Defining adherence

Adherence can be narrowly defined as the “extent to which patients take medications as prescribed by their health care providers” [16], or more broadly as the “the extent to which a person’s behaviour—taking medication, following a diet, or executing lifestyle changes, corresponds with agreed recommendations from a health care provider” [1]. The former definition refers to compliance with medication as prescribed by the provider while the latter goes beyond medication to include all the recommendations jointly agreed upon by both provider and client to improve health care. The latter definition also highlights or acknowledges additional factors such as diet or lifestyle which may affect adherence. It also recognises that the client may not agree with the recommendations of the health care provider. Such factors are even more relevant in HIV where dietary restrictions are required for the uptake of certain medications [17, 18] and that certain lifestyles (e.g. substance abuse) can lead to non- adherence to ART [19]. In this paper, all degrees of adherence which are not optimal are referred to as non-adherence.

The factors associated with adherence to ART exist in many categories, defined in literature [1, 16, 19]:

Patient factors such as substance abuse, being male (i.e. gender), depression, lower levels of education, lack of self efficacy, extreme anxiety, extreme pain, no change in health status despite ART and non-white race are significantly associated with non-adherence [1, 19].

Medication factors like dose frequency, pill burden, type of drug, inability to take medication when away from home, food requirements, side effects are also responsible for less than optimal adherence [1, 19]. In other words the complexity of the regimen and its side-effects are associated with non-adherence.

Provider related factors such as a poor patient-health care provider relationship can affect the patient’s overall satisfaction and trust in the provider. The quality of these relationships is significantly associated with better adherence [1, 20, 21].

Disease characteristics, notably stage and duration of HIV infection, symptoms experienced and the presence of opportunistic infections play a role in adherence to ART. HIV related symptoms like nausea may impede a patient from swallowing pills. Some studies report that patients who have experienced an opportunistic infection tend to be more adherent than those who have not [1, 22, 23].

Clinical setting and health system factors may influence use of services and adherence [19].

Adherence to ART in Cameroon

Cameroon has one of the highest rates of HIV in west and central Africa, with 5.3% of the adult population living with HIV. Of these, only 30% are receiving ART [24]. Adherence to ART in Cameroon has changed over the years for a multitude of reasons. Health system changes like the decentralization of HIV treatment centres and the subsidy of costs of medication and testing over the years have contributed to the variations in the reported adherence rates. A substantial reduction in costs occurred in 2004 [25] and ART has been provided for free since 2007 [26]. As a result, the designs and the contexts of the studies reporting rates of adherence to ART in Cameroon also vary. The set of similar and varying contextual factors in Cameroon make it a worthwhile case study to examine adherence to ART and learn lessons that could be applied across the country. Adherence rates in Cameroon have been measured using patient reports, pharmacy refill data, attendance at scheduled visits and blood testing [25–35]. These different methods have spawned adherence rates varying from 10.1% to 97.5%, between 2000 and 2010. Faced with these enormous discrepancies, we sought to use data from a recent trial [36] to describe adherence to ART and the associated determining factors. Adherence rates are not only a predictor of treatment success, but also an indication of how many resources should be invested in adherence enhancement research and practice. This cannot be done without stable estimates for adherence or a comprehensive list of locally relevant determinants.

Additionally, in a changing world where new methods and technologies arise to improve adherence rates, and levels of stigma are declining, there may be a shift in patients’ attitudes and practices related to adherence.

The objectives of this paper are:

Literature review of ART adherence rates in Cameroon

We searched electronic databases (PubMed, Google Scholar, Web of Science, CINAHL, EMBASE and PSYCINFO) for publications on adherence to ART in Cameroon (from January 1999 to May 2012) using appropriate combinations of keywords such as Cameroon, Cameroun (French), adherence, compliance, antiretroviral therapy, antiretroviral treatment and HIV. ART was introduced in Cameroon in 1999 [37]. No language restrictions were set. Citations were screened and full text obtained if they reported any measure of adherence to ART in Cameroonian subjects. Completeness and adequacy of reporting in the selected studies were assessed in duplicate by noting their compliance to the 22-item Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist [38]. Data were also extracted in duplicate: citation details; year study was conducted, city in which study was conducted, adherence rate, method of measure of adherence, study design, sample size and the cost of ART at the time. Discrepancies were resolved by discussion. We examined the evolution of adherence rates taking into account the above characteristics. We also identified the reported factors associated with adherence and classified them as patient variables, medication variables, patient-provider relationship and health system characteristics. These categories have previously been identified in literature [16, 19]. No statistical pooling was performed. The factors associated with adherence were extracted only if they resulted from multivariable analyses. The reported odds ratios and p-values were used to confirm the associations reported.

Adherence to ART in CAMPS

We conducted a cross-sectional analytical study based on the baseline data for the Cameroon Mobile Phone SMS (CAMPS) trial [36]. Data were collected from a subset of patients (eligible for the trial) aged 21 or older, who owned mobile phones and who agreed to take part in the trial. The independent variables included socio-demographics like age (in years), gender (male or female), level of education (none and primary or secondary and university), disclosure of status to family members (family aware or not aware of HIV status); clinical features like presence or absence of an opportunistic infection, Centers for Disease Control and Prevention (CDC; [39]) classification (categorised as AIDS defining condition or not), regimen (first line or second line), duration on treatment (in months); the number of drug side-effects experienced and the number of recall methods used. All the independent variables were those reported in literature and collected for the CAMPS trial [36]. The dependent variables were adherence measured using a visual analogue scale (0–100), self report on the number of missed doses and whether the doses were taken on time in the week preceding the interview. We also designed a composite measure of adherence based on all three measures, which we dichotomized as adherent (VAS=100; number of missed doses=0; all doses taken on time=yes) and non-adherent (VAS<100; missed doses >0; all doses taken on time=no).

We performed multivariable logistic regression with adherence (dependent variable) dichotomized at the VAS cut-offs of 90 and 95% (Model I and II). These cut-off points represent the levels of adherence at which sustained viral suppression can occur [5], and those often used in literature [40, 41]. We also analyzed the data using the self reports of adherence as the dependent variable, in which case one or more missed doses was considered as non-adherence and no missed dose was considered as good adherence (Model III). We repeated these analyses using the composite measure of adherence (Model IV). The interaction between independent variables was investigated based on associations reported in the general literature and specifically for Cameroon. Variables were introduced into the model in blocks defined by the following categories: socio-demographic, disease-related and medication-related. Model fit was assessed using the Hosmer and Lemeshow goodness of fit statistic [42].

Data were analysed using Predictive Analytic Software (PASW) Version 17.0 (SPSS, Inc., 2009, Chicago, IL, USA). Statistical significance was set at alpha= 0.05. Adjusted odds ratios (AOR), 95% confidence intervals (CI) and p-values are presented. The forest plot was drawn using Stata Statistical Software (STATA) Release 12 (StataCorp LP, 2011, College Station, TX, USA).

Literature review

We identified 9 studies conducted between 2000 and 2010 which reported adherence rates in Cameroon. The mean number of reported items on the STROBE checklist was 19.7 (standard deviation [SD] 1.38). Data from the CAMPS trial was included in the analysis [36]. Five of them were conducted in Yaoundé, the capital city of Cameroon. The characteristics of these studies are reported in Table 1. We observed an increasing trend in adherence rates over time, which corresponds to the reductions in cost of ART (Figure 1). Two studies were outliers to this quasi-linear relationship. The first was Kouanfack et al. [27], which reported high adherence rates (88.7% and 97.5%, using biological markers and self report respectively). This study was a drug trial and participants received medication free of charge at a time when medication costed 51.2 USD/month. The second, Mosoko et al. [25], reported very low adherence rates (10.1% using hospital records of the number of scheduled clinic visits attended). These data may be affected by inaccuracies in hospital records (incompleteness was handled using imputation techniques); participants who lived far away were sometimes given medication for more than one month; patients who lived in the nearby city of Douala had other ART opportunities and this measure (scheduled visits) considers all deaths and lost to follow-up as non-adherent. Ideally, the cases of death should be excluded from the denominator. The other study using attendance of scheduled visits [29] also reported low adherence rates (<50%).

Reference	Year	City	Study design (sample size)	Adherence rate	Method	Cost of ART/month for that year in USD*
Laurent et al.	2000-2003	Douala	Cross-sectional retrospective (788)	<50% μ	Scheduled visits attended	128
Kouanfack et al.	2002-2003	Yaoundé	Prospective cohort (60)	88.7%, (97.5%)	Biological markers, (Self Report)	51.2 (free for participants)
Mosoko et al.	2002-2005	Limbe	Cross-sectional retrospective α (2920)	10.1%	Scheduled visits attended	51.2
Boyer et al.[28]	2006	Yaoundé	Cross-sectional (532)	56.6%	Self Report β	12.8
Rougemont et al.	2006	Yaoundé	Prospective cohort (312)	64% (78%)	Pharmacy data (Self report)	12.8
Boyer et al.[32]	2006-2007	Multiple locations	Cross-sectional (3151)	53.9%	Self report β	12.8
Roux et al.	2006-2008	Yaoundé	Prospective cohort (401)	61-73%	Self report β	12.8
Mahy et al.	2007-2008	Tokombere	Cross-sectional (56)	95% (80%)	self report β (Pharmacy data)	0
Mbopi-Keou et al.	2010	Dschang	Cross-sectional (356)	80.2% (48.7)	Self report β (Pharmacy data)	0
CAMPS	2010	Yaoundé	Cross-sectional (200)	90.5%	Self report β	0

*Costs vary per regimen, the highest costs are reported.

^α Cross- sectional analysis of previously collected data without planning for needs of an investigation.

^β All methods in which all information used to measure adherence was obtained by interviewing the participant with either one question or a series of questions.

^μ 50% was imputed to represent the trend graphically.

Determinants of, and factors associated with ART adherence

These factors were found to reduce adherence to ART: binge drinking, drug use, tobacco use, lack of family support, experience of stigma, switching regimen, advanced stage of disease, cost of care, distance from clinic, large hospital size and no task shifting from physician to other staff [25, 28, 30–35].

These factors increased adherence to ART: Female gender, age greater than 49 years, higher levels of education, positive perceptions of treatment, high motivation, using reminder methods, satisfaction with information provided by physician, higher CD4 count at initiation of ART, and being transferred-in from another clinic [30, 31, 33–35].

High monthly income [31, 32] and increased duration on medication [30, 33, 35] were both reported to either increase or reduce adherence.

Factors associated with adherence to ART at baseline in the CAMPS Trial

In the first model, we dichotomized adherence using the VAS at 95%. Using multiple recall methods was significantly associated with adherence >95% (AOR 6.7, 95% CI 2.69, 16.56; p<0.001). In our second model, adherence was dichotomized at 90%. Female gender was associated with adherence >90% (AOR 0.28, 95% CI 0.09, 0.90; p=0.032), while secondary education (AOR, 4.4 95% CI 1.64, 11.92; p= 0.003) and multiple recall methods (AOR, 7.89 95% CI 3.22, 19.33; p<0.001) were associated with adherence >90%. In the third model we used the number of missed doses as our measure of adherence. Experiencing more side effects (AOR 2.25, 95% CI 1.13, 4.50; p= 0.021) and using multiple recall methods increased the odds for not missing doses (AOR 7.28, 95% CI 3.06, 17.32; p<0.001). In our last model we used a composite measure of adherence that incorporated the VAS, number of missed doses and timing of doses. Only the use of multiple recall methods was associated with adherence (AOR 4.11, 95% CI 1.89, 8.91; p<0.001). The Hosmer and Lemeshow goodness-of-fit statistic for all the four models had p-values greater than 0.05 implying a good fit [42]. These models are displayed in Figure 2.