Predictors of antidepressant adherence: Results of a Japanese Internet-based survey

Authors


*Jun Shigemura, MD, Department of Psychiatry, National Defense Medical College; 3-2 Namiki, Tokorozawa 359-8513, Japan. Email: shige@ndmc.ac.jp

Abstract

Aim:  The purpose of the present study was to identify the psychosocial/pharmacological predictors of antidepressant (AD) adherence.

Methods:  An Internet-based survey was conducted among 1151 Japanese individuals with major depressive disorder. Subjects were asked to report their degree of non-adherence for each AD taken using a 5-point Likert scale: 0, never forget; 1, rarely forget; 2, occasionally forget; 3, sometimes forget; and 4, often forget. The highest number reported among each subject was assigned as their low adherence index (LAI). Individuals with an LAI ≥ 3 were defined as members of the low adherence (LA) group. Predictors of LA was analyzed using bivariate and multivariate models, both among the total number of subjects and single AD subgroup (n = 657).

Results:  Nearly one-third of subjects (n = 381, 33.1%) reported LA. On bivariate analysis, LA was associated with lower age, worker or student status (vs unemployed or housewife), higher daily dosing frequency (DDF), low drug satisfaction, and a neutral/negative doctor–patient relationship (DPR; P < 0.001). In a multivariate model, LA was predicted by age (≤34 years: odds ratio [OR], 1.64), worker or student status (OR, 1.87), higher DDF (≥twice daily: OR, 1.61), and neutral/negative DPR (OR, 1.54; P < 0.01). Among the single-AD subgroup, adherence was similar between those on selective serotonin reuptake inhibitors/serotonin-noradrenaline reuptake inhibitors and tricyclics. Use of neither medication was associated with adherence in a multivariate model.

Conclusion:  LA was predicted by lower age, worker or student status, higher DDF, and neutral/negative DPR. Adherence was not significantly different between subjects on newer agents and tricyclics.

DEPRESSION IS A major public health concern worldwide. Given the often chronic and recurrent nature of the disorder, the treatment of depression requires a comprehensive approach; various treatments are provided, alone or in combinations. Treatment options include antidepressant (AD) therapy, augmenting and adjunctive medications, psychotherapy, and complementary/alternative medicine.1 Among them, the most significant treatment element is AD pharmacotherapy. AD treatment provides an amelioration of acute symptoms as well as a decrease in the risk of relapse and recurrence.2

Not all patients, however, take their medications as prescribed, thus resulting in premature treatment dropout and poorer health outcomes. In a review of papers published between 1976 and 2001,3 the prevalence of AD non-adherence ranged from 10% to 60%. Studies also show that non-adherent patients stop taking AD not only in the early treatment phases, but also in the maintenance stages,3–6 and that non-adherence predicted time to recurrence.7 Therefore, maintenance of adherence is a crucial element in the successful treatment of depression.

AD adherence is determined by multidimensional factors. Various studies show that certain demographic factors (including male gender, younger age, lower education, lower income, and non-married status) are risk factors for low adherence, although results tend to be inconsistent between studies.8–12 Factors related to course of the disorder include symptom severity; comorbid physical disorders; and comorbid psychiatric disorders, such as personality disorder and substance abuse.4,12,13

Other studies have focused on relationships between low adherence and pharmacological characteristics, including the use of older AD, such as tricyclic antidepressants (TCA);4,9,14 a complex treatment regimen;15 and the presence of unpleasant side-effects.16 Growing evidence also suggests that cognitive and subjective factors are related to adherence, including patient education;17 treatment perceptions, such as drug satisfaction;18,19 and the state of the doctor–patient relationship.20,21

To ascertain the relationships between AD adherence, sociodemographic factors, pharmacological characteristics, and subjective perceptions, we examined the results of an Internet-based study among Japanese individuals.

METHODS

Methods have been previously reported in detail.22 The study was conducted using an Internet-based survey in the Japanese language, and the protocol was in accordance with the Declaration of Helsinki. Potential participants were members of an Internet research panel service provided by a website (Yahoo! Japan Research; http://research.yahoo.co.jp). In the survey the following information was assessed: (i) demographic information; (ii) the names of currently prescribed AD; (iii) the degree of adherence and perceived satisfaction with each AD that the subject reported taking; (iv) the presence of perceived side-effects from the AD; and (v) the perceived state of the doctor–patient relationship. In our previous report, 1191 subjects were enrolled; in the present study we further excluded individuals who did not provide their prescribed drug name or their degree of adherence. Following this process, 1151 subjects were enrolled for the final analysis.

Questions were asked regarding degree of adherence to each taken AD; the English translation for the question was as follows: ‘How often do you forget to take your medication?’. The responses were quantified on a 5-point Likert scale: 0, never forget; 1, rarely forget; 2, occasionally forget; 3, sometimes forget; and 4, often forget. The highest number reported among each subject was assigned as their low adherence index (LAI). Individuals with an LAI ≥ 3 were defined as members of the low adherence group.

The degree of perceived drug satisfaction was assessed using a 5-point Likert scale, ranging from 1, very dissatisfied, to 5, very satisfied, using the following question: ‘How satisfied are you with the medication you have been taking?’. For subjects on two or more AD, the average of their scores was defined as their satisfaction score. Those with a satisfaction score ≥4.0 were identified as members of the ‘high satisfaction group’.

The presence of perceived AD side-effects was categorized as either ‘none’ or ‘present’; we determined this by asking if there were ‘any side-effects that negatively affected your work or daily activities’. The quality of the perceived state of the doctor–patient relationship was evaluated using the following question: ‘Please describe the relationship between you and your prescribing doctor’. Their responses were measured on a 7-point Likert scale ranging from 1 to 7: 1, very good; 2, good; 3, somewhat good; 4, neither good nor bad; 5, somewhat bad; 6, bad; and 7, very bad. Their responses were collapsed into three categories: positive (1, 2, or 3), neutral (4), and negative (5, 6, or 7).

Statistical analyses were conducted using SPSS 12.0.1 for Windows (SPSS, Chicago, IL, USA). The significance level was set at P < 0.05 (two-tailed). On bivariate analysis, correlations between continuous variables were analyzed using Pearson's correlation coefficients. The relationships between continuous variables and categorical variables were calculated using t-tests and one-way analysis of variance (anova) tests with post-hoc least significant difference tests. The relationships between categorical variables were measured using χ2 tests.

To understand better the relationship between AD selection and adherence, a bivariate analysis was performed among all subjects, as well as among subjects on a single AD therapy (i.e. monotherapy). Within the monotherapy group, the subjects were categorized into subgroups according to their AD (amitriptyline, amoxapine, clomipramine, fluvoxamine, imipramine, maprotiline, milnacipran, nortriptyline, paroxetine, sulpiride, and trazodone). When the number of subjects was ≤10 for the corresponding AD, the AD was categorized as ‘others’. Independent variables significant in the bivariate model were compared between AD subgroups.

On multivariate analysis, a stepwise logistical regression model (the Wald forward selection method) was used to identify significant and independent variables predicting low adherence. Independent variables significant on bivariate analysis were entered into the model as potential predictors. For continuous independent variables, the median values were used to transfer the variable to categorical variables. Multivariate analyses were performed among all the subjects, as well as the monotherapy subjects alone. For the monotherapy subjects, those with AD who had a significant difference in LAI were differentiated from those who did not; this categorical variable was entered into the multivariate model.

RESULTS

Nearly one-third of the subjects (n = 381, 33.1%) reported low adherence, that is, high LAI scores. The distribution of LAI was as follows: 0, n = 247 (21.5%); 1, n = 277 (24.1%); 2, n = 246 (21.4%); 3, n = 238 (20.7%); and 4, n = 143 (12.4%). The average and median ages of the subjects were 34.6 ± 7.1 years and 34 years, respectively. Younger subjects were more likely to be non-adherent; age negatively correlated with LAI (r = −0.138, P < 0.001). The subjects were on AD for 2.70 ± 3.00 years, which did not correlate with LAI (r = −0.05, P = 0.113).

Demographic features, pharmacological profiles, subjective responses, and adherence among all subjects are listed in Table 1. The age distribution was as follows: 20–29 years, n = 377 (32.8%); 30–39 years, n = 371 (32.2%); 40–49 years, n = 388 (33.0%); and 50–59 years, n = 15 (2.0%).

Table 1.  Subject characteristics vs LAI
 All subjects (n = 1151)Subjects on single antidepressants (n = 657)
SubjectsLAIAnalysisSubjectsLAIAnalysis
n%Mean ± SDn%Mean ± SD
  • *

    P < 0.05;

  • **

    P < 0.01 (post-hoc least significant difference).

  • Worker > housewife, unemployed; student > housewife, unemployed.

  • Worker > housewife, unemployed.

  • §

    Once > twice, thrice.

  • positive < neutral.

  • ††

    n vary due to missing values.

  • LAI, low adherence Index.

Total1151100.01.79 ± 1.33 657100.01.79 ± 1.33 
Age (years)        
 ≤3457950.31.93 ± 1.35t = 3.81**35654.21.90 ± 1.35t = 2.17*
 ≥3557249.71.64 ± 1.29(d.f. = 1149)
P < 0.001
30145.81.67 ± 1.29(d.f. = 655)
P = 0.030
Gender        
 Female64556.01.82 ± 1.38t = 1.0042063.91.80 ± 1.37t = 0.24
 Male50644.01.74 ± 1.26(d.f. = 1149)
P = 0.316
32736.11.78 ± 1.25(d.f. = 655)
P = 0.810
Occupation        
 Worker79769.21.90 ± 1.31F = 6.73**45168.61.94 ± 1.32F = 5.02**
 Housewife15213.21.47 ± 1.34(d.f. = 4)9714.81.37 ± 1.28(d.f. = 4)
 Unemployed12210.61.40 ± 1.26P < 0.001619.31.48 ± 1.29P = 0.001
 Student544.71.93 ± 1.32 3351.67 ± 1.32 
 Other262.31.58 ± 1.42 152.31.60 ± 1.45 
Years on antidepressants††        
 <2 years52845.91.72 ± 1.33t = 1.0446170.21.83 ± 1.34t = 0.99
 ≥2 years53946.81.80 ± 1.31(d.f. = 1065)
P = 0.299
19629.81.71 ± 1.29(d.f. = 665)
P = 0.322
No. antidepressants        
 165757.11.79 ± 1.33F = 0.763    
 236331.51.72 ± 1.33(d.f. = 3)    
 3998.61.93 ± 1.27P = 0.515    
 4322.81.91 ± 1.45     
Daily dosing frequency        
 138233.21.58 ± 1.35F = 7.03**§29444.71.61 ± 1.36F = 5.92**§
 240034.81.86 ± 1.33(d.f. = 2)22434.11.89 ± 1.28(d.f. = 2)
 336932.11.92 ± 1.28P = 0.00113921.22.04 ± 1.29P = 0.003
Perceived side-effect        
 Yes95983.31.77 ± 1.33t = 0.9153180.81.77 ± 1.33t = 0.83
 No19216.71.86 ± 1.32(d.f. = 1149)
P = 0.365
12619.21.88 ± 1.34(d.f. = 655)
P = 0.409
Perceived drug satisfaction        
 High61153.11.67 ± 1.30t = 3.26**40361.31.69 ± 1.30t = 2.58*
 Low54046.91.92 ± 1.33(d.f. = 1149)
P = 0.001
25438.71.96 ± 1.35(d.f. = 655)
P = 0.010
Perceived doctor–patient relationship
 Positive94882.41.72 ± 1.31F = 7.64**53481.31.72 ± 1.31F = 4.56*
 Neutral14012.22.17 ± 1.33(d.f. = 2)8913.52.18 ± 1.32(d.f. = 2)
 Negative635.51.94 ± 1.40P = 0.001345.21.85 ± 1.48P = 0.011

On bivariate analysis, low adherence was associated with younger age, occupation status (i.e. worker or student), higher daily dosing frequency, low drug satisfaction, and a negative/neutral doctor–patient relationship. Those with higher AD satisfaction were more likely to be adherent; satisfaction score negatively correlated with LAI (r = −0.115, P < 0.001). Adherence was not relevant to gender, the number of years on AD, perceived side-effects, or number of AD (single vs multiple drugs: n = 657 vs 494; LAI = 1.79 ± 1.33 vs 1.78 ± 1.32; t = 0.22, d.f. = 1149, P = 0.823).

In the monotherapy group (n = 657), the proportion of individuals with low adherence was 33.0% (n = 217). Variables associated with low adherence in the monotherapy group were similar to those of the whole subject group (Table 1). Table 2 shows the relationship between these variables and AD; subjects taking amoxapine or paroxetine had higher adherence compared to those taking other AD. Adherence was not significantly different between subjects on newer medications (e.g. selective serotonin reuptake inhibitors [SSRI] and serotonin-noradrenaline reuptake inhibitors [SNRI]) and TCA (n = 449 vs 94; LAI = 1.77 ± 1.33 vs 1.82 ± 1.37, t = 0.35, d.f. = 541, P = 0.73).

Table 2.  Drug type and subject characteristics vs LAI for mono-antidepressant therapy (n = 657)
AntidepressantsSubject age (years)Daily dosing frequencyLAI
 n%Mean ± SDMean ± SDMean ± SD
  • *

    P < 0.05,

  • **

    P < 0.01. Not significant: work status (worker or student status), drug satisfaction (satisfaction score), and perceived doctor–patient relationship. Analysis of variance with post-hoc comparison (least significant difference test).

  • 4 < 2, 3, 5, 10, 12*; 9 < 2, 3, 5, 7, 10, 12*.

  • 4 < 7**; 9 < 1–8, 10, 12**; 11 < 1–7, 10, 12*.

  • §

    2 < 3–5 *; 9 < 3–5, 7*.

  • LAI, low adherence Index.

Total657100.033.9 ± 7.61.76 ± 0.781.79 ± 1.33
 1 Amitriptyline203.035.2 ± 6.62.00 ± 0.861.80 ± 1.28
 2 Amoxapine264.036.5 ± 7.92.04 ± 0.771.38 ± 1.47*
 3 Clomipramine203.038.3 ± 8.02.15 ± 0.882.20 ± 1.44
 4 Fluvoxamine16324.833.1 ± 7.4*1.92 ± 0.76**1.96 ± 1.36
 5 Imipramine172.637.2 ± 8.42.06 ± 0.902.29 ± 1.11
 6 Maprotiline162.435.3 ± 7.22.00 ± 0.822.13 ± 1.46
 7 Milnacipran9113.934.7 ± 7.42.21 ± 0.611.92 ± 1.20
 8 Nortriptyline111.735.6 ± 7.91.82 ± 0.871.45 ± 1.29
 9 Paroxetine19529.732.2 ± 7.2*1.18 ± 0.42**1.53 ± 1.33*
10 Sulpiride619.335.3 ± 8.12.05 ± 0.781.82 ± 1.25
11 Trazodone121.833.8 ± 6.91.42 ± 0.79*2.17 ± 1.53
12 Others253.836.5 ± 7.22.00 ± 0.711.72 ± 1.24
Analysis     
 F (d.f. = 11)  3.03**21.1**1.88*
 P  0.001<0.0010.040
 Post-hoc tests  §

For the whole subject and monotherapy groups, multivariate analysis showed that predictors of low adherence were (i) younger age (≤34 years); (ii) occupational status (worker or student); (iii) dosing frequency (regimen of more than twice a day); and (iv) perceived state of the doctor–patient relationship (neutral/negative; Table 3). In the monotherapy group, adherence was not associated with the use of either amoxapine or paroxetine.

Table 3.  Multiple logistic regression analysis: predictors of low adherence (LAI ≥ 3)
 Wald χ2 (d.f. = 1)POdds Ratio95% Confidence Interval
  • *

    P < 0.05,

  • **

    P < 0.01.

  • Variable rejected from stepwise equation: low drug satisfaction.

  • Variables rejected from stepwise equation: low drug satisfaction, subjects on neither amoxapine nor paroxetine.

  • LAI, low adherence index.

All subjects (n = 1151)    
 Age (≤34 years)**14.48<0.0011.641.27–2.11
 Occupation (worker or student)**16.15<0.0011.871.38–2.53
 Higher dosing frequency (≥twice per day)**11.530.0011.611.22–2.12
 Perceived doctor–patient relationship** (neutral/negative vs positive)7.140.0081.541.12–2.12
Single antidepressant therapy subjects (n = 657)    
 Age (≤34)*4.940.0261.471.05–2.07
 Occupation (worker or student)**15.64<0.0012.321.53–3.51
 Higher dosing frequency (≥twice per day)**7.570.0061.611.15–2.27
 Perceived doctor–patient relationship* (neutral/negative vs positive)5.400.0201.641.08–2.49

DISCUSSION

The present results showed that younger age was a predictor of low AD adherence. This result is consistent with a study of 4312 depressed subjects on AD4 and a study of 634 individuals on psychotropic drugs, in which younger individuals were particularly likely either to forget their medication or to stop taking it because of side-effects.9 Previous studies suggest that comorbid mental disorders, especially personality disorders and substance abuse, are risk factors for low AD adherence.4,12,13 Due to methodological issues, we were not able to ascertain the presence of comorbid mental disorders other than bipolar disorder and schizophrenia, thus we could not investigate the confounding effects of these comorbid disorders or of age, either of which may have influenced the presents findings.

We also need take into account a study limitation regarding age and Internet-based sampling. The present subjects were relatively younger than a sample in a Japanese community study,23 in which individuals with major depressive disorder were uniformly distributed between the age groups 20–34 years, 35–44 years, 45–54 years, and 55–64 years. In contrast, according to the Japanese Internet White Paper 2006, the majority of the Japanese Internet users had been in their 20s, 30s, and 40s.24 Therefore, the present study may not represent the Japanese population of subjects diagnosed with major depressive disorder, but the population of Internet users with major depressive disorder.

The present study also found that workers and students were more likely to forget to take their medication than housewives or unemployed individuals. This finding is in contrast to those of past studies, which reported that lower socioeconomic status and social adjustment are predictors of low AD adherence.4,8,11 Individuals with full-time engagements outside of their homes may be hesitant to take medications in public or may forget to take their medication during their daily activities.25 Selection bias due to Internet-based study design, however, needs to be taken into consideration, and further studies are necessary to support these hypotheses.

The present study failed to show any relationships between adherence and other demographic variables. Overall, relationships between demographic profiles and AD adherence remain inconsistent,3 and could serve as a fruitful topic for future research.

We found that subjects on once-daily prescriptions had a higher AD adherence than those on twice- or thrice-daily prescriptions, suggesting that a reduction in dosing frequency is favorable in improving AD adherence. In a study comparing once-, twice-, and thrice-daily regimens of bupropion users, reduction in daily dosing frequency was associated with higher adherence.19 This trend may be due to treatment simplicity and a concomitant lower burden on the patient, given that a complex treatment regimen was found to be associated with low adherence.15 Another explanation is that controlled-release formulations have higher tolerability due to fewer side-effects, compared to immediate-release formulations.16 In the present study, however, adherence was affected by neither the presence of perceived side-effects nor the number of AD used. Lin et al. reported that side-effects were associated with adherence when they occurred at a severe level,26 but we did not assess the level of severity of side-effects nor asymptomatic adverse effects; we simply asked a dichotomous question on perceived side-effects, which is a study design drawback.

We determined that individuals on amoxapine or paroxetine reported higher adherence in the bivariate but not the multivariate model. In the present sample the mean age of amoxapine users was relatively higher. Therefore, the adjustment of age may have affected the significance of amoxapine in relation to adherence, on multivariate analysis. Likewise, paroxetine users were significantly younger and had lower dosing frequencies than other subjects on other agents, in the bivariate model. After adjusting for age and daily dosing frequency in the multivariate model, however, use of paroxetine was not associated with high adherence.

The present study found that the use of older AD, such as TCA, was not relevant to lower adherence. There are studies stating that the advent of newer agents has not changed adherence patterns,5 while some studies reported the use of TCA as a risk factor for lower adherence.4,9,14 The present study suggests that certain pharmacological profiles, such as dosing frequency, are more relevant to adherence than drug selection, although caution is necessary; we were unable to assess socioeconomic status, history of used AD, and severity of subjects' conditions, due to methodological limitations. Japanese treatment guidelines recommend SSRI and SNRI as first-line pharmacological treatments of depression,27 so it is also possible that those on TCA were so treated as a consequence of an unsuccessful first-line treatment with SSRI and/or SNRI. Further studies are necessary, to understand better the relationships between adherence, treatment course, and AD selection.

According to the present results, the subjective components of treatment are related to AD adherence, especially with respect to the perceived state of the doctor–patient relationship. This result is consistent with those of previous studies, which stated that physician communication style positively influences patient knowledge and perception of the medication.20,21 Such communication leads to better patient education and more in-depth knowledge of both the disorder and the medication, thus resulting in higher treatment adherence.17

Past studies have shown associations between adherence and AD satisfaction,18,19 and that predictors of higher AD satisfaction were perceived efficacy and fewer side-effects.22 In the present study, however, high AD satisfaction was associated with better adherence in the bivariate but not the multivariate model. Demyttenaere et al. reported preserved autonomy and partner agreement as relevant factors to AD adherence, along with drug satisfaction and the state of the doctor–patient relationship.21 Adherence may be related to satisfaction – not only in the subjects' AD, but in the whole of their treatment regimen – and there is a need for further study, to investigate this relationship.

The present study is constrained by several limitations, other than the aforementioned issues. Due to the Internet-based survey methodology, a sampling bias may have occurred; furthermore, responses were based on self-report.28 Bias may have also occurred as a result of characteristics unique to Japanese samples and differences in the availability of AD in Japan at the time of the study, compared to those in other countries. We were unable to assess information related to patient disorders and treatments, including comorbid disorders, severity of depression, duration of received treatment, AD dose and response, and adjunctive therapies.

Even with these limitations, however, the present findings are unique in terms of its large-scale, Internet-based study design and its examination of the Japanese population. To the best of our knowledge, there are no other studies examining demographic, pharmacological, and subjective factors related to AD adherence in Japan. Also, there is a dearth of comparison studies into adherence rates for different AD. Further studies would be useful in better understanding the factors associated with adherence among patients with depression.

ACKNOWLEDGMENTS

This study was conducted by Glaxo SmithKline. The questionnaire was originally conducted by the company as a part of their consumer survey. The authors do not have the information on how the company originally used the results obtained from the questionnaire. The authors subsequently received the raw data from the company solely for research purposes. The company had no role in analysis and interpretation of the data; in the writing of the report; or in the decision to submit the paper for publication. Dr Shigemura has received speakers' honoraria from Glaxo SmithKline, Otsuka Pharmaceutical, and Pfizer Japan. Dr Nomura has received speakers' honoraria from Asahi Kasei Pharma, Astellas Pharma, Glaxo SmithKline, Meiji Seika Kaisha, Mitsubishi Pharma, and Pfizer Japan.

Ancillary