Among kidney transplant recipients, girls and young women have been shown to have a higher graft failure risk than boys and young men, whereas older women have a similar or lower risk compared with older men.1,2 The reasons for these relationships are not clear. Although gender hormones may play a role in this relationship, with estrogens enhancing and androgens suppressing immune activation,3,4 another possibility is that medication adherence differs by gender, and that gender differences in adherence vary by age. Several prior studies of kidney transplant recipients reported associations between gender and medication adherence. However, none specifically addressed the question of whether medication adherence differs by gender. Furthermore, past studies were restricted to either children or to adults; no prior study focused on adolescents and young adults. Some studies of adult kidney transplant recipients showed better adherence in women than in men,5-9 whereas others showed no difference by gender.10-13 In contrast, most studies of pediatric kidney transplant recipients showed no difference in adherence between boys and girls; this may reflect the fact that parents take responsibility for medication adherence in children.14-16 Adolescence and young adulthood may represent a period distinct from other ages with respect to gender differences in adherence. As young people take greater responsibility for their own care throughout adolescence, the differing rates of cognitive maturation in boys versus girls17,18 may influence gender differences in adherence.19
The primary aim of this study was to determine whether medication adherence, as measured using electronic monitoring, differs by gender among adolescent and young adult kidney transplant recipients, and whether gender differences vary by age. We hypothesized that adherence would be similar for boys and girls aged 11 to 16 years, but that women would have higher adherence than men among those aged 17 to 24 years. We also considered gender differences in medication adherence measured by self-report and by the SD of tacrolimus trough levels. We expected that gender differences in adherence would be similar regardless of the method used to assess adherence.
MATERIALS AND METHODS
This study is a secondary analysis of data from the Teen Adherence in Kidney transplant Effectiveness of Intervention Trial (TAKE-IT),20,21 which was a multicenter, prospective randomized trial of an adherence-promoting intervention for adolescent and young adult kidney transplant recipients. The present study is an observational cohort study, focusing exclusively on the 3-month run-in period between the enrollment visit and initiation of the intervention. Neither study personnel nor participants were aware of group assignment (intervention or control) during the run-in. The intervention interval was not included in this study because the impact of the intervention on adherence may have differed by gender, and we had insufficient numbers of participants to consider a gender by age by intervention interaction.
TAKE-IT was approved by the research ethics boards of all sites. Written informed consent was obtained from all participants and parents (for those <18 y).
Prevalent kidney-only transplant recipients aged 11 to 24 years, 3 months or longer posttransplant, who had a functioning graft and were expected to be followed up in 1 of the 8 participating centers in Canada and the United States for the 15-month study were eligible for TAKE-IT. Exclusion criteria included impending graft failure, severe neurocognitive disabilities, lack of electronic pillbox connectivity, use of liquid immunosuppressive medications, having a sibling participating in the study, participating in another adherence-promoting intervention study, or inability to communicate comfortably in English (or French-Montreal site only).
Electronic Adherence Assessment
All participants were given an electronic multidose pillbox in which to store all immunosuppressive medications. During the first 4–6 months of recruitment, participants received a Medminder pillbox (Medminder, Needham, MA). However, due to technical difficulties with this device for some participants, subsequent participants were given a SimpleMed device (Vaica Medical, Tel Aviv, Israel); the Medminder and SimpleMed were similar, with the same types of adherence tracking and reminder functions. The prescribed immunosuppressive medication dosing times were recorded in a web-based pillbox record for each participant. The date and time of each pillbox compartment opening was registered in the electronic record of the patient to whom the device was assigned. Participants were instructed to inform study staff if they were not using the pillbox for a period (ie, weekend or vacation travel).
Other Adherence Measures
We also described tacrolimus trough levels variability and self-reported adherence. The SDs of all tacrolimus trough levels done for clinical care (except during hospitalizations or illnesses) during a 6-month period (including the 3 months before enrollment and the 3-month run-in period) were calculated for participants with 3 or more tacrolimus levels. Self-reported adherence was measured using the Medical Adherence Measure-Medication Module (MAM-MM)22,23 which is a semistructured interview, developed in children and adolescents with end-stage kidney disease including kidney transplantation, and administered by a trained interviewer at the enrollment visit and at the end of the run-in. MAM-MM captures adherence during the previous week.
Participant gender was the primary exposure. We considered the possibility that gender differences in adherence may be modified by age by including a gender by age interaction term.
The primary outcomes were daily “taking adherence” (proportion of prescribed doses taken) and “timing adherence” (proportion of prescribed doses taken within 1 h before to 2 h after the prescribed dosing time), as measured using electronic monitoring. A taking adherence score and a timing adherence score were calculated for each day of the study for each participant. Therefore, each participant had repeated outcome measures. Each score was either 0%, 50%, or 100%, because immunosuppressives were prescribed a maximum of 2 times per day. No score was calculated for days that the pillbox was not in use (turned off, not communicating with the server, or participant-reported nonuse). The first 2 weeks of electronic adherence data following enrollment were deleted to allow adaptation to the pillbox.24
Association Between Gender and Adherence
We used ordinal logistic regression with generalized estimating equations (proportional odds model) to estimate the association between gender and each of taking and timing adherence.21 This approach accounts for correlation between the repeated adherence scores for each day of observation within each participant. The assumption of proportional odds was assessed graphically. To consider effect modification by age, we included an age by gender interaction term. Age was categorized as 11 to 16 years versus 17 to 24 years based on assessment of plots of adherence versus age (in 1-year intervals), showing a clear gender difference starting at approximately 17 years of age. Figure S1, SDC, (http://links.lww.com/TP/B600) shows the results of a model, including gender, age, and the age by gender interaction in which age was treated as a continuous variable. This model showed odds ratios (OR) of smoothly increasing magnitude with increasing age in the comparison between girls/women and boys/men. However, the confidence intervals at older ages were very wide; therefore, we opted to dichotomize age. To identify potential confounders in the association between gender and adherence, we assessed univariate associations between baseline participant characteristics and adherence (Table S1, SDC,http://links.lww.com/TP/B600) as well as standardized differences25 in baseline participant characteristics by gender (and age category) (Table S2, SDC,http://links.lww.com/TP/B600). Only those variables associated with both gender (the exposure) and adherence (the outcome) were considered potential confounders. Covariates were selected based on the magnitude of the associations with adherence and gender rather than on statistical significance of the associations. Given the small sample size, important potential confounders may not have a statistically significant association with the exposure or outcome. Covariates associated with adherence with an OR <0.9 or greater than 1.1 and for which the absolute value of the standardized difference by gender was >0.3 in at least 1 age category (ie, substantial imbalance by gender) were included in multivariate models. For highly correlated characteristics, we selected only 1 to avoid collinearity.
The SD of tacrolimus trough levels was determined for each participant with 3 or more levels available. Using the MAM-MM, self-reported taking adherence was scored as the proportion of prescribed doses taken in the previous week; self-report timing adherence was scored as the proportion of prescribed doses taken up to 2 hours after the prescribed time in the previous week. The MAM-MM scores for each patient were summarized as the mean of the 2 scores determined during the run-in. Analyses were conducted using SAS 9.4.
The characteristics of the 83 boys and men (sixty-one 11–16 y; twenty-two 17–24 y at baseline) and 53 girls and women (thirty-five 11–16 y; eighteen 17–24 y at baseline) with electronic data available for analysis are summarized in Table 1. Both boys/men and girls/women were followed for a median of 2.5 months (male interquartile range [IQR], 1.8–2.7; female IQR, 1.9–2.8) in the run-in (after excluding the first 2 wk). Among the 136 participants, 68% were white. The median (IQR) time since transplant at baseline was 2.9 (0.8–7.1) years, with a median (IQR) age at transplant of 12.0 (8.3–14.9) years. Fifty-two percent of participants had a living donor. Concerning treatment characteristics, 65% of participants received 3 immunosuppressive medications per day; 92% took 2 doses of immunosuppressives per day. Detailed participant characteristics by gender within each age categories are presented in Table 1.
Association Between Gender and Adherence
Among participants 11–16 years, the median (IQR) overall taking adherence during the run-in was 85.6% (75.7–95.5) in girls and 88.3% (71.6–95.7) in boys; the median (IQR) overall timing adherence was 81.2% (69.1–92.0) in girls and 81.0% (62.5–92.7) in boys. Among those aged 17 to 24 years, overall taking adherence was 91.6% (76.0–95.7) in women and 70.0% (56.3–88.3) in men; the median (IQR) overall timing adherence was 87.7% (63.6–91.3) in women and 59.1% (49.5–79.4) in men. Figure 1 shows the proportion of boys/men and girls/women with 100% taking adherence on each day of follow-up, stratified by age category. Figure 2 shows the proportion of boys/men and girls/women with 100% timing adherence on each day of follow-up, stratified by age.
Table 2 shows the results of the ordinal logistic regression models used to estimate the association between gender and each of taking and timing adherence, accounting for age. Race, healthcare insurer, dialysis duration, primary disease, and number of immunosuppressive medications were identified as potential confounders. We did not adjust for number of prior transplants or number of past acute rejections because both may be markers of poor adherence. Because of the small numbers of patients, it was not possible to include all variables in a single multivariate model. Therefore, we present 3 models with different combinations of variables. Results were similar across all models. There were significant interactions between gender and age in both the taking (model 2 P = 0.019) and timing (model 2 P = 0.021) adherence models. Among participants aged 11 to 16 years, there were no significant gender differences in taking or timing adherence. In contrast, among those aged 17 to 24 years, women had significantly greater odds of higher taking adherence scores (model 2 OR, 3.06; 95% confidence interval [CI], 1.35–6.95) than men. Women aged 17 to 24 years also had significantly greater odds of higher timing adherence scores (model 2 OR, 3.06; 95% CI, 1.48–6.35) than men of the same age.
When we compared adherence by age category, stratified on gender, we found that, compared with boys aged 11 to 16 years, men aged 17 to 24 years had significantly lower odds of taking prescribed medication on time (model 2 OR, 0.46; 95% CI, 0.27–0.78). Results were similar but not significant for taking adherence. Girls/women showed no significant difference by age category for either taking or timing adherence.
In the multivariate models, only race and dialysis duration were significantly associated with taking and timing adherence (Table 3). Black participants had lower taking (OR, 0.48; 95% CI, 0.28–0.95) and timing (OR, 0.51; 95% CI, 0.31–0.84) adherence scores compared with white participants. Participants without dialysis before their current transplant had higher taking (OR, 1.71; 95% CI, 1.09–2.66) and timing (OR, 1.80; 95% CI, 1.21–2.68) adherence scores compared with participants with dialysis.
To consider a potential confounding effect of participants living with versus without a parent on the relationship between gender and adherence, we repeated the analysis on a sample restricted to the 127 participants not living on their own (Table S3, SDC,http://links.lww.com/TP/B600). Among participants aged 11–16 years, there were no significant gender differences in taking or timing adherence. Among participants aged 17–24 years, results were similar to the primary analyses, with women having significantly greater odds of higher taking and timing adherence scores than men.
Secondary Adherence Outcomes
Self-reported taking and timing adherence scores were similar across both gender and age categories among the 134 patients with self-reported adherence data (Table 4).
For the vast majority, self-reported adherence was higher than adherence measured electronically during the same week as the reporting period, but the magnitude of the difference between measurement methods differed by age and gender. Among those aged 11 to 16 years, the median (IQR) absolute difference between self-reported and electronic taking adherence was 14.3% (3.6–21.4) for girls and 7.1% (0.0–21.4) for boys. Similarly, the median (IQR) absolute difference between self-reported and electronic timing adherence was 14.3% (0.0–23.2) for girls and 7.1% (0.0–25.0) for boys aged 11 to 16 years. In contrast, among those aged 17 to 24 years, the median (IQR) absolute difference between self-reported and electronically-measured adherence was larger for men (taking 17.9% [1.8–42.9]; timing 21.4% [1.8–48.2]) than women (taking 7.1% [0.0–15.7]; timing 7.1% [−1.8 to 20.5]).
Variability in Tacrolimus Trough Levels
Among the 136 patients with electronic monitoring data in the run-in period, there were 100 with at least 3 tacrolimus trough blood levels (median number of levels [IQR] = 8.0 [5.0–16.0], similar across gender and age categories), allowing calculation of an SD. In participants aged 11 to 16 years taking twice per day tacrolimus (Prograf), the median (IQR) tacrolimus level SD was 1.6 (0.9–2.2) for boys and 1.5 (0.8–2.2) for girls (Table 4). Among participants 17–24 years, the median (IQR) tacrolimus level SD was 1.7 (1.0–2.3) for men and 1.8 (1.1–3.3) for women. Tacrolimus SD was >2.0 in 30% of girls and 36% of boys aged 11 to 16 years, and in 50% of women and 33% of men aged 17 to 24 years. The number of patients taking once per day tacrolimus (Advagraf) was insufficient (n = 14) to allow comparisons by gender and age.
Among adolescent and young adult kidney transplant recipients, we found a significant impact of age on the association between gender and electronically measured adherence. Whereas there were no gender differences in adherence measured by electronic monitoring among those aged 11 to 16 years, both taking and timing adherence were higher in women than men aged 17 to 24 years. These results are consistent with previous studies showing no difference in adherence by gender in pediatric kidney transplant recipients14,15 and better adherence in adult women than men.5,8 Our results underline the important modifying effect of age on gender differences in adherence.
There are several possible explanations for our findings. The lack of gender differences in younger adolescents likely reflects the fact that parents take some or all responsibilities for medication adherence in this age group. The better adherence observed among young women than young men may be related to earlier cognitive maturation in females than males,17 resulting in a greater capacity for self-care in young women than men. A greater influence of social desirability (a wish to comply with social expectations) among women than men26,27 may also contribute to the differences observed. Although self-reported adherence was not higher in women than men, this may reflect the fact that participants knew that their adherence was being monitored electronically; young women may have been more concerned that their reports accurately reflect the electronic record than young men. We must also consider the possibility that the observed gender differences represent different reactions to being observed between boys/men and girls/women in the context of a relatively brief study period.
Our findings also highlight possible age- and gender-based differences in the accuracy of different adherence measures. Self-report has been previously reported to overestimate adherence.6,28 We observed a large difference between self-reported and electronically measured taking adherence. We also observed what appeared to be a greater overestimation among girls than boys aged 11 to 16 years, and a greater overestimation among men than women aged 17 to 24 years. However, we cannot exclude the possibility that there are also gender differences in adherence to using the electronic monitoring pillbox. Failure to detect significant differences in self-reported adherence by gender may reflect a ceiling effect, whereby variability is not well captured, diminishing the possibility of finding differences.
We found no gender differences in either age group in adherence assessed using the SD of tacrolimus trough levels, which is often considered an objective measure of adherence.29 In fact, a slightly higher proportion of women than men had an SD of tacrolimus levels >2.0, which is opposite to what one would expect based on the electronic monitoring data. There may be several explanations for this. First, patients may have better adherence in the days preceding a planned blood level,30 masking gender differences. This observation also raises the question as to whether tacrolimus metabolism may be influenced by gender hormones, with changes in metabolism depending on stage of the menstrual cycle.31-33 If this were the case, tacrolimus trough levels in women may vary with the menstrual cycle, despite good adherence to a constant dose, leading to a higher SD. This has not been studied but deserves investigation. Finally, given the small numbers of participants aged 17 to 24 years, only a few participants with SD of tacrolimus levels >2.0 may make a substantial difference to the proportions with SD of tacrolimus levels >2.0.
The lack of concordance between the different adherence assessment methods may also reflect gender differences in willingness to consistently use an electronic monitoring device. However, we found no literature suggesting that use of electronic monitors differs by gender.
This study has limitations. First, the period of observation was relatively short. Although the first 2 weeks of electronic data were excluded from analyses (as is conventional with electronic monitoring), it is possible that participants’ behavior continued to be influenced by the fact that they were under observation or that their medication-taking routine had been altered. Second, the short period of observation meant that we were unable to observe changes in adherence with changing age; rather, we observed adherence among individuals of different ages. A third limitation is the lack of concordance in results when different methods were used to assess adherence. Although we have outlined possible explanations for these disparities, it is not possible to know which method most accurately represents behavior. There is no consensus regarding the best method to use to measure adherence.34-36 Although imperfect, electronic monitoring is generally considered the gold standard, with good accuracy,24,37 reflecting the participants’ behaviors.36 Residual confounding due to factors, such as gender differences in education, cannot be excluded. Finally, the relatively small sample, despite being derived from 8 centers in 2 countries, may limit generalizability.
Based on electronic monitoring, there are no apparent differences in medication adherence by gender in adolescents aged 11 to 16 years old, but poorer adherence in young men aged 17 to 24 years than young women of the same age. These findings may help with interpretation of studies showing differences in graft outcomes by gender.2 It may be important to take these gender differences into consideration when targeting patients for adherence-promoting interventions. This study also highlights potentially important gender differences in self-reporting of medication adherence, as well as the possibility that variability in tacrolimus trough levels may depend on factors other than adherence. Larger cohort studies using multiple methods to assess adherence are needed not only to confirm our findings but also to further explore gender differences in reporting of adherence. Pharmacokinetic studies to identify biologic factors that may contribute to variability in tacrolimus trough levels would also be helpful. A longitudinal study following younger adolescents into young adulthood would also allow assessment of the evolution of adherence behavior over time.
The authors would like to acknowledge the devotion of the study coaches and the generous participation of the patients and their families without whom this study would not have been possible.
1. Kabore R, Couchoud C, Macher MA, et al. Age-dependent risk of graft failure in young kidney transplant recipients. Transplantation. 2017;101:1327–1335.
2. Lepeytre F, Dahhou M, Zhang X, et al. Association of sex with risk of kidney graft failure differs by age. J Am Soc Nephrol. 2017;28:3014–3023.
3. Bouman A, Heineman MJ, Faas MM. Sex hormones and the immune response in humans. Hum Reprod Update. 2005;11:411–423.
4. Fish EN. The X-files in immunity: sex-based differences predispose immune responses. Nat Rev Immunol. 2008;8:737–744.
5. Chisholm-Burns MA, Spivey CA, Tolley EA, et al. Medication therapy management and adherence among US renal transplant recipients. Patient Prefer Adherence. 2016;10:703–709.
6. Denhaerynck K, Steiger J, Bock A, et al. Prevalence and risk factors of non-adherence with immunosuppressive medication in kidney transplant patients. Am J Transplant. 2007;7:108–116.
7. Prihodova L, Nagyova I, Rosenberger J, et al. Adherence in patients in the first year after kidney transplantation and its impact on graft loss and mortality: a cross-sectional and prospective study. J Adv Nurs. 2014;70:2871–2883.
8. Spivey CA, Chisholm-Burns MA, Damadzadeh B, et al. Determining the effect of immunosuppressant adherence on graft failure risk among renal transplant recipients. Clin Transplant. 2014;28:96–104.
9. Weng LC, Yang YC, Huang HL, et al. Factors that determine self-reported immunosuppressant adherence in kidney transplant recipients: a correlational study. J Adv Nurs. 2017;73:228–239.
10. Chisholm MA, Lance CE, Mulloy LL. Patient factors associated with adherence to immunosuppressant therapy in renal transplant recipients. Am J Health Syst Pharm. 2005;62:1775–1781.
11. Marsicano EO, Fernandes NS, Colugnati FA, et al. Multilevel correlates of non-adherence in kidney transplant patients benefitting from full cost coverage for immunosuppressives: a cross-sectional study. PLoS One. 2015;10:e0138869.
12. Vlaminck H, Maes B, Evers G, et al. Prospective study on late consequences of subclinical non-compliance with immunosuppressive therapy in renal transplant patients. Am J Transplant. 2004;4:1509–1513.
13. Weng FL, Israni AK, Joffe MM, et al. Race and electronically measured adherence to immunosuppressive medications after deceased donor renal transplantation. J Am Soc Nephrol. 2005;16:1839–1848.
14. Dobbels F, Ruppar T, De Geest S, et al. Adherence to the immunosuppressive regimen in pediatric kidney transplant recipients: a systematic review. Pediatr Transplant. 2010;14:603–613.
15. Chisholm-Burns MA, Spivey CA, Rehfeld R, et al. Immunosuppressant therapy adherence and graft failure among pediatric renal transplant recipients. Am J Transplant. 2009;9:2497–2504.
16. Dew MA, Dabbs AD, Myaskovsky L, et al. Meta-analysis of medical regimen adherence outcomes in pediatric solid organ transplantation. Transplantation. 2009;88:736–746.
17. De Bellis MD, Keshavan MS, Beers SR, et al. Sex differences in brain maturation during childhood and adolescence. Cereb Cortex. 2001;11:552–557.
18. Blakemore SJ, Choudhury S. Development of the adolescent brain: implications for executive function and social cognition. J Child Psychol Psychiatry. 2006;47:296–312.
19. Stilley CS, Lawrence K, Bender A, et al. Maturity and adherence in adolescent and young adult heart recipients. Pediatr Transplant. 2006;10:323–330.
20. Foster BJ, Pai A, Zhao H, et al. The TAKE-IT study: aims, design, and methods. BMC Nephrol. 2014;15:139.
21. Foster BJ, Pai ALH, Zelikovsky N, et al. A randomized trial of a multicomponent intervention to promote medication adherence: the Teen Adherence in Kidney Transplant Effectiveness of Intervention Trial (TAKE-IT). Am J Kidney Dis. 2018;72:30–41.
22. Zelikovsky N, Schast AP. Eliciting accurate reports of adherence in a clinical interview: development of the Medical Adherence Measure. Pediatr Nurs. 2008;34:141–146.
23. Zelikovsky N, Schast AP, Palmer J, et al. Perceived barriers to adherence among adolescent renal transplant candidates. Pediatr Transplant. 2008;12:300–308.
24. Denhaerynck K, Schafer-Keller P, Young J, et al. Examining assumptions regarding valid electronic monitoring of medication therapy: development of a validation framework and its application on a European sample of kidney transplant patients. BMC Med Res Methodol. 2008;8:5.
25. Yang D, Dalton JE. A unified approach to measuring the effect size between two groups using SAS®.
SAS Global Forum 2012 - statistics and data analysis2012.Orlando, FL
26. Bernardi RA. Associations between Hofstede’s Cultural Constructs and Social Desirability Response Bias. J Bus Ethics. 2006;65:43–53.
27. Hebert JR, Clemow L, Pbert L, et al. Social desirability bias in dietary self-report may compromise the validity of dietary intake measures. Int J Epidemiol. 1995;24:389–398.
28. Liu H, Golin CE, Miller LG, et al. A comparison study of multiple measures of adherence to HIV protease inhibitors. Ann Intern Med. 2001;134:968–977.
29. Shemesh E, Fine RN. Is calculating the standard deviation of tacrolimus blood levels the new gold standard for evaluating non-adherence to medications in transplant recipients? Pediatr Transplant. 2010;14:940–943.
30. Cramer JA, Scheyer RD, Mattson RH. Compliance declines between clinic visits. Arch Intern Med. 1990;150:1509–1510.
31. Kuypers DR, Claes K, Evenepoel P, et al. Time-related clinical determinants of long-term tacrolimus pharmacokinetics in combination therapy with mycophenolic acid and corticosteroids: a prospective study in one hundred de novo renal transplant recipients. Clin Pharmacokinet. 2004;43:741–762.
32. Soldin OP, Mattison DR. Sex differences in pharmacokinetics and pharmacodynamics. Clin Pharmacokinet. 2009;48:143–157.
33. Fadiran EO, Zhang L. Harrison-Woolrych M. Effects of Sex Differences in the Pharmacokinetics of Drugs and Their Impact on the Safety of Medicines in Women. In: Medicines For Women. 2015Cham: Springer International Publishing; 41–68.
34. Lieber SR, Helcer J, Shemesh E. Monitoring drug adherence. Transplant Rev (Orlando). 2015;29:73–77.
35. Williams A, Low JK, Manias E, et al. Trials and tribulations with electronic medication adherence monitoring in kidney transplantation. Res Social Adm Pharm. 2016;12:794–800.
36. Lehmann A, Aslani P, Ahmed R, et al. Assessing medication adherence: options to consider. Int J Clin Pharmacol. 2014;36:55–69.
37. De Bleser L, De Geest S, Vandenbroeck S, et al. How accurate are electronic monitoring devices? A laboratory study testing two devices to measure medication adherence. Sensors (Basel). 2010;10:1652–1660.