Frequency and Effectiveness of Empirical Anti-TNF Dose Intensification in Inflammatory Bowel Disease: Systematic Review with Meta-Analysis

Loss of response to antitumor necrosis factor (anti-TNF) therapies in inflammatory bowel disease occurs in a high proportion of patients. Our aim was to evaluate the loss of response to anti-TNF therapy, considered as the need for dose intensification (DI), DI effectiveness and the possible variables influencing its requirements. Bibliographical searches were performed. Selection: prospective and retrospective studies assessing DI in Crohn’s disease and ulcerative colitis patients treated for at least 12 weeks with an anti-TNF drug. Exclusion criteria: studies using anti-TNF as a prophylaxis for the postoperative recurrence in Crohn’s disease or those where DI was based on therapeutic drug monitoring. Data synthesis: effectiveness by intention-to-treat (random effects model). Data were stratified by medical condition (ulcerative colitis vs. Crohn’s disease), anti-TNF drug and follow-up. Results: One hundred and seventy-three studies (33,241 patients) were included. Overall rate of the DI requirement after 12 months was 28% (95% CI 24–32, I2 = 96%, 41 studies) in naïve patients and 39% (95% CI 31–47, I2 = 86%, 18 studies) in non-naïve patients. The DI requirement rate was higher both in those with prior anti-TNF exposure (p = 0.01) and with ulcerative colitis (p = 0.02). The DI requirement rate in naïve patients after 36 months was 35% (95% CI 28–43%; I2 = 98%; 18 studies). The overall short-term response and remission rates of empirical DI in naïve patients were 63% (95% CI 48–78%; I2 = 99%; 32 studies) and 48% (95% CI: 39–58%; I2 = 92%; 25 studies), respectively. The loss of response to anti-TNF agents―and, consequently, DI―occurred frequently in inflammatory bowel disease (approximately in one-fourth at one year and in one-third at 3 years). Empirical DI was a relatively effective therapeutic option.


Introduction
Biologic therapies have become the mainstay of treatment in inflammatory bowel disease (IBD). Antibodies targeting tumor necrosis factor-alpha (anti-TNF) have become essential in the armamentarium for the treatment of both ulcerative colitis (UC) and Crohn's disease (CD). TNF is a key proinflammatory cytokine that plays an important role in several autoimmune disorders, including IBD. Elevated stool and mucosal TNF concentrations in UC and CD patients have been shown to correlate with the disease activity [1]. Anti-TNF drugs operate via a multitude of mechanisms: they bind and clear soluble TNF but, also, cell-bound TNF, inducing cytotoxicity on immune cells, like T-cell apoptosis [2]. They are effective at inducing symptom relief, disease remission and mucosal healing and reducing the need for surgery and hospitalizations among patients with moderate-to-severe IBD. The current clinical guidelines recommend anti-TNF agents for patients who are refractory to other treatments [3][4][5][6]. The description of each included study is summarized in Table 1. There were six randomized, placebo-controlled trials (RCTs) [10][11][12][13][14][15], 48 prospective open-label observational trials and 119 retrospective studies.
A total of 157 studies assessed the need for DI; the response rate was evaluated in 52 studies, and the remission rate was reported in 33 studies.
One hundred and one studies focused on naïve patients, and 29 evaluated non-naïve patients, while 50 studies included both naïve and non-naïve patients in their assessments. In six studies, prior anti-TNF exposure was not reported. One hundred and seven studies reported the data from IFX users and 92 from ADA users. Only five studies included patients receiving golimumab [16][17][18][19][20], and four studies evaluated patients receiving certolizumab [21][22][23][24]; thus, a meta-analysis was not performed.

Anti-TNF Use by Medical Condition in Naïve Patients
The DI requirement rate after the 12-month follow-up with all the anti-TNF agent data was statistically higher in UC than in CD patients (test for subgroup differences: χ2 = 5.29, p = 0.02, I2 = 81.1%). No other subgroup differences were reported by the medical condition or anti-TNF used ( Table 2). Table 2. Dose intensification rate after the 12-month follow-up by the anti-TNF agent In non-naïve patients, the DI rate ranged from 7% (26) to 81% (111), with an overall pooled rate of 39% (95% CI 31-47, I2 = 86%, 18 studies) (Figure 2). The DI requirement after the 12-month follow-up was statistically higher in non-naïve than in naïve patients (test for subgroup differences: χ 2 = 6.13, p = 0.01, I2 = 83.7%).

Anti-TNF Use by Medical Condition in Naïve Patients
The DI requirement rate after the 12-month follow-up with all the anti-TNF agent data was statistically higher in UC than in CD patients (test for subgroup differences: χ 2 = 5.29, p = 0.02, I2 = 81.1%). No other subgroup differences were reported by the medical condition or anti-TNF used ( Table 2).

Thirty-Six Month Follow-Up
A total of 25 studies with a median follow-up of 36 months were analyzed. There was only one study reporting the DI rate in non-naïve patients, and therefore, no subgroup analysis was performed.

Anti-TNF Use by Medical Condition in Naïve Patients
No statistical differences (p>0.05) in the medical conditions or the anti-TNF drug used were found between the subgroups (Table 3).

Anti-TNF Use by Medical Condition in Naïve Patients
No statistical differences (p > 0.05) in the medical conditions or the anti-TNF drug used were found between the subgroups (Table 3).

Short-Term Follow up
A total of 17 studies with a median of three to nine months of follow-up were included. The DI rates in naïve patients ranged from 14% (130) to 71% (79) with an overall pooled rate of 29% (95% CI 31-37, I2 = 96%, five studies).
A subgroup analysis evaluating the follow-up time (short-term vs. 12 months vs. 36 months) showed no statistical differences (p > 0.05) in terms of the DI requirements in naïve patients.
The overall rate of the short-term response to the empirical DI was 63% (95% CI: 48-78%, I2 = 99%, 32 studies) and 58% (95% CI: 47-70%, I2 = 68%, nine studies) in the naïve and non-naïve patients, respectively (Figure 4). No statistical differences were found between the groups (p>0.05). No statistical differences were found when comparing CD vs. UC patients or the anti-TNF drugs used (Table 4). Neither were found (p>0.05) between different intensification regimens (i.e., intensification of dosing vs. reduction of the interval of administration). No statistical differences were found when comparing CD vs. UC patients or the anti-TNF drugs used (Table 4). Neither were found (p > 0.05) between different intensification regimens (i.e., intensification of dosing vs. reduction of the interval of administration).

Pediatric Population
A total of 24 studies reported data on children (<18 years) ( Table 1). When compared to the adult population, no statistical differences were found in terms of the DI required or its efficacy. The random-effects pooled DI rate in naïve patients after a 12-month follow-up was 29% (95% CI 21-37%, I2 = 81%, n = 9).

Randomized Controlled Trials
A total of five randomized controlled trials (Table 1) assessed the DI requirements after a 12-month follow-up in naïve patients. The random-effects pooled DI rate was 29% (95% CI 18-41%. I2 = 88%, five studies). No statistical differences were found when this subgroup was compared to the group of observational studies.

Sensitivity Analyses and Risk of Bias
We further investigated potential sources of heterogeneity by excluding studies that included extreme or diverging values in certain subgroups, such as the DI requirements after 12 months [34,100,123,127,147,165] and 36 months [85,113] of follow-up or the response [147] and remission [61,145,168,183] rates. The effects of including different followup periods in the same subgroup [34,147,149] or the use of different induction dosing regimens [13,126,138] were also explored. In all cases, the results were stable, with no significant variations after the sensitivity analysis, although the heterogeneity remained considerable.
Among the six RCTs evaluated for a potential risk of bias, five had a low risk of bias for randomization, and four of them reported on the implementation of the random allocation sequence preserving concealment. Four studies also reported the adequate blinding of participants and personnel. Three studies showed low risks of attrition bias; in two of them, the number of excluded patients was not specified, and in the remaining one, there was a difference in the proportion of the outcome data. Finally, none of the studies was considered to show reporting biases. In conclusion, for most of the RCT items assessed, there was a low potential risk of bias detected.

Discussion
A LOR to the anti-TNF agents represents a therapeutic challenge to gastroenterologists, as these drugs are usually indicated in severe forms of the disease, and the remaining treatment options in such situations are limited. However, there is no unanimous definition of LOR in the literature [185,186]; it has been defined as an increase in clinical activity (which can be assessed by numerous activity indices) or, alternatively, as the need to modify or discontinue the current treatment. Thus, several authors have proposed that the DI requirement, which has been shown to recapture the response in multiple studies [187], would be a more objective and reliable measure [188] and, therefore, a useful surrogate for the LOR. Several reviews have previously assessed the incidence of a LOR, mainly in CD [185][186][187][188][189][190][191]. When compared to previous reviews, our study includes a considerably higher number of studies, up to January 2020, assessing both UC and CD patients and, therefore, conferring more robustness and reliability to our work.

Prior Anti-TNF Exposure
Several studies have estimated that approximately one-third of inflammatory bowel disease patients experience LOR and require DI, and that occurs more frequently in patients with prior anti-TNF exposure [188][189][190][191].
In our study, the overall rate of the DI requirements at a one-year follow-up was 28% in naïve and 39% in non-naïve patients, respectively. This shows no relevant differences with the previous data and constitutes one main finding of our study: dose escalation was needed more often in patients with prior anti-TNF use. In fact, the vast majority of the included studies evaluating both naïve and non-naïve patients showed a greater incidence in the loss of response in those non-naïve [30,34,35,60,81,84,111,121,132,161,163,168,192,193].

Time of Follow-Up
Additionally, the time course of LOR remains poorly understood. The median time from the first anti-TNF exposure to the need for a DI varied widely among the studies, from 2.7 to 18 months. However, there is increasing evidence showing that such events occur mostly within the first year of anti-TNF therapy [186].
In our study, no differences were found in the rate of DI for the short term, 12 and 36 months of follow-up, supporting the fact that the LOR and consequent DI occur mainly during the first year of treatment.

Medical Baseline Condition
Another relevant finding in our study was that a DI was required more frequently in UC than in CD patients. Previous data indicated that some patients with active UC have a higher inflammatory burden and accelerated anti-TNF clearance [194][195][196]; therefore, they could require a higher drug exposure to achieve a response to TNF antagonists. This could be the rational explanation UC patients need for an earlier and more frequent DI than CD patients [110,120,167]. However, there is also evidence not supporting these results [174]. Further research should be conducted, as no randomized trials have focused on this subgroup of patients; they seem to have the highest DI rate and could benefit the most from alternative treatment strategies.

Anti-TNF Agent
The comparison between the IFX and ADA DI rates is also a matter of interest. Immunogenicity is believed to be a common cause of LOR due to the formation of antidrug antibodies. Some authors have argued that the chimeric nature of IFX, as opposed to the fully humanized ADA, could render the former more prone to generate an antibody response. However, in our study, we did not find significant differences in the DI rate between IFX and ADA patients, as in previous comparative reports [115].

Dose Intensification Efficacy
Several clinical trials and open-label cohorts included in a previous review reported DI to restore the response in 50-70% of patients [186]. Billioud et al. also found that DI restored the response in 71% and remission in 40% of the patients [189].
In our study, the response and remission rates to empirical DI in naïve patients were 63% and 48%, respectively. Although no significant differences were reported between the naïve and non-naïve patients, either in the response or remission rates, a trend towards a reduced DI efficacy in the patients with prior anti-TNF exposure was shown.
Our findings support that using all the available treatment options with the first anti-TNF agent through DI (even if it is not based on therapeutic drug monitoring) should be considered before switching to another anti-TNF agent or to another therapeutic target. Nevertheless, it should be noted that almost all studies do assess the DI efficacy in the short term; additional research regarding the long-term response and remission rates after DI should be performed.

Limitations
Our study had some limitations. First of all, the DI can result in an equivocal interpretation of the LOR if it is done without accurately confirming the disease activity. In addition, there were also some possible predictors for the LOR or DI that were not evaluated in our study, such as the concomitant use of immunomodulators. However, recent guidelines (three) have suggested monotherapy with anti-TNF in patients with long-term remission rather than the use of a combination therapy. Finally, we excluded studies in which the DI was made based on therapeutic drug monitoring, with the aim to assess the effectiveness of empirical DI. In this respect, the current guidelines (three) do not recommend either proactive or reactive therapeutic drug monitoring as a standard clinical practice due to insufficient evidence. Finally, we did not perform a quality assessment of all the included studies given the high heterogeneity of the observational studies encountered in terms of the design and number. It was decided to perform a risk of bias assessment exclusively in RCTs, which represented no more than 1.5% of the total of patients included in our systematic review but, including 512 patients, was a sufficient sample size to drawn robust conclusions. In terms of quality, most studies showed a low risk of bias for the majority of the items assessed, highlighting both an adequate random sequence generation and allocation concealment, as well as blinding: items that were usually preserved. Additionally, a subgroup analysis was performed to control for heterogeneity in terms of study design, and no significant differences in the DI requirement between the RCTs and observational studies were reported.

Conclusions
A LOR to anti-TNF agents-and, consequently, DI-occurs frequently in inflammatory bowel disease, with an overall rate of DI requirement of approximately one-fourth at one year and one-third at three years. DI is required more frequently in patients with prior exposure to anti-TNF agents and in UC patients. Empirical DI is a relatively effective therapeutic option, achieving a response in two-thirds and remission in one-half of those patients naïve to anti-TNF treatment.