Discussion

In this study of two large national databases, we observed that discontinuation of 5-ASA after initiation of anti-TNF therapy in patients with UC did not increase the risk of adverse clinical outcomes, including corticosteroid use, hospitalisation and surgery. These findings held after controlling for potential confounders and baseline healthcare utilisation. Importantly, results were consistent across two independent cohorts from different countries. This observation is clinically relevant as there are currently no published data on whether patients with UC who initiate anti-TNF therapy can safely discontinue 5-ASA therapy. To our knowledge, no prior publications have examined the impact of stopping 5-ASA treatment in patients requiring escalation to anti-TNF therapy. Given the potential benefits of discontinuing 5-ASA drugs, including decreased costs, reduced risk of medication side effects (although rare) and less polypharmacy, our study is the first to provide data supporting 5-ASA discontinuation after escalation to anti-TNF. However, clinical trials are needed to further establish this as a preferred medication de-escalation strategy in UC.

Our study provides new information on the often-encountered clinical scenario of patients with UC already on 5-ASA who then require escalation of therapy. Prior literature investigating the impact of 5-ASA discontinuation in UC is limited. Previous studies have been on adherence or discontinuation in patients already in remission. A prospective cohort of patients with UC in remission on 5-ASA found that patients who were not adherent to their medication had more than a fivefold increased risk of flare, defined by recurrence of symptoms, compared with adherent patients.11 An analysis of health insurance claims data found that patients with UC who were non-adherent to 5-ASA, defined as a medication to possession ratio less than 0.8, were significantly more likely to be hospitalised over a 12-month follow-up period.12 One clinical trial examined the impact of 5-ASA continuation on patients with UC who were in clinical, endoscopic and histological remission for at least 1 year on mesalamine or sulfasalazine and found that patients randomised to placebo were significantly more likely to have a relapse over 12 months compared with those randomised to mesalamine.6 A retrospective study, published in abstract form only, similarly found that over half of patients with UC in remission on 5-ASA who withdrew maintenance 5-ASA experienced a new flare within a median of 27 months.13 These studies suggest that 5-ASA discontinuation or non-adherence increases the risk of adverse clinical outcomes but did not address the impact in the setting of concomitant immunosuppressant treatment. One prior retrospective study did investigate the effect of 5-ASA treatment in patients with IBD with clinically quiescent disease on azathioprine and found no benefit to concomitant 5-ASA in maintaining remission.14 A recent study pooling anti-TNF clinical trial data examined if concomitant 5-ASA treatment at the start of the trial improved outcomes but found no difference among patients on or off 5-ASA.15 This is a related question to our study but is distinct from examining the impact of stopping 5-ASA among patients who were previously on treatment with these agents. Our results suggest that continuation of previously prescribed 5-ASA does not provide benefit after escalation to more potent anti-TNF biologic medications. We suspect that this is most likely because the anti-inflammatory effects of anti-TNF alone (or in combination with an immunomodulator) are more than sufficient for controlling UC with no additive or synergistic effect from co-treatment with 5-ASA, likely reflecting the higher potency of biologic agents. Another potential explanation for our observation is that dominant molecular pathways in moderate to severe UC are more directly targeted by anti-TNF as opposed to 5-ASA treatments.

Since clinical trial evidence supports improved outcomes with combination therapy in UC, we explored if concomitant immunomodulator therapy may modify risk of adverse events after 5-ASA discontinuation.16 In addition, some prior studies have suggested that 5-ASA interacts with immunomodulator metabolism and can increase 6-thioguanine levels.17–19 The presence or absence of concomitant immunomodulators did not generally modify the risk of major adverse outcomes when discontinuing 5-ASA. In the US cohort, there was an apparent reduction in risk of intestinal surgery when discontinuing 5-ASA in patients who initiate biologic monotherapy. This finding was not seen in the Danish cohort. This statistically significant finding in the US cohort may be explained by an artefactual effect of the low incidence of intestinal surgeries, where the relative (proportional) difference is great, but the absolute difference is very low (1.0 per 100 person-years).

This study has several strengths. First, the study included two completely independent population cohorts that permitted cross-validation of results. Although we present the results within the same study, the cohorts were analysed separately by two separate statisticians to minimise risk of bias. The rigour of this approach and the consistency in the results from two different countries provide greater confidence in the validity of the findings. Second, the use of large population-based cohorts provided the ability to select many patients in the specific clinical scenario of interest as well as improve the precision and power of effect estimates. This is particularly important given our null findings when comparing the discontinuation and continuation of 5-ASA. Third, the source data were derived from national research infrastructures that have well-characterised population databases with robust longitudinal follow-up and complete medication histories. Given the longer term follow-up of patients in these cohorts we were able to observe the progression of medical history over time and capture potential short and long-term adverse outcomes following discontinuation of 5-ASA.

We acknowledge several limitations of this study. First, we used databases that relied on administrative health claims to define patient characteristics and outcome. There is therefore the potential for misclassification from improper coding of diagnoses and we had to rely on prescriptions as a measure of patients actually taking medications. However, findings were consistent in both databases and inclusion criteria required UC-specific medications so misclassification was likely low. The Danish cohort has been validated and demonstrated at least 90% validity of UC diagnoses comparing charts with diagnosis codes.20 The US database has unfortunately not had similar validation of case definitions but our definition was strict based on ICD-9 code and two UC-related medication codes (5-ASA and anti-TNF) so misclassification of cases should likely be low. Second, both cohorts lack granular UC-specific data, such as clinical symptoms, laboratory values, disease extent by Montreal classification, smoking status and endoscopic findings that would allow for more detailed control of disease severity. We controlled for markers of disease severity prior to 5-ASA discontinuation and all patients required escalation to anti-TNF therapy so disease severity should be balanced between groups. If disease severity was not comparable between those who stopped or continued 5-ASA we would expect to see significant differences in risk of adverse events which was not observed. Third, given the retrospective nature of this study, there may have been unmeasured confounders that contributed to the decision about continuing 5-ASA in some patients, which could have attenuated risk of major adverse outcomes. However, given variability in practice patterns with the management of UC and the lack of consensus on when to discontinue 5-ASA, it is highly unlikely that clinicians would have employed similar algorithms for discontinuing or continuing 5-ASA. Since our study examined to different countries, the US and Denmark, clinical practice patterns in each cohort were likely heterogeneous given differing medical systems and similar outcomes were still observed. Consequently, it is highly unlikely that any one unmeasured confounder would have significantly biased the results.

In addition, we did not include stopping anti-TNF as an endpoint due to insufficient data on reason for discontinuation. Using anti-TNF discontinuation as an endpoint may have misclassified patients as treatment failures when the biologic was stopped for reasons other than failure to improve their UC (e.g. side effects, cost, non-adherence). In fact, a study from Denmark found that the most common reason for infliximab discontinuation in UC was remission so this would have been an inappropriate endpoint given our aims.21 An important consideration with any longitudinal database study is immortal time bias. Immortal time is the span of follow-up in which the outcome of interest could not occur due to the exposure definition. We think the potential impact of this is likely limited in our study. Exposure was defined as stopping 5-ASA within 90 days of starting anti-TNF and the definition of starting anti-TNF treatment was restricted to at least 90 days of anti-TNF before the start of follow-up. Therefore, the exposure (5-ASA stop or continue) was determined during the initial 90 days in which patients were required to be on anti-TNF before inclusion in the cohort. Outcomes of interest for our analysis could not have occurred during this initial 90-day anti-TNF treatment period in both groups (stop and continue 5-ASA). However, we do not capture very early events after starting anti-TNF (within 90 days of starting). We suspect that these patients are a specific sicker subpopulation in which the question of impact of 5-ASA discontinuation is of less clinical relevance.

A key issue that we were unable to fully evaluate is that 5-ASA may be continued longer term because of its potential benefit as a preventive agent against dysplasia and colorectal cancer. We were unable to assess this given a very low number of events of advanced colorectal neoplasia in both cohorts (fewer than five diagnoses of colorectal cancer) which precluded meaningful analysis. However, rates of colorectal cancer in IBD have significantly decreased in recent decades at the same time immunosuppressant use increased and data are conflicting on the impact of 5-ASA as a chemopreventive agent.22–24 Further studies are needed to help delineate the relative contribution of different IBD medications to decreasing rates of colon cancer.

In conclusion, using two population-based databases, we observed that discontinuation of 5-ASA after initiating anti-TNF therapy in UC was not associated with an increased risk of major adverse clinical outcomes. Our findings suggest that patients with UC can safely discontinue 5-ASA after escalating therapy to an anti-TNF. Prospective studies investigating the benefits and risks of 5-ASA discontinuation after biologic initiation are needed to confirm our findings.