Exploring risk factors of non-adherence to immunosuppressive medication in kidney transplant recipients : improving methodology & reorienting research goals

8.1. Background and aim of the research program
Non-adherence to the immunosuppressive therapy is an important issue in kidney
transplant patients. About 20% of the kidney transplant patients are non-adherent to
the immunosuppressive regimen. Non-adherence contributes to 20% of late acute
rejection episodes and 16% of the graft losses, and results in a decreased number of
quality adjusted life years. A strategy to increase long-term successful outcome after
transplantation is to identify patients at risk for non-adherence and to target them for
preventive and adherence enhancing interventions. Comprehensive research on risk
factors of non-adherence addressing socio-economic, patient-, condition-, therapy-,
and health care system/worker-related factors is lacking. Especially health care-related
risk factors are understudied. Moreover, existing studies are hampered by a number of
methodological shortcomings. An important shortcoming is that accurate measurement
methods for detecting non-adherence, such as electronic monitoring (EM) are rarely
adopted. EM, currently the most sensitive adherence assessment method, uses
microchip technology to register date and time of openings of a pill bottle. Although
EM’s superior sensitivity to detect non-adherence makes it a potential gold standard of
adherence assessment, the lack of thorough validation as well as the lack of use of
appropriate statistical methods for multivariable and/or longitudinal data analysis of EM
data, hinder progress in the field.
The main purpose of this research program was to determine prevalence and risk
factors of non-adherence to immunosuppressive medication in kidney transplant
patients. As an additional purpose, we aimed to improve the validity of EM
measurement by mapping assumptions underlying correct EM measurement. We
tested these assumptions on adherence data of kidney transplant patients.
8.2. Methods
8.2.1. Prevalence and risk factors of non-adherence
To study prevalence and risk factors of non-adherence to immunosuppressives, we
conducted a prospective study, in which we measured adherence by EM over a 3-
month period in 250 adult renal transplant patients sampled from two outpatient
transplant centers in Switzerland. We calculated period prevalences of adherence and
expressed them as the percentage of prescribed doses taken (taking adherence), the
percentage of days with correct dosing (dosing adherence), the percentage of interdose
intervals not exceeding 25% of the prescribed interval (timing adherence), and
the number of drug holidays per 100 days (>48h no intake if once; >24h if twice daily
intake).
Selected risk factors were socio-economic, therapy related (e.g. number of
transplantations, use of medication aids, symptom occurrence and distress), patient
related (e.g. self-efficacy, health beliefs, coping styles, busyness, health behaviors),
condition related (e.g. depression, substance use), and health care system/team
related (e.g. regularity of follow up). Testing of the risk factors occurred by simple
mixed logistic regression analysis, using a sequence of daily binary adherence data.
Factors significant after correction for multiple testing were entered into a multiple
mixed logistic regression model.
Because the EM-study was not designed to extensively investigate health care system
or health care team-related factors, we performed an additional meta-analysis to look
whether non-adherence prevalences differed between continents/countries. This metaanalysis
on individual patient data pooled data from three studies in adult kidney
transplant patients from the US (n=1563), the Netherlands (n=85), Belgium (n=187)
and Switzerland (n=342). Adherence was measured by the Siegal scale, a self-report
instrument for measuring non-adherence to immunosuppressives. Patients were
categorized as non-adherent if they reported to have missed a dose of
immunosuppression in the last 4 weeks. Data were analysed using multiple mixed
logistic regression with center as a random effect and continent/country as fixed
effects, while controlling for several demographical and clinical characteristics of the
included samples.
8.2.2. Validation of EM assessment
To study the validity of the EM measurement, we summarized existing evidence on
processes that may bias non-adherence assessment. Unbiased EM assessment requires
fulfillment of four validity assumptions, being (1) correctly functioning EM equipment,
(2) correspondence of EM-bottle openings to the actual intake of the prescribed dose,
(3) absence of influence of EM on a patient’s normal adherence behavior, and (4)
sample representativeness.
We examined these four validity assumptions using the above mentioned sample of
250 kidney transplant patients whose adherence was measured by EM. More
specifically, we (1) determined the prevalence of non-functioning EM systems, (2)
examined the impact of patient-reported discrepancies between cap openings and
actual drug intakes on period prevalence, (3) explored whether non-adherence
increased over time after patients started EM, and (4) screened for differences
between participating patients and patients who refused to participate or who dropped
out of the study.
8.3. Results
Mean taking, dosing, timing adherence and drug holidays per 100 days were 98%,
96%, 93%, and 1.1 days, respectively. Variables associated with EM measured nonadherence
were: higher self-reported non-adherence (OR= 3.08; 95%CI: 1.69-5.61),
no usage of a pillbox (OR= 0.31; 95%CI: 0.16-0.61), male gender (OR= 0.46; 95%CI:
0.26-0.81), and lower self-efficacy (OR= 0.49; 95%CI: 0.22-1.07). Furthermore, a
gradually declining adherence could be observed between Monday and Sunday (OR=
1.04; 95%CI: 1.02-1.07).
The results of the meta-analysis examining self-reported non-adherence differences
between continents/countries showed that the prevalence of non-adherence to
immunosuppressives in the U.S. and Europe was 19.3% and 13.2.%, respectively The
higher prevalence of non-adherence in US patients was confirmed in the multiple
logistic regression analysis (OR=1.78; 95%CI: 1.10-2.89). Moreover, non-adherence
differed between Belgium (16%) and the Netherlands (14.1%) (OR=0.27; 95% CI:
0.09-0.80) and between Belgium and Switzerland (11.4%) (OR=0.17; 95% CI: 0.0-
0.42).
The validation study of EM showed that not all assumptions underlying EM
measurement were fulfilled: (1) one cap malfunctioned, (2) mismatches between bottle
openings and actual drug intake occurred in 62% of the patients (n=155), and (3)
non-adherence increased during the initial period of the monitoring, primarily during
the first 5 weeks, indicating EM had an intervention effect. The bias caused by this 5-
week intervention effect was minimal. The effect of mismatches between bottle
openings and actual drug intake on the measured adherence prevalence was larger,
but could be minimized by correcting the downloaded EM data using patient selfreports
(i.e., self-reported adherence to the EM guidelines and notes made by the
patient to correct mismatches between openings and ingestions).
8.4. Conclusions
This study program aimed to study risk factors of non-adherence in kidney transplant
patients. Its contribution to the literature lies in the fact that a comprehensive number
of non-adherence risk factors, including the currently neglected health care system
factors, have been explored, and in the fact that improvements of the methodological
approach for adherence studies have been proposed.
The profile of risk factor appearing in the final results suggest that forgetfulness was a
major driver of non-adherence. Moreover, system factors might also have an impact on
individual adherence behavior, as suggested by the found differences in prevalence of
non-adherence between European and US patients and among European patients.
These findings may change the focus of adherence research in the transplant
population.
Methodological improvements put forward throughout this study program primarily
concern the measurement of adherence behavior using EM. Novel statistical techniques
are proposed that allow multivariate analysis of EM data and inclusion of time-varying
variables into the statistical regression models. Besides, we showed that, although
assumptions underlying valid EM measurement may be violated, bias can to a certain
extent be prevented by correcting incorrect data or omitting them from the analysis.