By reviewing the literature on adherence rates in Cameroon, and analysing a new data set we have produced an explanatory account of the trends in adherence observed and the reported determinants. We also identified another plausible means of enhancing adherence to ART: using multiple reminder methods.
Despite the different study designs, locations and sample sizes there is a clear improvement in adherence rates over time irrespective of how it is measured. This finding confirms and supports health policies that reduce the cost of care to improve access and use. This is also in line with studies reporting that the main hindrances to adherence in Africa are related to health system weaknesses such as inadequate supply, human resource shortages and poor infrastructure [29, 32].
The determinants of adherence to ART identified in the Cameroonian literature are a subset of factors identified elsewhere . However, there are discrepancies as to the role of monthly income and duration on ART on adherence. Boyer et al.  reported better adherence with a higher monthly income, while Rougemont et al.  reported the contrary. Both studies were initiated in the same period (2006) and therefore the cost of medication is unlikely to be the reason for this discrepancy. On the other hand Boyer et al. conducted a multisite (27 sites) study with a larger sample size (n=3151) compared to Rougemont et al. (n=312) in a single site. The results from Boyer et al. may be more plausible in the Cameroonian context where, apart from free ART, other related services are still funded by out-of-pocket payments and having to pay for care has been identified as a cause of non-adherence [25, 28, 29, 32]. The larger sample is also more likely to have more accurate and generalizable data, especially since one of the sites included in Boyer et al. is the same site where the Rougemont study was conducted.
Roux et al.  noted a drop in adherence rates with increased duration on ART while Freeman et al. and Mbopi-Keou et al. [33, 35] noted the contrary. However, these two samples differ greatly. Roux et al. describe a multisite study in the Centre region of Cameroon, in 401 PLHIV over two years. Freeman et al. describe a cross-national cohort of 8419 women from Cameroon, Burundi and the Democratic Republic of Congo followed up for two years. Female gender has been reported as a factor for better adherence  and this discrepancy may explain the interaction between gender and duration on medication. Mbopi-Keou et al. suggest that this trend might be the effect of continuous psycho-social support .
One study reported higher levels of adherence for patients who initiated ART with high levels of CD4-positive-T-lymphocytes . This finding is in favor of earlier initiation of ART. Why patients who were transferred to the clinic would have better adherence rates is unclear. However, if they left their previous clinics due to service provision issues or unsatisfactory interactions with clinic staff, the latter clinic may provide a more favorable environment for adherence.
Using multivariable analysis we identified gender, education, side effects experienced, and number of reminder methods as factors that affect adherence rates.
The male gender has often been reported as the most likely to be non-adherent, maybe because males are more likely to engage in other behaviours that influence adherence such as binge drinking, tobacco use and drug use . This effect was not found in all models and is not consistent across studies .
Level of education seems to play a role in adherence behaviour. A significant difference in adherence exists between people with secondary and those with no education. Further increments in educational level show no effect. This may imply that as concerns adherence to ART there are no benefits to be gained from very high levels of education. The benefits of education on adherence to ART can be obtained from secondary education. This effect was not consistent across all models.
Our findings regarding side effects are contrary to what is reported in literature. People who experienced more side-effects were more likely to be adherent. No other Cameroonian studies have identified side effects as a determinant for adherence to ART. In the CAMPS trial only 1% of those who missed doses reported side effects to be the reason for not taking medication (Table 3).
Only the number of reminder methods was associated with adherence in all the models. Another Cameroonian study reported the use of reminder methods to be associated with better adherence . This is the first study to show that multiple reminders may have a cumulative effect. A randomized clinical trial in Kenya found alarms to have no effect on virological outcomes . Simple electronic alarms are not as complex as the reminder methods described in this population (personal verbal reminders by individuals, phone alarms, meal times, timing with TV shows and watches). Other studies have reported the use of mobile phone beeps and prayer times by Muslims to remind them of when to take their medication . The use of multiple and varied reminder methods may address other causes of non-adherence like forgetfulness and lack of social support.
The difference in the four models suggests that the factors associated with adherence depend somewhat on how adherence is defined and the thresholds for acceptable adherence. Different levels of adherence are associated with different factors. It is unclear how this issue can be resolved without a uniform and validated tool for measuring adherence rates in clinical practice or research, as all methods have their advantages and flaws [16, 45]. Self reported adherence is by far the most popular method used, but how it is used varies greatly [16, 46]. On the one hand, a simple visual analogue scale can be used to situate a patient’s adherence; and on the other a series of questions related to number of pills, timing, missed doses and identification of pills, all fall under the canopy of self report.
In low resource settings, self report, attendance-based and dispensing-based adherence measuring methods can predict important clinical outcomes  and should be collected routinely. Some measure of drug availability or the occurrence of drug stock-outs should be documented to explain trends in adherence behaviours.
Both sections of this paper may have limitations. Even though we observed an increasing trend in adherence as the cost of ART reduces, other temporal factors may contribute to this trend, notably health system improvements over the years and reductions in stigma and discrimination. These factors are reported as potential threats to optimal adherence and are very likely to change over time. The availability and use of reminder methods like mobile phones may also enhance adherence over time [25, 32]. Participants in this study (CAMPS) may not adequately represent all the people living with HIV, but a subgroup who are already on ART, and who own mobile phones. Even though mobile phone ownership is widespread in Cameroon, those who own them may differ significantly from those who don’t.