inflammatory bowel disease patients are frequently nonadherent to scheduled induction and maintenance infliximab therapy: a Canadian cohort study

patients are frequently to scheduled induction and maintenance A BACKGROUND: Adherence to maintenance medication regimens in inflammatory bowel disease patients has traditionally been poor. Although infliximab has demonstrated efficacy in inducing and main-taining disease remission, adherence to regularly scheduled infliximab infusions is required to maintain therapeutic trough drug levels and prevent the development of anti-infliximab antibodies. OBJECTIVES: To characterize patient adherence to regularly scheduled induction and maintenance infliximab infusions. METHODS: A retrospective cohort study was conducted evaluating adult outpatients with Crohn disease or ulcerative colitis on an induction or maintenance regimen of regularly scheduled infliximab from 2008 to 2010 at the University of Alberta (Edmonton, Alberta). Nonadherence was defined by a discrepancy of >72 h between the scheduled date of infusion and the actual date of administration. Patients were defined as nonadherent if they received <80% of their infliximab infusions per schedule. RESULTS: A total of 215 patients (173 Crohn disease, 42 ulcerative colitis) met the inclusion criteria. Patients received a median of 12.0 infliximab infusions (interquartile range 7.0 to 13.0) during the study period; 412 induction and 1837 maintenance infliximab infusions were administered. Of 140 patients, 109 (77.9%) were adherent to their infliximab induction regimen, while 68 of 215 (31.6%) were adherent to their infliximab maintenance regimen. One hundred ninety-eight of 215 (92.1%) patients received at least one delayed maintenance infliximab infusion and 20 (10.1%) received maintenance infusions, on average, >1 week late. three-quarters of patients are adherent to infliximab induction therapy, fewer than one-third remained adherent to their scheduled maintenance infliximab regimen. these are all surrogate measures for assessing medication adherence. Thus, in the present study, we assessed the ‘true’ adherence to regularly scheduled infliximab induction and maintenance therapy using manually searched medical chart review of scheduled and administered infliximab infusions. We define the proportion of patients who were nonadherent to infliximab therapy, characterize the incidence of early and delayed infusions and describe the variance around scheduled infusions. In a secondary analysis, we examined risk factors predicting nonadherence to long-term maintenance infliximab therapy. multivariate (OR 1.77 [95% CI 1.01 to 3.11]; P=0.046)

UC (6), its 'real-life' clinical effectiveness may be hampered by poor adherence. Specifically, secondary loss of response during maintenance therapy is a common phenomenon encountered in more than onehalf of IBD patients receiving anti-TNF agents (7), with development of antidrug antibodies playing a central role. In patients treated with infliximab, missing or delaying regularly scheduled infusions may contribute to drug immunogenicity (8), decreased serum trough drug levels (9) and, ultimately, adverse clinical outcomes. In fact, frequently missing or delaying infliximab infusions can mimic an episodic administration regimen, which compared with fixed scheduled infusion protocols in subgroup analysis of the A Crohn's Disease Clinical Trial Evaluating Infliximab in a New Long-Term Treatment Regimen (ACCENT I) trial resulted in increased likelihood of antidrug antibody development (28% versus 9%), reduced likelihood of mucosal healing (18% versus 44%), increased hospitalizations (38% versus 23%), and reduced clinical response and remission rates (10).
Several authors have evaluated adherence to infliximab therapy in patients with IBD. Recently, Lopez et al (11) reviewed 13 studies evaluating adherence in this cohort and demonstrated varying rates of adherence ranging from 36% to 96%. Nevertheless, there are several limitations to the results from these studies. Most findings are extrapolated from pharmacy refill data or administrative databases (12), medical insurance claims (13) or self-reported questionnaires (14), which are inherently subject to under-reporting, recall bias and coding inaccuracy. Furthermore, they may not reflect true medication adherence because these types of data do not capture administration of delayed or early infusions. Defining nonadherence is also particularly challenging and highly heterogeneous: previous authors have used the number of missed or delayed infusions, missing refill prescriptions, medication possession ratio <80% or inadequate total infusions over a defined time period; however, these are all surrogate measures for assessing medication adherence.
Thus, in the present study, we assessed the 'true' adherence to regularly scheduled infliximab induction and maintenance therapy using manually searched medical chart review of scheduled and administered infliximab infusions. We define the proportion of patients who were nonadherent to infliximab therapy, characterize the incidence of early and delayed infusions and describe the variance around scheduled infusions. In a secondary analysis, we examined risk factors predicting nonadherence to long-term maintenance infliximab therapy.

Study design and setting
The present retrospective cohort study was performed using data collected from CD and UC outpatients receiving maintenance infliximab therapy from August 2008 to August 2010, at the University of Alberta (Edmonton, Alberta). Patients were identified from the Division of Gastroenterology inflammatory bowel disease electronic database.

Patient population
Patients were eligible for inclusion if they met the following criteria: adult (>17 years of age); diagnosis of CD or UC; receiving either initial induction regimen of infliximab 5 mg/kg at weeks 0, 2 and 6; or advanced onto a regularly scheduled maintenance outpatient infliximab regimen. Disease phenotype, infliximab dose escalation, concurrent medical therapy and duration of infliximab before the study were not contraindications to inclusion.

Outcomes and definitions
The primary objectives of the present study were to: determine the proportion of IBD patients who are adherent to a regularly scheduled infliximab infusion protocol; and to characterize the incidence of early and delayed infliximab administration. The secondary objective was to assess risk factors predicting infliximab nonadherence.
For the present study, nonadherence to a single infliximab infusion was defined as a discrepancy of >72 h between the scheduled date of infliximab infusion and the actual date of administered infusion. There is no previous widely accepted definition for adherence to an individual infliximab infusion -a 72 h cut-off was decided on because this creates a six-day window for the patient to receive their infusion per protocol (three days before and three days after their scheduled infusion date). It was agreed that this was a sufficient time frame to accommodate statutory holidays, infusion clinic availability and minor personal reasons that may briefly delay or expedite the infusion schedule. A missed infliximab infusion was defined during maintenance therapy if the delay between administered infusions was >12 weeks.
Overall patient nonadherence was defined if <80% of infliximab infusions were received adherently. Thus, patients requiring a temporary delay in infliximab infusions due to infection, perianal disease, holidays, etc, were not deemed to be nonadherent to therapy. This cut-off was based on previous studies involving IBD patients in which definitions of nonadherence included medication possession ratio <80% and receiving <7 infliximab infusions in first year of therapy (seven infusions represents approximately 80% of the potential infusions for the first year of therapy, including three induction and six maintenance infusions) (15,16).
The scheduled date of infliximab infusion was determined by the patient's optimal infliximab infusion regimen (ie, the next scheduled infusion would be eight weeks from the previous administered infusion if the patients is receiving maintenance therapy every eight weeks). For induction therapy, the scheduled infusion dates were at weeks 0, 2 and 6. For maintenance therapy, the scheduled infusion dates were every eight weeks from the previous infusion. For patients requiring dose escalation during maintenance therapy and receiving infusions more frequently than every eight weeks, the scheduled date of infusion was adjusted accordingly.
Study follow-up time was defined from the date of first administered infusion to the date of last infusion received during the study inclusion period; total duration of infliximab therapy was determined from the date of first infliximab induction infusion.

Data collection
Data were extracted from official infliximab infusion records by two of the authors (CJE and GG) using a standardized case report form. These records included baseline patient demographic data, lifetime infliximab infusions received, infusion frequency (induction versus maintenance, maintenance infusion every eight weeks versus escalated/de-escalated schedule), actual date of administered infusion and previous medication exposure. Data were reviewed by two of the authors (CM and RNF).
Baseline patient data collected included sex, age, IBD diagnosis (CD versus UC), previous medical therapy (mesalamine or 5-aminosalicylate products, azathioprine or methotrexate) and total lifetime infliximab infusions received before inclusion. Subsequently, the scheduled infusion date was identified as described above and actual date of administered infliximab using the official infusion record. Clinical outcomes, including disease flares, hospital admissions and IBDrelated surgeries, were not available in the infusion records (these outcomes were outside the scope of the present study).

Statistical methods
For continuous variables, mean, SD, median and interquartile range (IQR) were calculated. The variance in days between date of actual administered and date of scheduled infusion was calculated for each infusion.
A post hoc sensitivity analysis was performed. Because there is no consensus definition of nonadherence to an individual infliximab infusion, adherence to maintenance infliximab therapy was determined if the definition of adherence was relaxed to allow a discrepancy between scheduled and administered infusion of >7 days (from three days). A sensitivity analysis was also performed whereby patients receiving <6 maintenance infusions were excluded; this enhanced the power of the analysis to better account for temporary causes of nonadherence.
Univariate and multivariate logistic regression analyses were performed to examine the association between patient variables and nonadherence, with results expressed as OR with 95% CI. Age (defined as >40 and ≤40 years), sex, IBD diagnosis (CD versus UC), previous medication exposure, requirement for infusion escalation and lifetime infliximab infusions (defined as ≤7 infusions, eight to 12 infusions and ≥13 infusions) were a priori included in the regression model. Fistulizing disease was not included in the logistic regression model because it would not apply to patients with UC.
Statistical analysis was performed using SPSS version 22.0 (IBM Corporation, USA).

Baseline patient characteristics
Patient demographics are summarized in Table 1. A total of 215 patients (173 CD, 42 UC) met the inclusion criteria. One hundred forty (65.1%) patients received infliximab induction dosing during the study period and all patients received at least one maintenance infusion of infliximab. Patients received a median of 12.0 infliximab infusions (IQR 7.0 to 19.5) lifetime and were followed for a median duration of 80.1 weeks (IQR 38.7 to 100.9 weeks) during the study.
Forty-one (23.7%) CD patients had fistulizing disease; 144 patients (67.0%) required dose escalation of infliximab to more frequent infusions than every eight weeks (escalations to maintenance infusion every four to seven weeks, depending on the clinical circumstances and clinician judgement). Almost all patients had previous exposure to mesalamine, azathioprine or methotrexate.

Adherence with infliximab induction therapy
Adherence to infliximab induction infusions (5 mg/kg at weeks 0, 2 and 6) is summarized in Table 2. One hundred forty patients received a mean (± SD) of 2.9±0.3 infliximab induction infusions. A total of 412 infliximab induction infusions were administered.
Of 140 patients, 109 (77.9%) were adherent to their infliximab induction infusions and received, on average, 2.7±0.6 induction infusions with adherence. Sixty-five (46.4%) patients received at least one delayed induction infusion, but mean delay time was short (1.5±1.8 days). A similar number of patients received at least one early infliximab induction infusion (n=67 [47.9%]).

Adherence to infliximab maintenance therapy
Adherence to infliximab maintenance therapy is summarized in Table 3. Two hundred fifteen patients received a total of 1837 infliximab maintenance infusions. Over the two-year study period, patients received a median of 9.0 maintenance infusions (IQR 4.0 to 13.0).
Sixty-eight of 215 (31.6%) patients were adherent to >80% of their infliximab maintenance infusions. On average, patients were adherent   Mean variance from the scheduled infliximab maintenance infusion was 4.0±4.6 days. As such, a patient would be expected to accrue a dropped infliximab maintenance infusion every 13.9 infusions (112.0 weeks) based on the standard maintenance regimen of infliximab infusion every eight weeks. Among the 20 patients experiencing a mean delay in administration of infliximab >7 days, a dropped infliximab maintenance infusion would be accrued every 33.3 weeks.

Sensitivity analysis
Exclusion of patients receiving <6 maintenance infusions in sensitivity analysis did not change overall adherence to therapy: only 46 of 145 (31.7%) patients in this cohort were adherent to their maintenance regimen. When the definition of nonadherence to an individual infliximab infusion was relaxed to allow a discrepancy between scheduled and administered infusion date of >7 days, 101 of 215 (46.5%) patients were adherent to >80% of their infliximab maintenance infusions.

Predictors of nonadherence during maintenance therapy
Univariate and multivariate ORs for risk factors predicting nonadherence to infliximab maintenance therapy are summarized in Appendix 1. Only male sex was associated with slightly increased risk for nonadherence in both univariate (

DISCUSSION
Infliximab has a central role in the management of patients with moderate-to-severe IBD and, although it has demonstrated efficacy in randomized clinical trials, real-life clinical effectiveness may be limited by poor adherence. Previous studies evaluating anti-TNF therapy compliance in this cohort have primarily used administrative data and surrogate markers of adherence such as medication possession ratio (15). While these measures capture substantial lapses in adherence, they lack the resolution to identify nonadherence on an infusion-to-infusion basis. Here, we present a large Canadian retrospective evaluation of 215 IBD patients treated with infliximab, and assess through detailed chart review, the true adherence to both induction and maintenance infliximab infusion regimens. We found that while 77.9% of patients were adherent to induction infusions, fewer than one-third of patients had sustained infusion adherence and 10% will experience delayed infusions, on average, >1 week from the ideal scheduled dosing protocol during maintenance therapy.
Adherence to infliximab was first examined by Kane et al (12) in 2006; they described 96% adherence to therapy after retrospectively reviewing 1185 infusions using administrative and pharmacy refill data. This incredibly high adherence rate relates to the use of 'no show' visits as the definition for nonadherence. Similarly, we found a very low incidence of completely missed infusions (1.7%). Other authors, using administrative database definitions for nonadherence including medication possession ratio <80% or <7 infusions during the first year of treatment, have also found varying rates of adherence, ranging from 57.1% to 79.8% (13,14,16).
In contrast to previous studies in which nonadherence was defined only by completely missed infusions, we believe that frequently delayed infliximab infusions are a reflection of poor patient compliance with the treatment regimen. Thus, we used a generous definition of adherence to both individual infusions (within six days of scheduled infusion) as well as overall adherence (80% of infusions received per schedule) to describe patient adherence to infliximab therapy with greater resolution. Other authors have used similarly constructed composite definitions of nonadherence (17). Our definition for nonadherence accounts for common barriers to compliance including patient inconvenience or statutory holidays (patients can schedule their infusion for the next available business day if the scheduled infusion date falls on a weekend or holiday and still be considered adherent to therapy) as well as unpredictable or excusable reasons for delaying infliximab infusions such as personal emergencies, infections or drug adverse effects.
Overall, adherence to anti-TNF therapy, given its necessity for scheduled intravenous administrations, may be slightly better compared with adherence to oral maintenance therapies including 5-aminosalicylate products and immunomodulators. A systematic review of oral therapy compliance in IBD patients by Jackson et al (1) found that most studies report approximately 30% to 45% nonadherence rates. This disparity may reflect differences in disease activity (highly active disease requiring biologic therapy may prompt better adherence), route and schedule of administration (intravenous versus oral, every eight week maintenance versus daily therapy), and more frequent clinician follow-up and monitoring. However, it should be highlighted that even among patients on infliximab therapy, there exists a significant rate of nonadherence. Exploring reasons for nonadherence was outside the scope of the present study but has been evaluated elsewhere (18).
Few studies have directly evaluated delays to individual scheduled infliximab infusions and are predominantly reported in abstract form. In a small retrospective cohort of 82 IBD patients on infliximab, Duncan et al (19) found that infusions were postponed in 14 of 82 (17%) patients over 12-month follow-up. Similarly, another abstract by Angelucci et al (20) found that 19.1% of infusions were delayed, median delayed administrations per patient was three infusions, and the median delay was 9.6 days (range one to 35 days). Both authors cite 'technical' reasons, including patient forgetfulness, inconvenience and intentional nonadministration as the reason for delayed infusion in >90% of cases. In our cohort, we report even higher rates of delayed infusions, delayed infusions per patient and total accumulated late days, most likely related to our increased sample size and duration of follow-up. Specific indication for delayed infliximab administration is not part of our infliximab infusion record. However, medically appropriate indications for delaying infusions, such as adverse events or infections, are accounted for by our definition of nonadherence. These specific indications for delaying infusions would presumably only affect a minority of the patient's total infusions and, thus, they should still be able to receive >80% of their infusions with adherence. In fact, Angelucci et al (20) report that only 3.3% of nonadherent infusions were due to adverse events in their cohort.
Delays and nonadherence with infliximab therapy may have significant implications on clinical outcomes in patients with IBD. In particular, we hypothesize that nonadherence has a substantial impact on long-term maintenance of clinical response and secondary loss of response. In both prospective (10) and retrospective (21) cohorts, it has been shown that patients managed with regularly scheduled infliximab compared with episodically administered therapy have improved outcomes for mucosal healing, disease-related hospitalizations and surgeries, and maintained clinical response. While all patients in our cohort were on scheduled therapy, prolonged and recurrent delays to infusion can mimic a more episodic pattern of administration. Unfortunately, clinical outcomes, including hospitalizations, surgeries and disease flares, are not tracked in our infliximab infusion records. Thus, we are unable to correlate infliximab adherence with disease outcomes, although this should be evaluated in future studies.
The mechanism behind primary and secondary loss of response to infliximab is influenced by subtherapeutic serum trough drug levels predisposing to the formation of neutralizing anti-infliximab antibodies (22). Recently, in a post hoc analysis of the ACCENT trial, Cornillie et al (23) found that CD patients with week-14 infliximab trough level ≥3.5 μg/mL were 3.5 times more likely to achieve durable sustained response through week 54 of therapy (OR 3.5 [95% CI 1.1 to 11.4). Similar evidence exists in UC, in which detectable serum infliximab drug levels are significantly associated with clinical remission and decreased risk for colectomy (24). We hypothesize that delayed and nonadherent infliximab infusions increase the time during which patients have subtherapeutic infliximab trough levels and are predisposed to forming antidrug antibodies. Therapeutic drug monitoring with infliximab trough levels and antibody prevalence was not clinically available during the time of the study in our jurisdiction. Future studies should evaluate the relationship between delays in administered infliximab, therapeutic drug monitoring and, ultimately, clinical outcomes.
Nonadherence to infliximab maintenance therapy also has significant implications on health care utilization (25). Using the Integrated Health Care Information Service claims database, Kane et al (16) found significantly increased health care utilization costs among nonadherent CD patients on infliximab, who were 2.5 times more likely to require CD-related hospitalization, had 90% greater CD-related medical costs and 115% greater hospitalization costs compared with infliximab-adherent patients. Wan et al (26)  Early infusions did occur in our cohort, although this was less common and patients accumulated fewer early days compared with delayed infusions (6.9 early versus 25.8 late days). Although earlier infusion is likely preferable to delayed infusion to ensure ongoing maintenance of clinical response, excessively early administration may result in elevated infliximab trough levels, which have been shown to be associated with adverse events including dermatological reactions (27).
Definitive predictors of infliximab adherence have not been consistently reported across multiple studies (11,28). Kane et al (12) previously found that female patients were more likely to be nonadherent to infliximab but interestingly, the opposite trend was observed in our cohort, and nonadherent behaviours have been well described in male patients by other authors (29,30). Whether this relates to differences in patient population, financial drug coverage, temporal trends in infliximab use or sampling error is unclear. Kane et al (12) also reported that patients were more likely to be nonadherent for an infusion >18 weeks after induction; although not statistically significant in our regression analysis, the higher rate of adherence observed during induction therapy in our cohort compared with maintenance therapy suggests attrition of patient adherence behaviours over time. Additionally, we found that requirement for infusion escalation trended toward increased nonadherence; we hypothesize that common 'technical' reasons for delayed infusions such as forgetfulness and inconvenience are magnified in patients requiring more frequent infliximab infusions.
There were several limitations to the present study. First, although detailed review of infusion records enabled us to analyze infliximab drug adherence with more resolution than previous studies, we were limited in collecting other clinical parameters, including disease phenotype (particularly for CD), severity of disease and disease activity, clinical response and remission status, concurrent medical therapy and noninvasive biomarkers of inflammation (eg, C-reactive protein). This limited the power of our regression analysis for identifying predictors of nonadherence. A second limitation was that reasons for nonadherence were not available based on the infusion record. Although previous authors have defined patient-related 'technical' factors as being the predominant reason for nonadherence, we are unable to confirm this here. Medical reasons for delaying infliximab infusions such as opportunistic infections could not be captured in our data set, but we have attempted to minimize this potential misclassification bias using a composite definition of nonadherence as described above. While the criterion for defining adherence to an individual infliximab infusion may appear strict (within 72 h of schedule), this actually creates a six-day window for the patient to receive their infusion per protocol. Even when the definition for adherence was extended to seven days in the sensitivity analysis, fewer than onehalf of patients were adherent.
Third, data are available to 2010 but may be less generalizable because experience with infliximab has evolved. Additionally, there may be referral bias because the present study was conducted at a tertiary care centre. In less specialized community gastroenterology practices, adherence to therapy may be even worse, particularly if infusion sites are not centralized, potentially amplifying the effects of preventable reasons associated with poor compliance (eg, inconvenience).
Finally, although we recognize that therapeutic drug monitoring will become a critical component of IBD management moving forward, data regarding serum infliximab trough levels and antibodies to infliximab were not available in our jurisdiction at the time of the study. Furthermore, clinical outcomes relevant to patients with IBD, including hospitalization, surgery and disease flare, are not recorded in our infliximab infusion database. These are major outcomes of interest associated with anti-TNF therapy adherence but were outside the scope of the present study and should be evaluated in future studies.

CONCLUSION
Although three-quarters of IBD patients were adherent to infliximab induction therapy, only one-third sustained adherence during maintenance regimens. Frequently, patients on infliximab experience delays to infusion, and future studies should evaluate whether these delays result in altered drug immunogenicity, increased antiinfliximab antibody formation or adverse patient outcomes.
DISCLOSURES: RNF has served as a consultant and advisory board member for AbbVie Canada Inc and Janssen Canada Inc, and received clinical/basic research grants from AbbVie and Janssen. KIK, LAD and BPH have served as consultants and speakers for AbbVie and Janssen. The remaining authors have no financial disclosures or conflicts of interest to declare.

FUNDING:
This study was funded, in part, by the Alberta Innovates Health Solutions, Alberta Inflammatory Bowel Disease Consortium, and the Centre of Excellence for Gastrointestinal Inflammation and Immunity Research (CEGIIR).
AUTHOR CONTRIBUTIONS: RNF is acting as the article guarantor and contributed to study design, data collection, data analysis and manuscript editing. CM contributed to data analysis, manuscript drafting, and editing. CJE contributed to study design, data collection and data analysis. GG contributed to data collection and data analysis. VH, DKF, KIK, LAD and BPH contributed to manuscript editing. All authors have approved the final version of the manuscript, including the authorship list.