A systematic review of the prevalence and determinants of nonadherence to phosphate binding medication in patients with end-stage renal disease

The most frequently assessed demographic predictors of phosphate binder adherence were age (24 studies), gender (22 studies), educational level (21 studies), marital status (11 studies), ethnicity (8 studies), income (6 studies) and employment status (3 studies). As shown in Figure 1, few studies found significant relationships between demographic factors and adherence to phosphate binding medication, with the exception of those exploring the impact of age on adherence, where 11 of the 24 studies (46%) exploring this variable found a significant result [12‐22]. In these studies, older age was consistently associated with higher levels of adherence. Suggested reasons for this finding are that older people may be more concerned about their mortality and have more structured lives in which to accommodate the demands of the treatment regimen [13], that younger patients may have more difficulty coming to terms with having a chronic condition [13] or simply that younger patients are more willing to report nonadherence than older patients [17].

The most frequently assessed clinical predictors of nonadherence were length of time on hemodialysis (22 studies), whether or not the patient was diabetic (9 studies) and the patient's transplant history (i.e. whether or not they had received a kidney transplant in the past; 6 studies). As shown in Figure 2, none of these clinical variables were consistently associated with adherence to phosphate binding medication. Given the large tablet burden associated with phosphate binding medication (patients on average take approximately eight phosphate binding tablets per day [3]), it is surprising that only three studies have explored the effects of regimen complexity on adherence [3, 17, 22], with two finding significant results. One found a correlation between complexity of the phosphate-binding medicine regimen and serum phosphorus [22] and one found that patients reporting low adherence and those above target phosphorus levels were prescribed significantly more phosphate binder pills per day [3]. Although high tablet burden may be a barrier to adherence for many patients, we cannot assume a causal relationship between high tablet burden and low adherence from these studies. One explanation is that high tablet burden leads to low adherence, but an equally plausible explanation is that low adherence results in poorer phosphate control and an increase in the number of prescribed tablets. A review across other therapeutic areas suggests that prescribed number of doses is inversely related to adherence [23] and this warrants further research in relation to phosphate binding medication. In addition, whilst qualitative and descriptive studies have indicated that the size and taste of the tablets may impact on adherence to phosphate binding medication [24, 25], none of the quantitative studies reviewed explored these variables. Further research is needed to determine the role of these tablet-related factors in predicting nonadherence.

Whilst psychosocial predictors of nonadherence were the least often assessed, they were more likely to be significantly associated with nonadherence to phosphate binding medication than demographic and clinical variables (see Figure 3).

Six of the nine studies investigating the relationship between health beliefs and adherence to phosphate binding medication reported significant relationships. These beliefs were all related to patients' perceptions of medication (e.g. concerns about potential adverse effects of medication [17], perceived barriers to and benefits of taking medication [22, 26], perceptions of self efficacy with regard to taking the medication [16, 27] and perceptions of others' expectations regarding adherence [28]). In a recent review of adherence to medication across chronic illnesses, such beliefs were identified as important potentially modifiable predictors of nonaderence that could be addressed within interventions to facilitate adherence [5].

Four of five studies found a relationship between social support and adherence to phosphate binding medication. This included support of friends and family [13, 29, 30] and of renal staff [22]. In addition, three of five studies exploring associations between family dynamics and adherence reported significant results. Family problems caused by the patient's illness [22], disorganisation and disagreements within the family [30] and lack of clear family structure [31] were associated with low adherence to phosphate binding medication. It is interesting that whilst marital status/living arrangements alone were not often associated with adherence, patients' perceptions of the actual support they received and the quality of their family relationships were more likely to be associated with adherence. This is consistent with findings in the broader social support literature that suggest that it is the quality rather than quantity of social support that is important in predicting mental and physical health outcomes [32].

Four of the eight studies exploring personality as a predictor of adherence to medication found significant results. Personality traits associated with nonadherence included low conscientiousness [14], high cynical hostility [33], and being expedient, venturesome, experimental and lacking self control [34].

Findings of studies looking at knowledge as a predictor of nonadherence were mixed. Two out of four studies found an association between knowledge of the purpose of the regimen and phosphate levels [20, 22]. However, the other two studies found no relationship between knowledge of treatment instructions and adherence to phosphate binding medication [26, 35]. Knowledge might be a prerequisite for adherence behaviour but the presence of knowledge alone may not bring about change in behaviour.

A variety of methods of assessing adherence were utilised in the studies, including tablet counts, electronic monitoring, patient self-report, health care professionals' reports and serum phosphorus levels. Each method has its own limitations, as discussed in a recent review of adherence, compliance and concordance [5]. Serum phosphorus was the most frequently used indicator of phosphate binder adherence (79% studies). This can be problematic as it reflects not only adherence to phosphate binding medication but also adherence to diet restrictions and dialysis attendance. It has also been suggested that serum phosphorus levels can be affected by 'residual renal function, urine output, co-morbid illnesses, hypercatabolism, nutritional status, hormonal and acid base status, type and intensity of dialytic treatment' [36], highlighting the lack of specificity of this measure. Where studies used more than one method of measuring adherence, rates of nonadherence and predictors of nonadherence varied depending on the adherence measure used [3, 22, 29]. This makes it very difficult to accurately estimate the levels of nonadherence in the renal dialysis population.

Definitions of nonadherence were inconsistent. Serum phosphorus levels that were considered acceptable ranged from 4.5 mg/dl [37] to 7.5 mg/dl [18] and this was reflected in the reported rates of nonadherence, with the study adopting the highest cut-off reporting the lowest rates of nonadherence (22%, [18]), and the study adopting the lowest cut-off reporting one of the highest rates of nonadherence (68%, [37]). Similarly, there was variation in the level of adherence that was considered acceptable in studies using self report measures of nonadherence, with definitions of nonadherence ranging from 'ever missing a dose' [38] to 'missing more than 20% of doses' [3]. More research is necessary to determine the level of adherence to phosphate binding medication required to prevent negative health outcomes.

Eight studies combined adherence to phosphate binding medication with adherence to other parts of the treatment regimen (e.g. attendance at dialysis, adherence to diet and fluid restrictions) for the analysis [28‐31, 34, 39‐41]. People may have different levels of adherence for different parts of the treatment regimen and therefore adherence to the individual components should ideally be considered in isolation. Indeed, studies that did assess adherence to different parts of the regimen separately not only reported different levels of adherence to the different aspects of treatment but also found that different factors predicted adherence to different parts of the regimen [12‐15, 18, 21, 22, 26, 27, 33, 35, 42‐48].

Only three studies utilised a prospective design [14, 27, 28], with the remainder using a cross sectional study design. Whilst cross-sectional studies enable the identification of associations between variables, prospective studies are required to determine causal links between potential predictor variables and adherence.

Many studies had small sample sizes, with a third including less than 50 people and 6 studies (18%) reporting sample sizes of 25 or less. Only one study included a power calculation [17] and it is likely that many of the other studies would not have had the power to detect predictors of nonadherence. It is therefore possible that actual predictors of nonadherence remain undetected. Future research should ensure sample sizes are large enough for the analysis to identify significant predictors of nonadherence, should they exist.

The vast majority of the studies were conducted in the United States of America (79%). It is possible that the health care system in the United States has unique characteristics that could influence adherence (e.g. prescription charges, private health insurance). It is therefore not possible to generalise the results to all health care systems and there is a need for further research outside of the United States.