in analyzing the results of a clinical trial: persistence, and quality of execution—commonly called compliance. Persistence is the amount of time between when a patient first uses a product and when she or he discontinues it. Quality of execu- tion is the correspondence between the prescribed regimen and the patient’s actual application history during the period of persistence (Vrijens and Urquhart, 2005). That is, quality of execution measures whether a patient has complied with instructions for using a product. Figure 5-2 illustrates these concepts in 20,000 patients who were pre- scribed regular doses of drugs for various diseases, aggregated across multi- ple clinical trials (Vrijens, 2007). The horizontal line at the top of the graph represents perfect adherence. The persistence curve reflects the proportion of patients who continued product use during the first year. While persis- tence varied across disease and trials, these data indicate that 40 percent of participants discontinued the prescribed intervention in the first 12 months. The adherence/compliance curve reflects on a daily basis the proportion of trial participants who executed product use according to the ideal regimen. About half of the participants did so. Describing and analyzing persistence separately from quality of execu- tion provides insights into whether a product is acceptable at both the individual and population levels, and suggests different modes of action. A trial population consisting of 50 percent nonpersisters who are otherwise perfect executers would have very different implications than if all subjects skip every other application. In the first instance, the product would work very well for a select subgroup. The second situation would require more complex analysis and intervention. Recommendation 5-1: Because simple measures of adherence can mask substantially different underlying adherence problems, investigators

DESIGN CONSIDERATIONS: ADHERENCE 123 Non acceptance 100 Perfect adherence Treatment 90 discontinuation 80 Percent Persis tence 70 Adher n io ence / 60 c ut omplia ec nce ex 50 n No 40 0 100 200 300 Days PKC: N > 20,000 patients Figure 5-2  Different aspects of adherence among 20,000 patients in multiple clinical trials. Source: Vrijens, 2007. 5-2 should develop and use adherence measures that can capture different adherence patterns over time. Measuring Adherence, Sexual Behavior, and Condom Use Investigators can gather information on product adherence and risk behavior through a variety of indirect and direct measures. Adherence and risk behavior are commonly measured by self-report through self- administered questionnaires, audio computer-assisted self-interviewing (ACASI), face-to-face interviews with participants, and participant diaries. Other indirect measures include pill counts, electronic product monitoring, pharmacy refills, and biomarkers of product exposure and risk behavior, such as applicator staining for vaginal insertion, or presence of semen

124 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS indicating unprotected sex. Direct measures of product adherence include pharmacokinetic studies (which measure drug levels or metabolites in sub- jects’ blood or bodily fluids), and directly observed therapy. These measures can vary substantially in expense, the effort required of participants and their partners, their perceived invasiveness, and their accuracy and reliabil- ity (Berg and Arnsten, 2006). Complicating matters is the lack of a “gold standard” for measuring adherence. Most research on measuring adherence in HIV studies has examined HIV-infected individuals’ adherence to ART. Measuring individuals’ risk behavior is also critical to knowing whether a product worked. Given the limited published empirical evidence on adherence measures for nonvaccine biomedical HIV prevention interventions, the committee draws heavily on the fields of ART adherence and sexual risk behavior assessment in making its assessment and recommendations. Indirect Measurement Methods Self-Reports of Product Adherence and Risk Behavior Self-report of product adherence and risk behavior is widely used in research and clinical settings because it is relatively inexpensive, easy to administer, allows for probing about nonadherence, and has low participant burden (Berg and Arnsten, 2006). However, the accuracy of self-reports is controversial and has been the focus of substantial research. Reliability and validity are the two most important psychometric aspects of self-report measurements. Reliability refers to whether the instru- ment is free of random error and validity refers to whether the instrument is measuring what it intends to measure (Pequegnat et al., 2000). Self-reports can be incorrect because a person fails to respond truthfully or does not accurately recall their behavior (Pequegnat et al., 2000). Participants may respond untruthfully to questions about their adherence and risk behavior because they want the interviewer, study staff, or other participants to view them more favorably—a phenomenon known as “social desirability” bias (Pequegnat et al., 2000; Schroder et al., 2003b; Simoni et al., 2006b). Even with truthful responses, the accuracy of self-reports can be affected by the length of the recall period (for retrospective reports), the question format, appropriateness of the assessment mode, and individual factors (e.g., such as the frequency of behaviors, educational level, age, or use of alcohol and drugs) (Pequegnat et al., 2000; Schroder et al., 2003a,b, 2007).   ee Pequegnat et al., 2000; Simoni et al., 2006b; and Schroder et al., 2003a,b, 2007, for a S detailed discussion of psychometric factors related to adherence and risk behavior assessment and associated references.

DESIGN CONSIDERATIONS: ADHERENCE 125 Studies have found that self-reports of adherence tend to be positively skewed, producing higher estimates of adherence compared to more objec- tive measures (e.g., electronic drug monitoring) (Berg and Arnsten, 2006; Simoni et al., 2006b). In contrast, studies comparing self-reports of sexual behavior with biomarkers of semen exposure indicate that individuals may underreport sensitive or stigmatized risk behavior (see, for example, Gallo et al., 2007). Researchers have identified several methods to help mitigate these problems. First, investigators can enhance the validity and reliability of self- reports by stressing the importance of truthful responses, ensuring anonym- ity or confidentiality of responses, and allowing for privacy in answering questions (Pequegnat et al., 2000). Guidelines for minimizing social desir- ability bias in ART adherence assessment include using self-administered measures with either open-ended or forced choices, acknowledging the difficulty of perfect adherence to the participant (“normalizing” imperfect adherence), querying reasons for nonadherence, focusing on recent behav- ior, clearly specifying a time frame, using recall aids and anchoring reports to salient events, and conducting reliability checks (Simoni et al., 2006b). Investigators can also use methods of collecting self-reported infor- mation that are more likely to promote reliable and valid self-reports. Face-to-face interviews can be prone to overreporting of adherence and underreporting of sensitive risk behaviors (Jadack et al., 2001; Rogers et al., 2005). Use of interviewers who are independent of the study staff can help create a neutral climate (UNAIDS, 2007). Self-administered question- naires may also decrease the likelihood of social desirability bias. While written self-administered questionnaires are inappropriate in areas with low literacy rates, the use of ACASI can address some of the problems with face-to-face interviews. A number of studies have suggested that participants are more likely to report sensitive behaviors, such as sexual behavior or drug use, if investiga- tors rely on ACASI rather than in face-to-face interviews. ACASI can also help researchers check the consistency of participants’ answers and reduce the number of missing data fields, or “don’t know” responses. Several stud- ies have shown that this technique is feasible and acceptable in a variety of international settings among different at-risk populations with varying literacy rates and computer skills (Van De Wijgert et al., 2001; Simoes et al., 2006a,b; NIMH Collaborative HIV/STD Prevention Trial Group, 2007). Contraception trials have used prospective methods, such as coital diaries, extensively to collect self-reported information, and to validate ret- rospective self-reports of sexual behavior (Schroder et al., 2003a). The first phase 3 trial of a microbicide (Nonoxynol-9, or N9) initially used diaries in the form of pictorial log charts (Van Damme et al., 2002). However, after finding that women sometimes filled in their diaries while waiting at the

126 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS clinic, the trial abandoned the diaries in favor of interviews (Van Damme et al., 2002). The Microbicide Development Program’s phase 3 trial of the PRO2000 microbicide is now using coital diaries in a subset of women to capture information on sexual behaviors and product use. In a feasibility study prior to the start of the phase 3 trial, investigators found that women tended to report sexual behaviors more frequently in diaries than in face- to-face interviews (Allen et al., 2007). Electronic diaries such as personal digital assistants (Bartley et al., 2004), mobile phones (Hays et al., 2001), pagers (Shrier et al., 2005), and Web applications (Baer et al., 2002) have also been tested, and could offer important advantages over paper diaries. Electronic diaries can allow internal checks on the validity of the data, and could record time and date information (Raymond and Ross, 2000). Electronic diaries also eliminate the need to transcribe and code data, allowing more detailed and timely analysis (Bartley et al., 2004). Studies have shown that subjects are willing to use electronic diaries, and that they may actually do so more often than paper diaries (Hufford, 2002). Despite extensive work, several psychometric aspects of self-report mea- surements remain unresolved. A key problem is the lack of standardization in adherence and risk behavior measurement instruments across research and clinical settings (Schroder et al., 2003b; Berg and Arnsten, 2006). These measures vary in terms of length of recall period, the type of measure (quali- tative versus quantitative measures), the question format, among other factors. There is poor agreement between various self-report measures and the variation makes it difficult to compare results across studies (Berg and Arnsten, 2006). Recent review papers suggest that cognitive interviewing, which examines how target audiences interpret questions, process infor- mation, and form responses to survey questions, and additional empirical evaluations of self-reported adherence questions can further improve the validity and reliability of self-reports (Berg and Arnsten, 2006). Pill and Applicator Counts Many clinical trials rely on less specific but possibly more objective measures of product adherence, such as pill counts. Asking participants to return unused pills and product applicators during routine visits is a rela- tively inexpensive measure. However, such measures can be time consuming and subject to bias if participants “dump” unused products prior to a visit (Berg and Arnsten, 2006). To address these concerns, some studies have found that unannounced pill counts at participants’ homes can provide a reliable measure of ART adherence (Bangsberg et al., 2000, 2001). This approach reduces the problems with asking participants to return unused product at scheduled visits, reduces their opportunity to empty pill contain-

DESIGN CONSIDERATIONS: ADHERENCE 127 ers, and does not require them to bring all their medications to the clinic (Bangsberg et al., 2001). Yet unannounced pill counts at participants’ homes can be expensive and logistically difficult, and some participants may prefer not to have home visits (Kalichman et al., 2007). In a recent study of 77 HIV infected individuals in Atlanta, Kalichman and colleagues (2007) found that unan- nounced telephone-based pill counts were a logistically feasible and eco- nomical method for monitoring adherence to ART medication. Pharmacy Refills Maintaining pharmacy refill records is another method for indirectly measuring product adherence. If patients do not receive timely refills from the pharmacy, the investigator can assume that the patient is either miss- ing doses or not taking the medication at all. “Medication gaps” or the period that the patient’s supply of product is assumed to be exhausted is determined by a comparison of the actual refill dates with the expected refill dates (Berg and Arnsten, 2006). This method of measuring adherence relies on two major assumptions. Participants who do not receive timely refills are not obtaining product from other sources, and participants who do receive timely refills use the product as prescribed. In the clinical setting, a patient may have multiple opportunities to access medication, such as through family members and friends or other pharmacies. However, in the trial setting, access to the product usually is limited to the study pharmacy and study population, though product shar- ing among trial participants has been raised as a concern in some studies. Given that distribution of the study product is more tightly controlled, the use of pharmacy refill records may have greater applicability for measuring adherence in the trial setting than in the clinical setting. Similar to pill counts, pharmacy refill records are not immune to bias, as participants may “dump” or share unused products. Timely refills do not guarantee that the participant took the product as instructed. Despite this limitation, several HIV treatment studies have shown a significant correla- tion between pharmacy refill adherence and HIV viral load (Maher et al., 1999; Low-Beer et al., 2002; Grossberg et al., 2004; Fairley et al., 2005) all studies can be found in Berg and Arnsten (2006). Electronic Medication Monitoring A commonly used approach to measuring adherence to ART is the “medication event monitoring” system (MEMs). MEMs uses microcircuitry in pharmaceutical packages to detect, time-stamp, analyze, and store infor-

128 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS mation on when users remove a dose, and communicates that information to investigators (Urquhart, 1997). MEMs provides an “objective” and time- specific measure of adherence, and thus is often considered more valid than other types of measures, such as self-reports (Berg and Arnsten, 2006). Yet MEMs has several important limitations. It requires participants to store all medications in the MEMs container, to remove only the correct number of pills at each dosing, to open the container only during dosing, and to close the monitor after dosing (Bangsberg et al., 2001). Several stud- ies have documented “pocket dosing,” in which participants remove mul- tiple doses from the container at one time, which can lead to underestimates of adherence (Bova et al., 2005). Other studies have identified “curiosity checks,” in which participants open the containers to test them, or to see how many tablets they have left, which may lead to overestimates of adher- ence (Bangsberg et al., 2001; Bova et al., 2005; Berg and Arnsten, 2006). MEMs may be valuable in assessing adherence to medication-based approaches such as PrEP or HSV-2 suppression therapy. Researchers have also tested modifications of MEMs-type monitors for topical medications and found them feasible (Tusa et al., 2006). Investigators may find it useful to adapt MEMs to microbicide applicators. Other approaches being devel- oped is the “smart” vaginal microbicide applicator, called Xigo, that stores information about product use (Rosenberg, 2007) and the “sexometer”—an intravaginal ring called Paragon that attempts to capture the time and date of sexual intercourse based on motion indicators (Rosenberg, 2007). Biomarkers of Product Use and Sexual Activity Researchers have developed biomarkers of product use and sexual activity that have potential for validating self-reported information in trial settings. To assess microbicide product use, researchers developed an assay that assesses lactobacillus growth on returned used applicators to determine exposure to the vagina (Wallace et al., 2004; Hogarty et al., 2007). This method was used in the Carraguard microbicide trial to distinguish appli- cators that had been inserted vaginally versus those that were not inserted (Skoler, 2007). The assay cannot reveal when the gel was used, which is important information for coitally dependent microbicides, or whether it was used by the participant or someone else. Nor can it determine whether the product was used each time it should have been. It can, however, reveal the absolute level of exposure to a product. A subsequent study of this approach in a 14-day microbicide trial of 0.5% PRO 2000 and placebo gel returned applicators found that there was high concordance between self- report and applicator staining (Hogarty et al., 2007). Although incidence of sexually transmitted infections (STIs) has been suggested as a potential biomarker for unprotected sexual behavior, trans-

DESIGN CONSIDERATIONS: ADHERENCE 129 mission of STIs also depends on individual susceptibility, sexual partners, and the characteristics of the infection (Pequegnat et al., 2000). Researchers have recently examined biomarkers of semen exposure as potential tools for validating self-reports of condom use and recent sexual activity. These candidate biomarkers can be classified into two categories: those that detect seminal plasma, and those that detect spermatozoa and other cells present in semen (see Mauck and Doncel, 2007 for a review of methods). The seminal plasma biomarker with the most extensive testing in clinical trials is the prostate specific antigen (PSA) (Lawson et al., 1998; Macaluso et al., 1999). Because PSA begins to clear from the vagina immediately after it is exposed to semen, PSA detection likely underestimates exposure (Gallo et al., 2007). Another biomarker for semen exposure is the Y-chromosome (Yc) DNA. Researchers developed a Yc polymerase chain reaction (PCR) assay (Jadack et al., 2001; Zenilman et al., 2005). Yc-DNA may offer advantages over PSA because it can be detected several weeks post-coitus (Mauck and Doncel, 2007). Such biomarkers have been used primarily to evaluate the effective- ness of contraceptives, but they may have a role in validating self-reports of sexual activity in HIV/STI prevention trials (Mauck and Doncel, 2007). For example, a recent study used PSA to assess the validity of self-reported condom use among female sex workers in Kenya, and found that 11 per- cent of samples from women who reported no unprotected sex in the prior 48 hours tested positive for PSA (Gallo et al., 2007). A follow-up study to the phase 3 MIRA diaphragm trial also used PSA to assess the validity of self-reported sexual behavior (Mauck and Doncel, 2007). Upon completion of the trial, investigators randomized a subset of women to an additional session using either ACASI or face-to-face interview. They then conducted a PSA analysis on women who reported no intercourse in the previous 48 hours, to assess whether misreporting differed by interviewing technique. The investigators have not yet reported the results of this analysis. These or future assays may help validate self-reports of sexual behavior, and possibly help investigators assess the safety and efficacy of microbicides (Mauck and Doncel, 2007). However, the use of vaginal products, vaginal washing, menses, and infection may affect the sensitivity of the assays, so the ultimate value of the approach requires further research (Zenilman et al., 2005; Mauck and Doncel, 2007). Direct Measurement Methods Drug Monitoring Biological assays of active drug, metabolite, or other markers in blood, urine, or other specimens can provide information on individuals’ expo-

130 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS sure to drug products. The value of examining drug levels depends on the half-life of the product (Osterberg and Blaschke, 2005). Such tests are thus most valuable for products with longer half-lives. Studies suggest that when exposure is measured during scheduled clinical visits, “white-coat compli- ance” can occur: that is, heightened awareness may spur participants to have better-than-normal adherence to treatment prior to their scheduled visit (Cramer et al., 1990). Measurements at those intervals may not be an accurate reflection of overall adherence. Directly Observed Therapy Directly observed therapy (DOT)—in which members of the study staff administer all doses of a product regimen to individuals, or observe their intake—is most commonly used as an intervention to increase adherence. It is also the only approach that provides near-perfect information on indi- viduals’ adherence and exposure to an HIV prevention product. DOT is quite feasible in some settings and populations. For example, DOT and modified DOT (in which only a portion of the product regimen is taken under supervision) have been used extensively with ART (Goggin et al., 2007), including in resource-poor settings (Pearson et al., 2007). One late-stage biomedical HIV prevention trial now using DOT is the CDC-sponsored phase 2/3 PrEP trial in injecting drug users in Thailand. Participants were given the option of daily DOT or monthly follow-up. Approximately 85 percent of participants opted for daily DOT of their PrEP dose of Truvada, along with their daily methadone treatment (Smith, 2007). Researchers are also considering using DOT for other products that require application once a day, such as a microbicide gel, or once a month or less often, such as the vaginal ring (Rosenberg, 2007). More widespread use of DOT is often limited by cost and logistical constraints, and by the fact that it cannot be used for coitally dependent products such as the first generation of microbicides. Nevertheless, DOT and modified DOT could be very useful in efficacy or proof-of-concept trials to minimize interpretation problems that result from nonadherence. However, the committee has concerns about using DOT in effectiveness tri- als of biomedical HIV interventions, if that approach cannot be sustained in real-world practice because the trial results may be poor predictors of the effectiveness of the interventions. Since no adherence measurement tool is perfect, several studies have found that using multiple measures to “triangulate” adherence levels and risk behaviors is helpful in reducing the error introduced by any particular method (Liu et al., 2001; Pool et al., 2006). However, investigators relying on that approach must directly address inconsistencies in the results from different measures, rather than simply identifying them (Pool et al., 2006).

DESIGN CONSIDERATIONS: ADHERENCE 131 And any adherence measures that rely on recall must entail short recall periods. Recommendation 5-2: In light of the uncertainty about the accuracy of various methods for collecting data on adherence and risk behavior, investigators of biomedical HIV prevention trials should strive to use multiple types of measures to triangulate adherence estimates. Rather than collecting detailed information on all participants, investigators could collect more detailed information on a well-chosen random sam- ple, and collect less detailed information on all participants. Recommendation 5-3: Although directly observed therapy or modi- fied DOT could be very useful in proof-of-concept trials, investigators should not use these methods in effectiveness trials if that approach will not be used in real-world practice, because the trial results may then be poor predictors of the effectiveness of the interventions. Strategies to Improve Adherence High levels of adherence to a product regimen by participants in a clini- cal trial are critical to determining that product’s efficacy—and ultimately its public health impact. However, little empirical evidence exists on the effectiveness of strategies to improve adherence to nonvaccine biomedical HIV prevention interventions. The committee was unable to identify any publications specifically evaluating adherence strategies for such interventions. However, evidence on the effectiveness of methods to improve the adherence of HIV-infected patients to ART may inform efforts to enhance adherence in biomedical HIV prevention trials. Studies of ART adherence interventions may have particular relevance for medication-based HIV prevention strategies, such as PrEP or acyclovir for HSV-2 suppression. Although some aspects of adherence undoubtedly differ between the two arenas, they have important similarities: biomedical HIV prevention trials target products to uninfected individuals, and HIV-infected individu- als often start ART while they are still asymptomatic. In addition, both treatment and prevention interventions may require patients to follow daily regimens indefinitely. This section reviews the effectiveness of strategies to improve adher- ence to ART, examines the lessons learned and knowledge gaps, and makes recommendations for applying such strategies in biomedical HIV preven- tion trials.

132 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS Effectiveness of Strategies to Improve Adherence to ART Given the advent of combination antiretroviral therapy in the 1990s, the field of improving adherence to HIV treatment is still relatively young (Simoni et al., 2006a). However, a number of literature reviews on strate- gies to increase adherence to ART have been completed. Early qualitative reviews found that most studies of adherence strategies lacked sufficient methodological rigor, such as inadequate sample size, lack of randomiza- tion or control conditions, and failure to use intent-to-treat analyses, among others (Simoni et al., 2003). Amico and colleagues (2006) conducted the first quantitative review of randomized and uncontrolled studies of strategies to improve adherence to ART published between 1996 and 2004 (24 studies). They found that these strategies had a significant (P < 0.05) aggregated effect on adherence to ART, but that the magnitude of the effect varied greatly across studies. For example, Amico and colleagues found that strategies targeting participants with poor pretest adherence had greater effects than strategies targeting groups with a variety of pretest adherence levels. The analysts conclude that to design effective strategies, investigators must carefully delineate the target population. Amico and colleagues found no evidence that the effects of adherence strategies decayed over time, but few of the studies included extended follow-up periods. Like earlier qualitative reviews, they found that many of the studies were generally underpowered and would not have been able to detect a small to moderate sized effects. In an update and extension of this work, Simoni and colleagues (2006b) conducted a meta-analysis of 19 randomized, controlled trials of strategies to improve ART adherence, measuring their impact on reported adherence and participants’ HIV-1 RNA viral load. They found that participants receiving adherence strategies were about 1.5 times as likely (95% CI: 1.16–1.94) to report at least 95 percent adherence as participants in com- parison conditions, and about 1.25 times as likely (95% CI: 0.99–1.59) to achieve an undetectable viral load. However, some strategies did not improve adherence, and most of those that did had only modest and short- term effects. A Cochrane Collaboration review examined 57 randomized, controlled trials of interventions to improve adherence to medication across a variety of medical conditions, and found similar limitations. The reviewers con- cluded that “almost all of the interventions that were effective for long-term care were complex,” and that “even the most effective interventions did not lead to large improvements in adherence and treatment outcomes” (Haynes et al., 2005, p. 1). While the two meta-analyses of strategies to improve ART adherence suggest that they can be somewhat effective, the reviews offer few guidelines

DESIGN CONSIDERATIONS: ADHERENCE 133 on which strategies are most effective (Simoni et al., 2007). Despite these limitations, several interventions to improve adherence to ART have shown promise, and could be worth further exploring in biomedical HIV preven- tion trials. These include cognitive-behavioral strategies (e.g., Pradier et al., 2003; Mannheimer et al., 2006; Petersen et al., 2007; Rueda et al., 2007), social support interventions (Remien et al., 2005; Williams et al., 2006), and directly observed therapy or modified directly observed treatment (e.g., Goggin et al., 2007; Pearson et al., 2007), and contingency management (Haug and Sorenson, 2006; Rosen et al., 2007). Applying ART Adherence Improvement Strategies to HIV Prevention The findings from studies and meta-analyses of strategies to improve ART use suggest ways to improve adherence to biomedical HIV prevention interventions. However, important knowledge gaps remain: • As noted in a recent meta-analysis (Amico et al., 2006) and several qualitative reviews (Haddad et al., 2000; Fogarty et al., 2002; Simoni et al., 2003), many studies of adherence strategies for ART lacked sufficient meth- odological rigor, were underpowered, and lacked theoretic underpinnings. • Publications on trials of strategies to improve adherence often do not describe in enough detail the nature, content, and intensity of the strategies. Yet this information is important in evaluating the adherence intervention, and comparing outcomes across studies (Amico et al., 2006). In reviewing the protocols for non-vaccine biomedical HIV prevention tri- als, the committee generally found a similar paucity of information on the types and frequency of planned adherence improvement strategies, and the factors that might trigger changes in those strategies. • Most studies evaluating ART adherence strategies have focused on individuals, even though substantial research indicates that factors at the provider, clinic, and sociocultural level can affect adherence to HIV inter- ventions (Gordon, 2006). • Because most trials evaluating ART adherence strategies were con- ducted in high-income countries, their applicability to resource-poor areas is uncertain (Gordon, 2006). • A key unknown is the extent to which strategies that improve adherence to HIV treatment apply to uninfected individuals who must take   ontingency management (CM) typically involves a voucher- or monetary-based reinforce- C ment technique in which individuals are rewarded for sustaining positive behavior. CM has been used successfully in substance abuse treatment settings to reinforce treatment goals such as drug abstinence or completion of certain activities. CM may also help promote adherence in the HIV field (Haug and Sorensen, 2006).

134 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS an experimental HIV prevention agent over a long period of time. While the consequences of imperfect adherence for HIV-infected individuals are tangible, those for uninfected individuals are less so. • Different subgroups within a given population may need different adherence strategies. Such strategies may also exert different effects across subgroups (Chesney, 2006; Simoni et al., 2006a). Studies of strategies for improving adherence to HIV prevention inter- ventions are lacking but essential. Evidence is especially needed in the context where adherence matters most, and where it may well be most challenging: for coitally dependent microbicides, and for interventions that individuals must use indefinitely, such as PrEP, microbicides, and HSV-2 suppression. Successful research on strategies to improve product adherence requires collaboration among behavioral, social, and quantitative scientists, despite the barriers to such collaboration. It is critical for multidisciplinary teams to study adherence challenges in a given setting, and to develop socially and culturally relevant strategies for improving adherence. Such teams also need to pursue research on how best to translate strategies that are effective in one setting to settings with different personal, economic, and sociocultural influences on adherence and risk behaviors (Chesney, 2006). Investigators designing clinical trials of biomedical HIV prevention interventions may need to adapt and combine various strategies to maxi- mize adherence (Haynes et al., 2005). As discussed in Chapter 10, facto- rial study designs, and dynamic designs, can allow such investigators to empirically evaluate alternative adherence strategies. Investigators of HIV prevention trials also need to consider using adherence strategies that target couples, groups, and communities as well as individuals (Gordon, 2006). Recommendation 5-4: Donors should fund and investigators should undertake empirical evaluations of strategies to increase adherence to biomedical HIV prevention products during and after a clinical trial. These evaluations should be adequately powered, methodologically rigorous, socially and culturally relevant, grounded in behavioral and social science theories, and conducted in the regions where the strate- gies will be utilized. Recommendation 5-5: Investigators should specify in the study proto- col detailed plans for monitoring, measuring, and analyzing adherence data, and steps they will take to improve adherence if it is poorer than anticipated.

DESIGN CONSIDERATIONS: ADHERENCE 135 ANALYZING Adherence Exposure to a product, adherence to instructions for using it, and behavior related to that use are key factors on the causal path from preven- tive intervention to HIV infection. Figure 5-1 illustrates how randomization of participants in a clinical trial has a direct effect on their product expo- sure, but is not itself influenced by their baseline characteristics, including pre-randomization sexual behavior. It is this particular feature that enables investigators to estimate the causal effect of study arm assignment in the traditional intention-to-treat analysis, which compares groups “as random- ized” (that is, based on the intended intervention). However, baseline char- acteristics of participants may influence how much product exposure they ultimately experience, and both factors can influence changes in behavior. In Figure 5-3, arrows emanating from each of these features point to a direct impact on HIV incidence. Additional variables, observed or unob- served, may also enter the picture. Analyzing patterns of exposure to the product, adherence to the prod- uct regimen, and accompanying behavior—including vaginal, anal, and oral sex, with or without various forms of protection (condom and/or product)—can yield important information about the study population. Such analyses also yield information on the acceptability and feasibility of the intervention, and the extent to which the trial results will apply to the target population. Analyzing the association between exposure or adherence patterns and HIV incidence can help investigators estimate the causal effect of different interventions and reveal whether the primary intention-to-treat analysis is estimating efficacy or some particular form of effectiveness, given the observed dosing schedule. However, “measurement error,” (less- than-perfect information on adherence and behavior, due to inaccurate reporting or measuring of adherence) will limit investigators’ ability to interpret intention-to-treat analyses, and to recognize and monitor adher- ence problems, distinguish nonresponders from nonadherers, and provide adherence-specific estimates of the effects of the product, which can guide further development. This section examines the major sources of variation in adherence patterns within a trial. In each case, the committee discusses how random and systematic measurement error introduced by subjective adherence mea- sures can affect the results and their interpretation. Table 5-1 summarizes this analysis. The committee also suggests adherence analyses that can reveal subject-specific baseline variables (sometimes called moderator vari- ables) and variations in behavioral responses (sometimes called mediator variables)—both of which can shed light on the potential impact of an intervention (Mackinnon and Dwyer, 1993).

136 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS Adaptation of Exposure to risk behavior intervention product Randomization HIV infection Baseline risk status behavior Figure 5-3  A causal diagram: some key variables and their potential direct effects (the arrows). Investigators of trials with low levels of adherence will have difficulty assessing a product’s efficacy. Such levels5-3 also indicate that the product Figure may will be unacceptable or infeasible for wider use in the community. Effect of Measurement Error Because measurements of adherence usually overestimate true adher- ence, low reported levels of adherence can usually safely be assumed to reflect low actual adherence. However, even if reported adherence levels in a trial are high, doubt may remain about true adherence levels, and thus about the product’s efficacy and value in the community. Comparing Adherence Between Study Arms Differences in adherence between the intervention and control arms in a randomized trial imply that different products have different side-effect profiles, or otherwise affect participants differently. These differences may emerge early in a trial or become apparent over time. If the latter occurs, blinding has not necessarily failed. A small increase in a mild side effect in the intervention arm could lead to lower adherence even if participants are not aware of which arm they are in. Such an outcome may produce a higher dropout rate in the intervention arm, and hence lower adherence. On the

DESIGN CONSIDERATIONS: ADHERENCE 137 Table 5-1  What Adherence Analyses Can Reveal Analysis Possible Results Interpretation Analyze level of product The measured levels Efficacy trial: This shows to adherence in each study indicate how much what extent the measured arm. participants deviate from effect should indeed be seen as the prescribed dose timing, a measure of efficacy. and possibly the method of using the product. Effectiveness trial: This may reveal problems with the feasibility and acceptability of the intervention in the study population within the trial setting. Analyze differences The measured levels of This might indicate that in product exposure, product adherence and the active product versus adherence, and risk sexual behavior differ placebo produce different side behavior between study between study arms. effects, prompting subjects to arms.a comply differently with the assignment.b Analyze differences People within study arms Some subpopulations might be in product exposure, differ greatly in how better suited to another type adherence, and risk they use the product, and of protection than the one behavior between subjects possibly also in their risk- under study. within randomized arms. taking behavior. Analyze changes over People within randomized If rates of adherence among time in product exposure, arms differ greatly in how individual subjects drop adherence, and risk they use the product over dramatically over time, behavior among subjects time. this might indicate that the within randomized arms. intervention is not sustainable. That, in turn, may indicate that side effects emerge after cumulative use, or the need for supportive measures to improve adherence. If long- term use of the intervention is envisaged, suggestions that sustainability is limited would require investigators to further examine the appropriateness of the intervention. If adherence rates increase over time, subjects may be getting better at adhering to the intervention. continued

138 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS Table 5-1  Continued Analysis Possible Results Interpretation Analyze how levels of Subjects appear to be This could reflect risky behavior differ from highly sexually active, or disinhibition, or that subjects those that can be expected they engage in more risky want to become pregnant outside the trial. behavior than expected within the healthy study outside the trial context. setting. Such behavior could lead to sexual acts against which the product is not designed to protect, and may put subjects at higher risk of HIV infection. Analyze how levels of risky Subjects differ between This could distort the behavior differ between arms in the amount of intention-to-treat effect. That, study arms. risky behavior they pursue in turn, could give a distorted against which the product view of the direct impact of is not designed to protect the product under real-world (such as anal sex). conditions. aAny interpretation of observed differences in adherence between arms will need to account for differential dropout rates over time. Such differences in dropout rates could occur, for in- stance, if a product is protective, and HIV incidence differs between arms. The least-protected people will tend to drop out first owing to HIV infection. bThis does not mean that blinding has failed. A small increase in a mild side effect in the study arm could lead to lower product adherence without participant awareness. These dif- ferences may emerge early in the trial or become apparent with accumulated product use over time. other hand, if a product is protective and HIV incidence differs between arms, the least-protected participants would tend to drop out sooner. Effect of Measurement Error Investigators would have no reason to expect measurement error to differ between study arms in a blinded randomized trial. They can there- fore expect significant differences in adherence between arms to reflect true differences. The interpretation of reported adherence differences between arms in an unblinded trial is usually less clear. For example, in the MIRA diaphragm/ Replens trial, observed rates of condom use were much higher in the “con- dom-only” control arm than in the intervention arm (Padian et al., 2007). One explanation is that subjects in the condom-only arm adhered to guide- lines on condom use more than did subjects in the condom-plus-diaphragm arm, because they felt that this was their only form of protection against

DESIGN CONSIDERATIONS: ADHERENCE 139 HIV infection. However, a different explanation is that they over-reported condom use because they could not report diaphragm use and wished to appear compliant. Comparing Adherence Patterns Within Randomized Arms Large variations in reported adherence between subjects within a study arm may indicate that some subpopulations are better suited for another type of protection. For example, if participants in one arm used a product for 80 percent of sexual acts, it matters whether all subjects failed to use the product for 20 percent of the acts, versus whether 20 percent of subjects did not use the product at all and 80 percent used it every time. In the lat- ter case, investigators would hope to recognize the 20 percent of nontakers early on and offer them a different intervention option. If most people use the product only 80 percent of the time, the intervention may be impracti- cal. Investigators need to address these differences in their analyses, rather than reporting only the percentage of people who comply with the regimen in each arm. Effect of Measurement Error If subjects within a study arm differ significantly in their reported expo- sure to the product, adherence to the product regimen, and sexual behavior, investigators will need to consider whether the differences reflect true dif- ferences, errors in measurement, or a combination of both. Adherence patterns for individuals may change over time. If compli- ance rates for individuals drop dramatically during a trial, investigators should determine whether side effects are emerging. If adherence declines are common across subjects within an arm, investigators should consider providing additional support or other interventions, especially if the prod- uct is intended for long-term use. If measurement errors remain stable, such differences probably reveal actual trends. When Behavior Differs from That Expected Outside the Trial Participants in a clinical trial might pursue higher rates of sexual activ- ity or other risky behavior because of disinhibition, or because they are trying to become pregnant in the healthy environment of the trial. Such behavior would put them at higher risk of HIV transmission, especially if they pursue more risky behavior against which the product is not designed to protect, such as anal sex. In such cases, the results of the trial’s intention- to-treat analysis could give a distorted view of the impact of product use outside the study.

140 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS Effect of Measurement Error This aspect of the adherence analysis is quite sensitive to measurement error, although systematic errors are again likely to reflect underreporting among trial participants. Association Analyses: Linking Adherence Patterns to HIV Incidence Investigators can attempt to analyze information on reported sexual activity, adherence, and HIV exposure to assess their association with HIV infection. Such analyses may allow them to interpret and explain the observed intention-to-treat effect, judge whether to explore different ways of administering a product, and estimate the effect the product will have in other populations. The observed association between reported product use and HIV inci- dence is prone to confounding, and does not necessarily reflect a real (causal) effect even if reported use is correct. For microbicides, for instance, higher sexual activity tends to increase the risk of HIV transmission as well as the number of gel applications. Investigators would therefore expect to find a positive association between the product use (such as applications per day) in the experimental arm and HIV incidence, even in the absence of any causal effect of the product. Nevertheless, information on variations in product use can help inves- tigators interpret the effect revealed by an intention-to-treat analysis, and help them estimate the product’s effectiveness in future populations. As in the N-9 trial, higher HIV incidence rates may occur at high levels of prod- uct use in the intervention arm than at corresponding levels of product use in the placebo arm (Van Damme et al., 2002), and this might reflect an increased risk of HIV infection with use of the intervention. However, investigators need to consider and exclude other explanations before regarding such an effect as causal. The main reason is that a mea- sured compliance level—such as use of gel in at least 80 percent of sexual acts—is a post-randomization category that may itself be influenced by the product, and need not be influenced in the same way among study arms. (In technical investigations of causality, this phenomenon is known as a lack of “exchangeability.”) For instance, if a product is associated with more sexually transmitted diseases, some women may start to use it less often, and hence different subpopulations may have similar use levels in both trial arms. However, when the study arms are blinded and investigators observe similar distributions of adherence between them, they can be more confident in the exchangeability of subpopulations with similar compliance levels among study arms. Errors in reported adherence will further complicate the interpreta-

DESIGN CONSIDERATIONS: ADHERENCE 141 tion of data. Such errors may depend on the actual level of adherence, as more bias may be expected when true adherence is low, and on the study endpoint, as participants may be more reluctant to admit to imperfect adherence after high-risk behavior or an HIV diagnosis, for example. If these biases are at play equally in both study arms, an adherence-adjusted comparison of study arms would tend to convey the correct signal, at least qualitatively. However, assessments of safe adherence levels would tend to be overestimates. In unblinded trials, such as the MIRA diaphragm trial, subjects in the unblinded arm may be under higher pressure to comply with condom use, and may therefore overreport adherence to a larger extent. An adherence-adjusted analysis would then tend to compare HIV incidence rates between somewhat different subpopulations. Statisticians have developed sophisticated causal models that attempt to address questions such as: “What would have occurred if exposure had been different?” One such model links potential responses to different levels of treatment and risk behavior, and allows estimation of the model’s parameters based on the randomization used to assign patients to study arms. The corresponding statistical tests are approximately valid under the null hypothesis of no causal effect, even if the assumed form for the causal model is incorrect (Robins and Tsiatis, 1991; Vandebosch et al., 2005). Such models thus provide a valuable level of protection against confounding. Linear structural mean models (Goetghebeur and Vansteelandt, 2005) also remain valid under the null hypothesis in the face of random mea- surement error, and can be adjusted to account for (known or modeled) average systematic error in measuring compliance. However, when there is a causal effect, the validity of these analyses rests on the assumption that the structural causal model is correct, and the latter is typically not testable based on available data. It is important to verify the assumptions made in causal models as much as possible, and to assess the sensitivity of the results to plausible devia- tions from the assumptions. These models allow investigators to predict, under additional assumptions, the expected effects of a given distribution of product exposure within a population, including under full compliance. The models also allow investigators to correct for differential condom use between the study arms, thus allowing estimation of the direct effect of the intervention. In some settings, investigators compare adherence between partici- pants who become HIV infected and matched controls who do not become infected. A comparison of recent behavior and product exposure could help explain the extent to which seroconverters are nonresponders or non- compliers. However, because such a comparison would be nonrandomized, other factors could confound an observed association. Detailed adherence data obtained retrospectively may also suffer from bias, given that a partici-

142 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS pant’s knowledge that he or she has become HIV infected may change his or her recollection of behavioral details. Investigators could thus complement this retrospective information with simple measures of adherence gathered prospectively, in real time. A substantial difference between prospective (reported by the participant before his or her HIV status is known) and retrospective (reported after the HIV status is known) measures of adher- ence across study arms may indicate differential measurement error. (See Appendix D for more detail on methods for analyzing adherence data.) Recommendation 5-6: Investigators should provide data on product adherence and risk behavior results to the data monitoring committee, as this information may influence the committee’s views of the relative efficacy and safety of the study arms, and the feasibility of the study. Recommendation 5-7: Investigators should analyze adherence and behavior as both outcomes in an HIV prevention trial and modifiers of the effect of the biomedical intervention on HIV infection risk. Recommendation 5-8: Investigators should analyze the potential impact of adherence by doing the following: • Perform a stratified analysis when adherence appears similar between study arms. Such analyses aim to provide unbiased compari- sons of subpopulations across study arms. • Postulate causal models and perform randomization-based analyses. • Perform matched case-control adherence analyses involving sub- jects who become HIV infected. When reporting model-based analyses of adherence, investigators should clearly state the model’s assumptions and discuss their plausibility, as well as the robustness of the analysis to violations of the assumptions. REFERENCES Allen, C. F., S. S. Lees, N. A. Desmond, G. Der, B. Chiduo, I. Hambleton, L. Knight, A. Vallely, D. A. Ross, and R. J. Hayes. 2007. Validity of coital diaries in a feasibility study for the microbicides development programme trial among women at high risk of HIV/AIDS in Mwanza, Tanzania. Sexually Transmitted Infections [Epub ahead of print]. Amico, K. R., J. J. Harman, and B. T. Johnson. 2006. Efficacy of antiretroviral therapy ad- herence interventions: A research synthesis of trials, 1996 to 2004. Journal of Acquired Immune Deficiency Syndromes 41(3):285-297. Baer, A., S. Saroiu, and L. A. Koutsky. 2002. Obtaining sensitive data through the web: An example of design and methods. Epidemiology 13(6):640-645.

DESIGN CONSIDERATIONS: ADHERENCE 143 Bangsberg, D. R., F. M. Hecht, E. D. Charlebois, A. R. Zolopa, M. Holodniy, L. Sheiner, J. D. Bamberger, M. A. Chesney, and A. Moss. 2000. Adherence to protease inhibitors, HIV-1 viral load, and development of drug resistance in an indigent population. AIDS 14(4):357-366. Bangsberg, D. R., F. M. Hecht, E. D. Charlebois, M. Chesney, and A. Moss. 2001. Compar- ing objective measures of adherence to HIV antiretroviral therapy: Electronic medication monitors and unannounced pill counts. AIDS & Behavior 5(3):275-281. Bartley, J. B., D. G. Ferris, L. M. Allmond, E. D. Dickman, J. K. Dias, and J. Lambert. 2004. Personal digital assistants used to document compliance of bacterial vaginosis treatment. Sexually Transmitted Diseases 31(8):488-491. Berg, K. M., and J. H. Arnsten. 2006. Practical and conceptual challenges in measuring an- tiretroviral adherence. Journal of Acquired Immune Deficiency Syndromes 43(Suppl 1): S79-S87. Bova, C. A., K. P. Fennie, G. J. Knafl, K. D. Dieckhaus, E. Watrous, and A. B. Williams. 2005. Use of electronic monitoring devices to measure antiretroviral adherence: Practical con- siderations. AIDS & Behavior 9(1):103-110. Chesney, M. A. 2006. The elusive gold standard. Future perspectives for HIV adherence as- sessment and intervention. Journal of Acquired Immune Deficiency Syndromes 43(Suppl 1):S149-S155. Cramer, J. A., R. D. Scheyer, and R. H. Mattson. 1990. Compliance declines between clinic visits. Archives of Internal Medicine 150(7):1509-1510. Fairley, C. K., A. Permana, and T. R. Read. 2005. Long-term utility of measuring adherence by self-report compared with pharmacy record in a routine clinic setting. HIV Medicine 6(5):366-369. Fogarty, L., D. Roter, S. Larson, J. Burke, J. Gillespie, and R. Levy. 2002. Patient adherence to HIV medication regimens: A review of published and abstract reports. Patient Education and Counseling 46(2):93-108. Friedland, G. H. 2006. HIV medication adherence. The intersection of biomedical, behavioral, and social science research and clinical practice. Journal of Acquired Immune Deficiency Syndromes 43(Suppl 1):S3-S9. Gallo, M. F., F. M. Behets, M. J. Steiner, S. C. Thomsen, W. Ombidi, S. Luchters, C. Toroitich- Ruto, and M. M. Hobbs. 2007. Validity of self-reported “safe sex” among female sex workers in Mombasa, Kenya—PSA analysis. International Journal of STD and AIDS 18(1):33-38. Goetghebeur, E., and S. Vansteelandt. 2005. Structural mean models for compliance analysis in randomized clinical trials and the impact of errors on measures of exposure. Statistical Methods in Medical Research 14(4):397-415. Goggin, K., R. J. Liston, and J. A. Mitty. 2007. Modified directly observed therapy for antiret- roviral therapy: A primer from the field. Public Health Reports 122(4):472-481. Gordon, C. M. 2006. Commentary on meta-analysis of randomized controlled trials for HIV treatment adherence interventions. Research directions and implications for practice. Journal of Acquired Immune Deficiency Syndromes 43(Suppl 1):S36-S40. Grossberg, R., Y. Zhang, and R. Gross. 2004. A time-to-prescription-refill measure of antiret- roviral adherence predicted changes in viral load in HIV. Journal of Clinical Epidemiol- ogy 57(10):1107-1110. Haddad, M., C. Inch, R. H. Glazier, A. L. Wilkins, G. Urbshott, A. Bayoumi, and S. Rourke. 2000. Patient support and education for promoting adherence to highly active antiretro- viral therapy for HIV/AIDS. Cochrane Database of Systematic Reviews (3):CD001442. Haug, N. A. and J. L. Sorensen. 2006. Contingency management interventions for HIV-related behaviors. Current HIV/AIDS Reports 3:144-149.

144 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS Haynes, R. B., X. Yao, A. Degani, S. Kripalani, A. Garg, and H. P. McDonald. 2005. Inter- ventions for enhancing medication adherence [systematic review]. Cochrane Database of Systematic Reviews (4). Hays, M. A., B. Irsula, S. L. McMullen, and P. J. Feldblum. 2001. A comparison of three daily coital diary designs and a phone-in regimen. Contraception 63(3):159-166. Hogarty, K., A. Kasowitz, B. C. Herold, and M. J. Keller. 2007. Assessment of Product Ad- herence to Product Dosing in a Pilot Microbicide Study. Sexually Transmitted Diseases 34(11). Hufford, M. R. 2002. Paper vs. Electronic diaries: Compliance and subject evaluations. Ap- plied Clinical Trials April:38-43. Jadack, R., G. Willis, and S. M. Rogers. 2001. Accurate responding to sensitive questions in an STI clinic population: Patient preference for computer-based administration. Annual Meeting of the Society of Behavioral Medicine, Seattle, Washington. Kalichman, S. C., C. M. Amaral, H. Stearns, D. White, J. Flanagan, H. Pope, C. Cherry, D. Cain, L. Eaton, and M. O. Kalichman. 2007. Adherence to antiretroviral therapy as- sessed by unannounced pill counts conducted by telephone. Journal of General Internal Medicine 22(7):1003-1006. Kerr, T., J. Walsh, E. Lloyd-Smith, and E. Wood. 2005. Measuring adherence to highly active antiretroviral therapy: Implications for research and practice. Current HIV/AIDS Reports 2(4):200-205. Lawson, M. L., M. Maculuso, A. Bloom, G. Hortin, K. R. Hammond, and R. Blackwell. 1998. Objective markers of condom failure. Sexually Transmitted Diseases 25(8):427-432. Liu, H., C. E. Golin, L. G. Miller, R. D. Hays, C. K. Beck, S. Sanandaji, J. Christian, T. Maldonado, D. Duran, A. H. Kaplan, and N. S. Wenger. 2001. A comparison study of multiple measures of adherence to HIV protease inhibitors. Annals of Internal Medicine 134(10):968-977. Low-Beer, N., R. Gabe, S. McCormack, V. S. Kitchen, C. J. Lacey, and A. J. Nunn. 2002. Dextrin sulfate as a vaginal microbicide: Randomized, double-blind, placebo-controlled trial including healthy female volunteers and their male partners. Journal of Acquired Immune Deficiency Syndromes 31(4):391-398. Macaluso, M., J. Kelaghan, L. Artz, H. Austin, M. Fleenor, E. W. Hook, 3rd, and T. Valappil. 1999. Mechanical failure of the latex condom in a cohort of women at high STD risk. Sexually Transmitted Diseases 26(8):450-458. Mackinnon, D. P., and J. H. Dwyer. 1993. Estimating mediated effects in prevention studies. Evaluation Review 17(2):144-158. Maher, K., N. Klimas, M. A. Fletcher, V. Cohen, C. M. Maggio, J. Triplett, R. Valenzuela, and G. Dickinson. 1999. Disease progression, adherence, and response to protease inhibitor therapy for HIV infection in an urban veterans affairs medical center. Journal of Acquired Immune Deficiency Syndromes 22(4):358-363. Mannheimer, S. B., E. Morse, J. P. Matts, L. Andrews, C. Child, B. Schmetter, and G. H. Friedland. 2006. Sustained benefit from a long-term antiretroviral adherence in- tervention. Journal of Acquired Immune Deficiency Syndromes 43(1):S41-S47. Mauck, C. K., and G. F. Doncel. 2007. Biomarkers of semen in the vagina: Applications in clinical trials of contraception and prevention of sexually transmitted pathogens includ- ing HIV. Contraception 75(6):407-419. NIMH Collaborative HIV/STD Prevention Trial Group. 2007. The feasibility of audio computer- assisted self-interviewing in international settings. AIDS 21(Suppl 2):S49-S58. Osterberg, L., and T. Blaschke. 2005. Adherence to medication. New England Journal of Medicine 353(5):487-497.

DESIGN CONSIDERATIONS: ADHERENCE 145 Padian, N., A. van der Straten, G. Ramjee, T. Chipato, G. de Bruyn, K. Blanchard, S. Shiboski, E. Montgomery, H. Fancher, H. Cheng, M. Rosenblum, M. van der Loan, N. Jewell, J. McIntyre, and The MIRA Team. 2007. Diaphragm and lubricant gel for prevention of HIV acquisition in southern African women: A randomised controlled trial. Lancet 370:251-261. Pearson, C. R., M. A. Micek, J. A. Simoni, P. D. Hoff, E. Matediana, D. P. Martin, and S. S. Gloyd. 2007. Randomized control trial of peer-delivered, modified directly observed therapy for HAART in Mozambique. Journal of Acquired Immune Deficiency Syndromes 46(2):239-244. Pequegnat, W., M. Fishbein, D. Celentano, A. Ehrhardt, G. Garnett, D. Holtgrave, J. Jaccard, J. Schachter, and J. Zenilman. 2000. NIMH/APPC workgroup on behavioral and biologi- cal outcomes in HIV/STD prevention studies: A position statement. Sexually Transmitted Diseases 27(3):127-132. Petersen, M. L., Y. Wang, M. J. van der Laan, D. Guzman, E. Riley, and D. R. Bangsberg. 2007. Pillbox organizers are associated with improved adherence to HIV antiretroviral therapy and viral suppression: A marginal structural model analysis. Clinical Infectious Diseases 45(7):908-915. Pool, R., A. Kamali, and J. A. G. Whitworth. 2006. Understanding sexual behaviour change in rural southwest uganda: A multi-method study. AIDS Care—Psychological and Socio- Medical Aspects of AIDS/HIV 18(5):479-488. Pradier, C., L. Bentz, B. Spire, C. Tourette-Turgis, M. Morin, M. Souville, M. Rebillon, J. G. Fuzibet, A. Pesce, P. Dellamonica, and J. P. Moatti. 2003. Efficacy of an educational and counseling intervention on adherence to highly active antiretroviral therapy: French prospective controlled study. HIV Clinical Trials 4(2):121-131. Raymond, S., and R. Ross. 2000. Electronic subject diaries in clinical trials. Applied Clinical Trials (March):48-57. Remien, R. H., M. J. Stirratt, C. Dolezal, J. S. Dognin, G. J. Wagner, A. Carballo-Dieguez, N. El-Bassel, and T. M. Jung. 2005. Couple-focused support to improve HIV medication adherence: A randomized controlled trial. AIDS 19(8):807-814. Robins, J., and A. Tsiatis. 1991. Adjusting for differential rates of pneumocystis-carinii pneu- monia prophylaxis in high versus low-dose AZT treatment arms in an AIDS randomized trial. American Journal of Epidemiology 134. Rogers, S. M., G. Willis, A. Al-Tayyib, M. A. Villarroel, C. F. Turner, L. Ganapathi, J. Zenilman, and R. Jadack. 2005. Audio computer assisted interviewing to measure HIV risk behav- iours in a clinic population. Sexually Transmitted Infections 81(6):501-507. Rosen, M. I., K. Dieckhaus, T. J. McMahon, B. Valdes, N. M. Petry, J. Cramer, and B. Rounsaville. 2007. Improved adherence with contingency management. AIDS Patient Care and STDs 21(1):30-40. Rosenberg, Z. F. 2007. Phase III clinical trial design. Presentation at the first public meeting for the Committee on Methodological Challenges in HIV Prevention Trials, Washington, DC. Rueda, S., L. Park-Wyllie, A. M. Bayoumi, A. M. Tynan, T. A. Antoniou, S. B. Rourke, and R. H. Glazier. 2007. Patient support and education for promoting adherence to highly active antiretroviral therapy for HIV/AIDS. Cochrane Database of Systematic Reviews (3). Sankar, A., C. Golin, J. M. Simoni, M. Luborsky, and C. Pearson. 2006. How qualitative meth- ods contribute to understanding combination antiretroviral therapy adherence. Journal of Acquired Immune Deficiency Syndromes 43(Suppl 1):S54-S68. Schroder, K. E. E., M. P. Carey, and P. A. Vanable. 2003a. Methodological challenges in research on sexual risk behavior: I. Item content, scaling, and data analytical options. Annals of Behavioral Medicine 26(2):76-103.

146 METHODOLOGICAL CHALLENGES IN HIV PREVENTION TRIALS Schroder, K. E. E., M. P. Carey, and P. A. Vanable. 2003b. Methodological challenges in research on sexual risk behavior: II. Accuracy of self-reports. Annals of Behavioral Medicine 26(2):104-123. Schroder, K. E. E., C. J. Johnson, and J. S. Wiebe. 2007. Interactive voice response technol- ogy applied to sexual behavior self-reports: A comparison of three methods. AIDS and Behavior 11(2):313-323. Shrier, L. A., M. C. Shih, and W. R. Beardslee. 2005. Affect and sexual behavior in adoles- cents: A review of the literature and comparison of momentary sampling with diary and retrospective self-report methods of measurement. Pediatrics 115(5):e573-e581. Simoes, A. A., F. I. Bastos, R. I. Moreira, K. G. Lynch, and D. S. Metzger. 2006a. Accept- ability of audio computer-assisted self-interview (ACASI) among substance abusers seek- ing treatment in Rio de Janeiro, Brazil. Drug and Alcohol Dependence 82(Suppl 1): S103-S107. Simoes, A. A., F. I. Bastos, R. I. Moreira, K. G. Lynch, and D. S. Metzger. 2006b. A random- ized trial of audio computer and in-person interview to assess HIV risk among drug and alcohol users in Rio de Janeiro, Brazil. Journal of Substance Abuse Treatment 30(3):237-243. Simoni, J. M., P. A. Frick, D. W. Pantalone, and B. J. Turner. 2003. Antiretroviral adherence interventions: A review of current literature and ongoing studies. Topics in HIV Medicine 11(6):185-198. Simoni, J. M., C. R. Pearson, D. W. Pantalone, G. Marks, and N. Crepaz. 2006a. Efficacy of interventions in improving highly active antiretroviral therapy adherence and HIV-1 RNA viral load. A meta-analytic review of randomized controlled trials. Journal of Acquired Immune Deficiency Syndromes 43(Suppl 1):S23-S35. Simoni, J. M., A. E. Kurth, C. R. Pearson, D. W. Pantalone, J. O. Merrill, and P. A. Frick. 2006b. Self-report measures of antiretroviral therapy adherence: A review with recommendations for HIV research and clinical management. AIDS & Behavior 10(3):227-245. Simoni, J. M., D. W. Pantalone, M. D. Plummer, and B. Huang. 2007. A randomized con- trolled trial of a peer support intervention targeting antiretroviral medication adherence and depressive symptomatology in HIV-positive men and women. Health Psychology 26(4):488-495. Skoler, S. 2007. Population Council: Phase 3 efficacy study of the vaginal gel Carraguard to prevent HIV transmission. Paper read at the first public meeting for the Committee on Methodological Challenges in HIV Prevention Trials, Washington, DC. Smith, D. 2007. CDC PrEP trials. Presentation at the first public meeting for the Committee on Methodological Challenges in HIV Prevention Trials, Washington, DC. Tusa, M. G., M. Ladd, M. Kaur, R. Balkrishnan, and S. R. Feldman. 2006. Adapting electronic adherence monitors to standard packages of topical medications. Journal of American Academy of Dermatology 55:886-887. UNAIDS. 2007. Ethical considerations in biomedical HIV prevention trials. Geneva, Swit- zerland: UNAIDS. Urquhart, J. 1997. The electronic medication event monitor. Lessons for pharmacotherapy. Clinical Pharmacokinetics 32(5):345-356. Van Damme, L., G. Ramjee, M. Alary, B. Vuylsteke, V. Chandeying, H. Rees, P. Sirivongrangson, L. Mukenge-Tshibaka, V. Ettiegne-Traore, C. Uaheowitchai, S. S. Karim, B. Masse, J. Perriens, and M. Laga. 2002. Effectiveness of col-1492, a nonoxynol-9 vaginal gel, on HIV-1 transmission in female sex workers: A randomised controlled trial. Lancet 360(9338):971-977. Van De Wijgert, J., M. Mbizvo, S. Dube, M. Mwale, P. Nyamapfeni, and N. Padian. 2001. Intravaginal practices in Zimbabwe: Which women engage in them and why? Culture, Health & Sexuality 3(2):133-148.

DESIGN CONSIDERATIONS: ADHERENCE 147 Vandebosch, A., E. Goetghebeur, and L. Van Damme. 2005. Structural accelerated failure time models for the effects of observed exposures on repeated events in a clinical trial. Statistics in Medicine 24(7):1029-1046. Vrijens, B. 2007. Presentation at the second meeting for the Committee on Methodological Challenges in HIV Prevention Trials, London, UK. Vrijens, B., and J. Urquhart. 2005. Patient adherence to prescribed antimicrobial drug dosing regimens. Journal of Antimicrobial Chemotherapy 55(5):616-627. Vrijens, B., A. Belmans, K. Matthys, E. de Klerk, and E. Lesaffre. 2006. Effect of intervention through a pharmaceutical care program on patient adherence with prescribed once-daily atorvastatin. Pharmacoepidemiology and Drug Safety 15(2):115-121. Wallace, A., M. Thorn, R. A. Maguire, K. M. Sudol, and D. M. Phillips. 2004. Assay for establishing whether microbicide applicators have been exposed to the vagina. Sexually Transmitted Diseases 31(8):465-468. Ware, N. C., M. A. Wyatt, D. R. Bangsberg. 2006. Examining theoretic models of adherence for validity in resource limited settings: A heuristic approach. Journal of Acquired Im- mune Deficiency Syndromes 43(1):S18-S22. Williams, A. B., K. P. Fennie, C. A. Bova, J. D. Burgess, K. A. Danvers, and K. D. Dieckhaus. 2006. Home visits to improve adherence to highly active antiretroviral therapy: A ran- domized controlled trial. Journal of Acquired Immune Deficiency Syndromes 42(3): 314-321. Zenilman, J. M., J. Yuenger, N. Galai, C. F. Turner, and S. M. Rogers. 2005. Polymerase chain reaction detection of Y chromosome sequences in vaginal fluid: Preliminary studies of a potential biomarker for sexual behavior. Sexually Transmitted Diseases 32(2):90-94.

Next: 6 Design Considerations: Recruitment and Retention »
Methodological Challenges in Biomedical HIV Prevention Trials Get This Book
×
Buy Paperback | $66.00 Buy Ebook | $54.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

The number of people infected with HIV or living with AIDS is increasing at unprecedented rates as various scientists, organizations, and institutions search for innovative solutions to combating and preventing the disease. At the request of the Bill & Melinda Gates Foundation, Methodological Challenges in Biomedical HIV Prevention Trials addresses methodological challenges in late-stage nonvaccine biomedical HIV prevention trials with a specific focus on microbicide and pre-exposure prophylaxis trials. This book recommends a number of ways to improve the design, monitoring, and analysis of late-stage clinical trials that evaluate nonvaccine biomedical interventions. The objectives include identifying a beneficial method of intervention, enhancing quantification of the impact, properly assessing the effects of using such an intervention, and reducing biases that can lead to false positive trial results.

According to Methodological Challenges in Biomedical HIV Prevention Trials, the need to identify a range of effective, practical, and affordable preventive strategies is critical. Although a large number of promising new HIV prevention strategies and products are currently being tested in late-stage clinical trials, these trials face a myriad of methodological challenges that slow the pace of research and limit the ability to identify and fully evaluate effective biomedical interventions.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!