- Top of page
- Materials and Methods
- Supporting Information
Ulcerative colitis (UC) is a chronic inflammatory disorder of the gastrointestinal tract with a prevalence of 200–250 per 100 000 persons. The disease course is unpredictable and is marked by periods of quiescent disease interspersed with relapses, which may require hospitalisation. The aim of treatment is to maintain remission and avoid corticosteroid therapy. The first line therapy in the treatment of UC is 5-Aminosalicylate (5-ASA) compounds that can be utilised to induce and maintain remission. One of the leading causes of 5-ASA therapy failure is non-adherence. Adherence to 5-ASA outside clinical trials is poor, ranging from 40% to 60%.[3-6]
Risk factors for non-adherence in UC have been explored in the literature and include the following: complex and inconvenient dosing regimens,[5, 7] extremes of age, not being married, male gender, left-sided colitis, multiple prescription medications, fulltime employment, depression and a new patient–physician relationship. Understanding these factors and their impact on non-adherence can help in identifying patients at high risk for non-adherence and developing strategies to counsel them. Non-adherence is associated with increased risk of clinical relapse, healthcare costs and the risk of complications including colorectal cancer.[9, 10]
There are a variety of ways to measure adherence. The most accurate way to measure adherence is through direct observation. Alternative methods include pill counts and electronic medication events monitoring systems.[11, 12] However, the drawbacks of these methods are costly and the difficulty of maintaining them over protracted periods. Adherence also can be measured using pharmacy refill data, one of the earliest papers described and validated pharmacy-based adherence indicators dated back to 1988 by Steiner et al. More recently, patient-reported adherence surveys have been validated against pharmacy-based adherence measurements, and have been used in a large internet-based cohort of inflammatory bowel disease patients to assess their medication adherence.
Adherence in UC patients varies considerably in the literature secondary to different measurement methods, varied follow-up periods and heterogeneous cohorts (geographical/education/demographics).[3-6] In UC, because of the long, lapsing and remitting course of the disease, patients are required to take medication for long periods of time and, in most cases, indefinitely. Over the course of therapy, adherence can vary based on changes in not only disease factors but also financial, psychological and cultural variables. Thus, measuring medication adherence over the long term is a challenging task in light of the multifactorial influences known to play a role in adherence behaviour.
Most of the literature on adherence in UC is from tertiary care centres and might not be nationally representative.[3, 4] To our knowledge, this is one of the first studies of long-term adherence among UC patients using a national database. The aims of this study are to evaluate the overall long-term adherence levels to oral mesalazine in a veterans affairs (VA) healthcare system, to determine the impact of non-adherence on disease flares that require high dose steroids and to evaluate the different methods of adherence measurement using pharmacy data to predict exacerbations.
- Top of page
- Materials and Methods
- Supporting Information
We identified 37 191 UC patients that have been treated in the VA system during the inclusion period. We included 13 062 patients into the analysis. The population was predominately white males (93.7% males, 76.2% Caucasians) with an average age at the first mesalazine prescription of 59 (median 61, range 19–102).
Included patients filled mesalazine on average 21 times (median 15, range 2–125) during a mean of 5.8 years (median 6.1, range 0.49–10) of follow-up and had a mean annual cumulative mesalazine dose of 559 gram/year (g/y) (median 494 g/y, range 2.3–2957 g/y). Average Daily Prescription Dose was 2.7 g/d (median 2.4 g/d, range 0.4–4.8 g/d). Clinical and demographic parameters by flare status are presented in (Table 1).
Table 1. Demographic and clinical factors by flare status
| || ||No flare (%)||Had flares (%)||Total (%)||P-value|
|Age at first oral mesalazine fill (Quartiles)||<50.5||2588 (23)||677 (42)||3265 (25)||<0.001|
|50.5–60.8||2759 (24)||508 (31)||3267 (25)|
|60.8–70||2987 (26)||279 (17)||3266 (25)|
| ||>70||3098 (27)||166 (10)||3264 (25)|| |
|Gender||Male||10753 (94)||1482 (91)||12235 (94)||<0.001|
|Female||679 (6)||148 (9)||827(6)|
|Race||Caucasians||8720 (76)||1233 (76||9953 (76)||<0.001|
|Non Caucasians||1216 (11)||294 (18)||1510 (12)|
| ||Unknown||1496 (13)||103 (6)||1599 (12)|
|Marital status||Not married||3850 (34)||737 (45)||4587 (35)||<0.001|
|Married||7567 (66)||893 (55)||8460 (65)|
|Cumulative annual dose (Quintile) in gram per year||0–20th percentile (2–157)||2437 (21)||175 (11)||2612 (20)||<0.001|
|20–40th percentile (157–370)||2310 (20)||303(19)||2613 (20)|
| ||40–60th percentile (370–613)||2293 (20)||319 (20)||2612 (20)|| |
| ||60–80th percentile (613–892)||2284 (20)||329 (20)||2613 (20)|
| ||80–100th percentile (892–2956)||2108 (19)||504 (31)||2612 (20)|| |
|Geographical region||Northeast||2346 (21)||266 (17)||2612 (20)||<0.001|
|South||4202 (37)||606 (38)||4808 (37)|
| ||Midwest||2720 (24)||380 (23)||3100 (24)|
| ||West||2054 (18)||360 (22)||2414 (19)|| |
|CSA||High adherers ≥80%||5542 (49)||680 (42)||6222 (48)||<0.001|
|Medium adherers 50 to <80%||4504 (40)||683 (42)||5187 (40)|
| ||Low adherers <50%||1342 (11)||258 (16)||1600 (12)|
|MPR||High adherers ≥80%||5036 (44)||669 (41)||5705 (44)||<0.001|
|Medium adherers 50 to <80%||3540 (32)||471 (29)||4011 (31)|
| ||Low adherers <50%||2812 (25)||481 (30)||3293 (25)|| |
|CMG||High adherers ≤20%||3683 (32)||438 (27)||4121 (32)||<0.001|
|Medium adherers >20–50%||4443 (39)||609 (38)||5052 (39)|
| ||Low adherers >50%||3262 (29)||574 (35)||3836 (29)|
Mean adherence indicators scores for the entire population were 74% (median 79%) for CSA, 69% (median 75%) for MPR and 37% (median 33%) for CMG. After classifying the cohort into high, medium and low adherers according to the aforementioned criteria, the unadjusted hazards ratios for adherence indicators showed that patients with low adherence were more likely to experience a flare over the follow-up period. Low adherence was associated with 1.17 times increased risk of UC flares when using MPR (P = 0.008), 1.34 when using CMG (P < 0.001), 1.6 when using CSA (P < 0.001).
Using Cox regression analysis, all the aforementioned variables were significant predictors of disease flares except for gender (Table 2). All the three adherence indicators predicted disease flares well while in the full regression model, with CSA showing the best trend (dose-response) pattern (Figure 2). Low adherence was associated with 1.7 times increased risk of flares when using MPR (Figure 3), 1.8 when using CMG, 2.8 when using CSA. By classifying adherence into three levels and using CSA in the final model, we improved the predictive value of the index markedly in the univariate level and slightly in the multivariate level as shown in (Table 3).
Table 2. Predictors of ulcerative colitis flares, Cox regression analysis
| ||HR||95% CI|| P-value|
|Age at first oral mesalazine fill||<50.5||1 (reference)|| || |
|Gender||Male||1 (reference)|| || |
|Race||Caucasians||1 (reference)|| || |
|Marital status||Not married||1 (reference)|| || |
|Geographical region||NE||1 (reference)|| || |
|Cumulative annual dose||0–20th percentile||1 (reference)|| || |
|CSAa ||High adherers ≥80%||1 (reference)|| || |
|Medium adherers 50% to <80%||1.346||1.2–1.5||<0.001|
|Low adherers <50%||2.814||2.33–3.4||<0.001|
|MPRa ||High adherers ≥80%||1 (reference)|| || |
|Medium adherers 50% to <80%||1.016||0.9–1.15||0.802|
|Low adherers <50%||1.772||1.51–2.1||<0.001|
|CMGa ||High adherers ≤20%||1 (reference)|| || |
|Medium adherers >20–50%||1.064||0.94–1.21||0.341|
|Low adherers >50%||1.870||1.59–2.2||<0.001|
Table 3. Predictive value of adherence indicators
| ||UV R2||MV R2|
|CSA||2 levels: high adherers ≥80%, low adherers <80%||0.0084||0.345|
|3 levels: high adherers ≥80%, medium adherers 50% to <80%, low adherers <50%||0.014||0.357a|
|MPR||2 levels: high adherers ≥80%, low adherers <80%||0.0003||0.336|
|3 levels: high adherers ≥80%, medium adherers 50% to <80%, low adherers <50%||0.0045||0.345|
|CMG||2 levels: high adherers ≤20%, low adherers >20%||0.0037||0.345|
|3 levels: high adherers ≤20%, medium adherers >20% to 50%, low adherers >50%||0.0092||0.35|
Other predictors of UC flares included the following: (i) Age at which first mesalazine was dispensed was negatively associated with the risk of UC flare; this is consistent with previous reports about the reduction in disease activity with advanced age.[21, 22] (ii) On a univariate level, females were more likely to encounter UC flares, this effect was insignificant in the multivariate level most likely due to the small percentage of females (6.3%). (iii) Caucasians showed less flares compared with other racial groups. (iv) Those who reported that they were married during most of the observational period had less risk of UC flares compared with those who reported having no partner (separated, divorced, widowed, never married). (v) Interestingly enough, patients from the west geographical area of the US showed the highest risk of UC flares.
- Top of page
- Materials and Methods
- Supporting Information
This retrospective cohort showed that low adherers were about two times more likely to suffer from disease flares that necessitated high-dose oral or parenteral steroids in the median of 6 years of follow-up after initiation of oral maintenance mesalazine. Furthermore, the study compared between three pharmacy data-based indicators of adherence in terms of their predictive values of disease flares on multivariate level with CSA showing the best trend pattern.
Overall mean adherence rates were 74%, 69% and 37% for CSA, MPR and CMG respectively. We notice that MPR and CSA represent medication possession over the follow-up period. However, CMG represents the medication gap and thus it is approximately the reciprocal of MPR and CSA. Long-term high adherers' percentages were 48%, 44% and 32% based on CSA, MPR and CMG respectively. These figures are consistent, but slightly lower than the previously reported range (40–60%).[3-6] A major proportion of the included cohort were classified as medium adherers (ranged from 31% as identified by MPR to 40% by CSA), a group that used to be classified as low adherers using a single 80% threshold in previous literature.[3, 4] We chose to report our results using the adherence indicators with three levels (high adherers ≥80%, medium adherers 50–80% and low adherers <50%) as this increased the predictive value of the model and showed a clear dose-response pattern, especially with CSA. Furthermore, more recent research has reported adherence in three levels.[23, 24]
Across all of the adherence parameters, the high-adherer and medium-adherer groups had almost the same (MPR, CMG) or slightly different (CSA) risk of flares in favour of high-adherers patients, whereas low-adherer patients showed increased in the risk of flares that ranged from 70% (MPR) up to 280% (CSA) P < 0.001 (Table 2).
In previous literature, researchers have examined the association between adherence and relapse defined as mild UC flare symptoms based upon clinical features. We investigated the effect of low adherence on developing a more severe UC flare that was associated with high doses of oral or parenteral steroids, which to our knowledge has not been previously reported. Our definition of UC flares is consistent with the American College of Gastroenterology guidelines, as high-dose systemic steroids are indicated in the management of relapses in which maximal doses of oral and rectal 5-ASA or topical steroids have failed. Fluctuation in steroid utilisation obtained from the patients' pharmacy data has been previously utilised to detect relapses in inflammatory bowel disease.
This study used the ACD as a surrogate for disease severity. Higher daily dose of mesalazine (4.8 g/d) is indicated in moderate disease, whereas lower dose (2.4 g/d) is enough for mild disease.[2, 26] This will result in higher ACD for moderate disease and lower ACD for mild disease, it was anticipated that the first subgroup will have an increased risk of UC flares. The data confirmed this assumption.
In a recent nationwide survey of gastroenterologists, 77% of the responders stated that they screen for adherence among inflammatory bowel patients; however, only 19% of them use accepted screening measurements (i.e. pill count, refill rates or validated adherence surveys). The majority of gastroenterologists were using interviewing to assess adherence, a method considered the least effective. These results highlight the importance of providing clinicians with feasible and validated tools to help identify patients at risk for low adherence to UC medications and prevent disease flare.
Adherence indicators can be reliably calculated from pharmacy data; these formulas can be uploaded to a website or to a smart phone application to help the treating physician to get objective scores for the patient adherence by following his pharmacy refill. Furthermore, with the continuous development of the electronic medical care records, large integrated health systems (as the VA and Kaiser) or health insurance programmes such as managed Medicare groups can help physicians in tracking patients adherence lapses over long follow-up time through incorporating these indicators as part of electronic medical records that are accessible by the treating physician and updated periodically as the patient fill–refill his/her prescription.[28, 29]
Once adherence is determined, there are several ways that adherence can be improved. Establishing good therapeutic relationship is critically important in improving adherence, and patients must be a significant part of medical decision regarding their treatment.[29, 30] A simplified drug delivery system and less intrusive dosing is another important factor that has been investigated.[31, 32] Scheduling frequent follow-up visits at 2 weeks, 3 months and then each 6 months after the initiation, if prescribing mesalazine therapy[8, 33] may aid in increasing adherence. During these visits, routine pill checks can be very helpful through asking the patient to bring in his medication to be inspected by the physician to assess the extent of home utilisation.[29, 34] Counselling about the importance of adherence in periods of remissions should be given in the follow-up visits.
The major strength of our study lies in the fact that (to our knowledge) this is the first report of nationwide data regarding adherence to mesalazine and the role of adherence in predicting disease flares. This study is also novel in that it had a relatively longer follow-up period compared with the study conducted by Kane and colleagues. In addition, we used an objective indicator of our outcome (pharmacy fill) rather than subjective definitions (frequency of bowel movements, urgency or pain), which are susceptible to individual variations. Pharmacy data allows the analysis of refill patterns in real life, outside the clinical trials in which adherence is expected to be higher due to the fact that patients might change their adherence because of the increased attention and monitoring provided in clinical trials.[7, 35] Furthermore, the study is unique because it is one of the first analyses that compared different indicators of adherence in predicting the outcome over long-term follow-up. We believe that the data captured most of the drug exposure as the veterans (depending upon their level of disability and income bracket) either receive their medication free of charge or have substantially lower co-pay at the VA pharmacy even if mesalazine was prescribed by an outside physician.
The investigators acknowledge some limitations of the present study as it is a retrospective database analysis. We have not looked at clinical parameters of disease exacerbations, but rather used pharmacological data as a proxy. We chose to use such steroids fills as a proxy for a flare as this information is readily available in the pharmacy database with high level of validity and objectivity. Mayo score for UC activity is a widely acceptable method to assess disease activity; however, in community-based general practices, it is rarely used. It has been well recognised in the literature that medication possession ratio and other pharmacy-based adherence indicators have inherent limitations and tend to overestimate adherence on average.[13, 37] Even with the presence of such limitation, non-adherence as measured by pharmacy data is significantly associated with disease flare, this indicates that the observed effect (two times more likely in this study) might be an underestimation of the actual non-adherence effect.
We were able to detect adherence level by tracking pharmacy refill records, but we cannot be certain about the reason for non-adherence. It is possible that patients may have stopped taking the medication (resulting in low adherence) because of lack of medication efficacy. In this case, low adherence could be a marker of exacerbation in patients who stopped taking the drug because it did not prevent worsening of symptoms rather than a cause of exacerbation. Further research is needed to better understand patterns of low adherence that might differentiate between true low adherence and discontinuation of drugs due to low effectiveness.
As these data are from the VA healthcare system, which encompasses a population predominately comprised of Caucasian males, this can limit the external validity of the study results. Our starting date was 1 October 2001, and we included all those who filled oral mesalazine prescription for the first time after this date. The investigators did not have access to data that predated the start of the study period. Our study is susceptible to data entry errors because it is an administrative dataset. We do not anticipate these errors to create any significant bias as previous studies have shown that the VA database is robust,[38-40] and because these errors would be random they should not alter the direction of the observed adherence effect. Furthermore, all included patients had a combination of ICD-9 codes for UC and a minimum of two fills of mesalazine. This increased the captured diagnosis specificity as shown in previous work in other chronic diseases as hypertension[18, 41] and chronic obstructive pulmonary disease.
Lastly, we were able to capture the high-dose steroid filling, but we cannot ensure that the intention for the prescribed drug was for acute exacerbation of UC. Other diseases are treated with oral corticosteroids, such as rheumatoid arthritis and systemic lupus. However, we anticipate that such fills would affect both dose range groups equally and would be a random error.