Introduction
The window of opportunity concept states that there are superior clinical responses and the potential for remission when patients with rheumatoid arthritis (RA) are managed early and aggressively with disease-modifying anti-rheumatic drugs (DMARDs) [
1‐
3]. Early RA clinics are the perfect setting to address such concept, as patients with recent disease onset are referred to experts in rheumatologic care without delay. Nonetheless, it seems intuitive to assume that the full benefit of the pharmacologic intervention will be achieved if patients follow prescribed regimens reasonably closely.
Medication adherence refers to the patient’s act of conforming to the recommendations made by the health care provider with respect to timing, dosage and frequency of medication-taking [
4,
5]. Medication adherence can be divided into three major components: (1)
persistence, defined as the length of time a patient fills prescriptions without permissible gaps; (2)
initiation adherence, defined as the patient starting with the intended treatment; and (3)
execution adherence that results from the comparison between the prescribed drug regimen and the patient’s actual drug-taking behavior [
4].
Poor adherence with therapy affects 20 to 70 % of patients with RA, sometimes during their follow-up [
6‐
14]. Local experience within a cohort of Mexican Mestizo patients with early RA (ERAP) confirmed these data and additionally showed that poor adherence with traditional DMARDs was associated with increased disease flares, decreased rates of remission and worse patient-reported outcomes [
15‐
17], although the mechanisms were not examined. In particular, there may be aspects within the adherence and persistence constructs, such as duration (of persistence) or timing (of first non-persistence period), that may have a different (and/or additive) impact on outcomes. We sought to explore this hypothesis in a well-defined population of ERAPs in whom persistence was prospectively evaluated and the concept of window of opportunity was extended to the construct of persistence.
The following were the specific objectives of the study:
1.
To investigate if the timing of first non-persistence period and/or duration of persistence during the first 4 years of follow-up were predictors of the level of disease activity at the fifth year of follow-up and of achieving sustained remission (SR).
2.
If the first objective was achieved, we planned to extend the analysis to a patient-reported outcome and investigate a similar impact on patient function as evaluated per the Health Assessment Questionnaire (HAQ) [
18].
3.
To define additional predictors of disease outcomes.
Methods
Setting and study population
The Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán is a national referral center for rheumatic diseases that belongs to the National Institutes of Health in Mexico City. In 2004, an early arthritis clinic was established. Patients entering the clinic had disease durations of less than 1 year when first evaluated and no specific rheumatic diagnosis other than RA. Patients were evaluated every 2 months during the first 2 years of follow-up and thereafter every 2, 4 or 6 months (fixed for all the patients from the baseline evaluation), depending on patient and disease characteristics. Treatment was prescribed by the rheumatologist in charge of the clinic and was given using a treat-to-target (T2T) approach. Briefly, at every medical encounter, the Disease Activity Score in 28 joints (DAS28) was calculated and the level of disease activity defined. If remission was achieved, no major treatment modifications were made; if not, treatment was intensified with the aim of achieving remission, which was defined as DAS28 less than 2.6. Traditional DMARDs were used in 99 % of our population, with or without corticosteroids (around 50 % of the patients). From the beginning of the clinic, patients’ medication behavior was prospectively assessed, initially through a structured interview and starting in November 2008 by using an instrument locally designed, the Concordance Questionnaire (CQ) formerly called the Compliance Questionnaire, which evaluates both constructs: adherence to and persistence with therapy [
16].
Up to February 2015, 165 ERAPs had been evaluated, of whom 107 had at least 5 years of follow-up (19 patients were lost to follow-up, and 2 additional patients died). The 5 years of follow-up was deemed to be sufficient to accomplish the objectives described.
Rheumatic evaluations
At study entry, a complete medical history and demographic data, along with disease-specific autoantibodies, were recorded. Medical evaluations were standardized and included swollen and tender joint counts, acute reactant-phase determinations, patient- and physician- reported outcomes, comorbidity established by record review and treatment assessment (names, doses and schedules of all drugs taken since last visit), along with the evaluation of persistence.
Persistence evaluation
From 2004 to 2008, persistence was assessed through an interview conducted at every visit by the same rheumatologist. Patients were directed to report the names, doses and schedules of DMARDs and corticosteroids they had taken since last visit (fixed at 6 months apart), initially spontaneously and if necessary directly. Then, patients were asked about any missing and/or incorrect medications, doses and/or schedules since their last visit. The number of days of missing medication was recorded. The rheumatologist compared the last prescription and the actual treatment, and, if inconsistencies were found, they were resolved. Data were collected in standardized formats.
Since 2008, persistence (and adherence) was evaluated through the CQ. The performance of the CQ has shown high sensitivity and satisfactory specificity to detect persistence [
16]. CQ was fulfilled without help by 95 % of the patients.
Definitions
Sustained remission was assessed at the fifth year of follow-up and defined if DAS28 was below 2.6 at the three consecutive evaluations within that year [
19,
20].
The DAS28 at the fifth year of follow-up was calculated as the mean of individual DAS28 from visits performed during the fifth year.
Patient function was also evaluated at the fifth year of follow-up. A patient was considered to be without disability if the HAQ score was sustained at or below 0.20 at all consecutive evaluations within that year. In addition, HAQ score at the fifth year was derived from the mean of individual HAQs from visits during the fifth year.
According to the interview, non-persistence with medication was defined as omission of at least one DMARD and/or corticosteroid for at least 7 consecutive days. Regarding methotrexate, at least one missing weekly dose was considered non-persistence. Treatment modifications because of adverse events and/or indicated by a different physician for any reason (e.g., insufficient response, pregnancy, schedule surgery) were not considered non-persistence under the construct. According to the CQ, a patient was considered to be non-persistent if, in item 10 (“In the past 6 months, how often did you completely stop taking your DMARDs?”), boxes 2 (“Sometimes”), 3 (“Almost always”) and 4 (“Always”) were filled. Persistence was evaluated at 6-month periods (fixed for all the patients) and defined by an independent observer according to the information recorded in standardized formats.
Ethics
The study was approved by the institution’s internal review board (Comites de Ética e Investigación del Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán). Written informed consent was obtained from all the patients when entering the clinic. Also, specific written consent was obtained to have each patient’s charts reviewed and data presented in scientific forums or publications.
Statistical analysis
Descriptive statistics, Student’s t test and χ
2 test were used as appropriate. Sociodemographic data are presented as mean ± standard deviation (SD), and disease and treatment characteristics are described as median and 25th–75th interquartile range (IQR).
For each patient, persistence was evaluated at fixed 6-month intervals during the first 4 years of follow-up, with a total of eight consecutive persistence evaluations (persistence 1 to persistence 8); accordingly, for each patient, duration of persistence varied from 0 to 48 months.
Also, for each patient, the timing of (or follow-up to) first non-persistence period was obtained and scored as 1 month (if first non-persistence was detected at persistence 1 evaluation), 7 months (if first non-persistence was detected at persistence 2 evaluation), 13 months (if first non-persistence was detected at persistence 3 evaluation), 19 months (if first non-persistence was detected at persistence 4 evaluation) and successively (adding 6 months to each consecutive persistence evaluation) up to 43 months if first non-persistence was detected at the persistence 8 evaluation.
Linear regression analysis was used to investigate the impact of the timing of the first non-persistence period and the duration of persistence during the first 4 years of follow-up (independent variables) on DAS28 and HAQ at the fifth year of follow-up (dependent variables). Also, Cox regression analysis was used to investigate whether the timing of the first non-persistence period and the duration of persistence during the first 4 years of follow-up (independent variables) made a contribution to SR and sustained function (SF) at the fifth year of follow-up (dependent variables). Variables included in the different models tested were selected based on their statistical significance in the univariate analysis and also on their clinical relevance (e.g., age and comorbidity). In particular, age was forced into the models based on previous local report, where it has been found a predictor of poor adherence and subsequently worse outcomes [
15]. Correlation between variables to be included was also examined, and the final number of variables was limited by the number of outcomes of interest. Significant variables were finally isolated using stepwise selection. Analysis was repeated for each outcome and in the subpopulation of female patients with RA.
All statistical tests were two-sided and evaluated at the 0.05 significance level. Statistical analysis was performed using the IBM SPSS software program (v.17.0; IBM, Armonk, NY, USA).
Discussion
This study was developed in a well-characterized cohort of Mexican Mestizo patients with early RA disease and substantial comorbidity. Standardized and complete follow-up was performed by the same rheumatologist in a real clinical setting. Follow-up included periodic and prospective evaluations of comorbidity and of treatment persistence. Conventional DMARDs given according to a T2T strategy were the mainstay of treatment, and a substantial follow-up was included. For all these reasons, we consider the population described to be representative of “real-life patients”, so the results presented here have clinical and practical implications and can be generalized to populations with similar characteristics.
The duration of persistence with DMARDs during the first 4 years of follow-up was a predictor of disease activity at the consecutive year (in addition to baseline disease activity and sex) and of disability (in addition to age). Also, persistence duration during the first 4 years (along with lower baseline disease activity) predicted SR at the consecutive year in women. The timing of first non-persistence period (early vs. late) did not impact disease outcomes. Forty-two months of persistence within the first 4 years of follow-up that corresponded to 87.5 % of the complete potential persistence length was the best cutoff for persistence duration to predict SR in the subpopulation of women, who were highly represented in our cohort. Finally, lower disease activity and younger age were the only predictors of SF.
Our study confirms prior literature reviews which highlighted that adherence to and persistence with traditional and biologic DMARDs among patients with RA are suboptimal [
20‐
23]. Persistence impacted outcomes in our population of early RA patients. Viller et al. also showed better outcomes in European patients with early RA (≤5-year disease duration) in whom compliance with drug dosages and dosing times was assessed yearly using a questionnaire [
13]. We recently showed, in the same inception cohort, better outcomes in adherent and persistent patients in two different clinical scenarios: patients with high disease activity and disability and patients in remission or who had low disease activity [
15,
16]. The present study adds additional information to the field of the burden of inadequate therapeutic behavior in patients with RA patients. In addition to our examination of sustained and major outcomes, we show that the duration of persistence impacted disease outcomes, whereas the timing of non-persistence did not. Also, in the female subpopulation, a substantial persistence duration with DMARDs was required to achieve SR (almost 88 % of the entire length of persistence). Several studies have shown that consistent adherence and persistence among patients with chronic conditions drop dramatically after the first 6 months of therapy [
5,
24]. This may be particularly relevant in RA, where aggressive treatment in the early phases of the disease has been shown to prevent structural damage and to favor better outcomes, including a higher remission rate [
25‐
28]. In such a clinical context, it may be intuitive to assume that earlier non-persistence will affect favorable outcomes, although non-persistence timing (early vs. late) did not appear to be a predictor. Interestingly, persistence duration was highly and negatively correlated with non-persistence timing (Spearman’s ρ = −0.8,
p ≤ 0.001), which may explain the relevance of the former aspect of persistence over non-persistence timing. Finally, the cutoff at least 87.5 % of persistence duration to predict SR in the subpopulation of women highlights the patient’s need to adhere closely to prescriptions if major outcome benefits are desired. This cutoff is higher than the arbitrary categories of good and poor compliance, often set at 80 % [
29], although that figure is based on a major and sustained outcome.
In addition to persistence duration, baseline DAS28, age and sex impacted the outcomes of interest.
Lower disease activity at RA diagnosis was found to be a predictor of subsequent disease activity, of SR and of function. Combe et al. [
30] also found that patients with RA for whom therapy with one, two or three or more DMARDs had failed and who had higher disease activity at baseline were less likely to achieve remission after 6 months of golimumab. ten Kloster et al. [
31] identified baseline predictors of achieving satisfactory improvement in pain in 209 ERAPs after 6 months of T2T therapy; among the predictors was 12 or fewer tender joints at baseline. Two additional studies performed in early RA confirmed a lower initial number of tender joints (which may be considered a surrogate of disease activity) as a predictor of remission [
32,
33].
Age was found to be a predictor (in addition to disease activity) of SF. Data from two large inception cohorts identified that older age (in addition to other factors) was associated with increased likelihood of membership in subgroups with worse HAQ progression [
34].
Young et al. [
35] found that among 732 ERAPs, 9.4 % had marked functional loss at 5 years of follow-up, and this adverse functional outcome was more likely to occur in patients older than 60 years of age. Finally, the relationship between disability and demographic and clinical variables was analyzed in 684 patients with inflammatory polyarthritis referred to the Norfolk Arthritis Register. Older age at symptom onset (≥64 years) was one of the factors associated with disability as defined as HAQ of 1 or above [
36].
Finally, female sex was an additional predictor of unfavorable outcome, as previously reported in other populations [
37,
38]. Interestingly, women were highly represented in our cohort, as a female/male ratio of 7–8:1 has been described in Latin American surveys, which is well above the nearly 3:1 ratio reported in the United States and Europe [
39].
The present study has some limitations. The first is that it is a single-center study, which potentially could decrease the generalizability of the results. Second, persistence was defined according to physician-reported and patient-reported discontinuation of treatment, and the threshold used to define non-persistence was arbitrarily chosen. Third, we analyzed the impact of a persistence construct on outcomes, but we did not examine the potential impact of a different construct such as adherence (i.e., compliance with) to medication. Fourth, we defined SF according as HAQ score of 20 or less. A different cutoff (0) on the HAQ-DI has been recommended on the basis that it may be (more) intuitive and sensitive and has face validity [
40]. We repeated our analyses according to such suggestions and obtained similar results. Fifth, the study had a relatively short observation period of 5 years. Sixth, we did not include local controls in whom to assess the progression of functional disability largely explained by the ageing process [
41]. Seventh, we did not include all the potential variables linked to worse disability, such as body mass index [
42]. Finally, medication-taking behavior is a complex process in which frequency, characteristics and predictors may vary over the course of follow-up. Favorable and unfavorable disease courses have both been associated with adherence to therapy [
12,
43], the latter reflecting, perhaps, the fact that active patients regard complex therapy as useless. Additional predictors are related to the disease itself, the population enrolled and the indicated therapy. In such a dynamic environment, the hypothetical causal chain of outcomes–adherence–outcomes is debatable and limits comprehensiveness of the topic.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
ICY participated in the conception and design of the study, performed the statistical analysis and revised the manuscript. She is in charge of database integrity of the early arthritis clinic. VPR participated in the conception and design of the study, performed the statistical analysis and drafted the manuscript. She is in charge of the early arthritis clinic and performed patients’ clinical evaluations. Both authors read and approved the final manuscript.