Systematic review with meta‐analysis: efficacy and safety of oral Janus kinase inhibitors for inflammatory bowel disease
1 | INTRODUC TION
Over the past two decades, biologic agents targeting tumour ne‐ crosis factor alpha (TNF), leucocyte integrins and interleukin (IL) pathways have become the mainstay of treatment for moderate‐to‐ severe Crohn’s disease (CD) and ulcerative colitis (UC).1 However, these agents have some important limitations. First, approximately 30% of patients are primary nonresponders to biologics and up to half will lose response in long‐term follow‐up due to sensitisation and development of anti‐drug antibodies, intolerance, or “mechanistic escape”.2 Second, currently available biologics must be intravenously or subcutaneously administered, which is a burden for patients that compromises treatment durability or complicance.3 Third, despite the introduction of lower cost biosimilars, the costs of biologics re‐ main an important challenge for payors.4 Finally, treatment with TNF antagonists is associated with an increased risk of serious infections.5 Thus, novel therapies with alternative mechanisms of action that are effective, well‐tolerated and safe are needed for patients with moderate‐to‐severe inflammatory bowel disease (IBD). Multiple orally administered, non‐immunogenic small molecule drugs hold promise for the treatment of IBD. Of these, several inhibitors of Janus kinase (JAK) are now in clinical development with tofacitinib recently being approved for treatment of moderate‐to‐severe UC.6 The JAK family consists of four intracellular tyrosine kinase (TYK) proteins (JAK1, JAK2, JAK3 and TYK2). Paired JAK activation is responsible for ini‐ tiating the intracellular signalling associated with different cytokine receptors. JAK phosphorylation subsequently activates intracytoplas‐ mic signal transducer and activator of transcription (STAT) pathways to control downstream target gene expression of inflammatory medi‐ ators.7,8 Depending on the selectivity of the JAK inhibitor for certain TYK proteins, different inflammatory pathways mediated by interfer‐ ons, ILs and colony stimulating factors, can be targeted. Genome‐wide association studies have demonstrated an association between poly‐ morphisms encoding JAK‐STAT proteins with exaggerated immune responses in patients with IBD.9‐11 Conversely, blocking JAK‐STAT sig‐ nalling can result in profound, broad‐spectrum immunosuppression.12 Blocking the JAK‐STAT signalling pathway has proven successful
in other immune‐mediated disorders and several oral JAK inhibitors have now received regulatory approval for the treatment of rheuma‐ toid arthritis, psoriasis, allergic dermatitis, myelofibrosis and polycy‐ thaemia vera.13 Considerable attention is currently being paid to the clinical development of JAK inhibitors for IBD. Two previous network meta‐analyses have demonstrated that tofacitinib, an oral JAK1/3 in‐ hibitor, is effective for achieving clinical, endoscopic and quality of life outcomes compared to placebo in patients with UC.14,15 However, negative trial results from two phase II studies of tofacitinib in CD16,17 and one trial of peficitinib in UC,18 as well as an increased risk of in‐ fections, particularly of herpes zoster,19 has raised questions regard‐ ing the risk‐benefit profiles of this class of therapy. To help inform this debate, we conducted a systematic review and meta‐analysis of all placebo‐controlled randomised trials evaluating JAK inhibitors for CD or UC to determine their pooled efficacy and safety relative to placebo.
2 | MATERIAL S AND METHODS
A systematic review and meta‐analysis were conducted in accord‐ ance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) recommendations.20
2.1 | Data Sources
PubMed (1946‐2018), Embase (1947‐2018) and the Cochrane CENTRAL Library (1996‐2018) were searched without language re‐ striction from inception to November 1, 2018. The search strategy captured terms relating to JAK inhibition combined with IBD, UC, or CD. Complete search strings are presented in File S1. The search was supplemented by reviewing conference proceedings from Digestive Disease Week, United European Gastroenterology Week, and the Congress of the European Crohn’s and Colitis Organization (2013‐2018) for relevant abstracts.
2.2 | Study Selection
Eligible studies were randomised, placebo‐controlled trials assess‐ ing the efficacy and/or safety of any JAK inhibitor for the treatment of adult patients with IBD. Trials without a placebo arm or enroll‐ ing paediatric patients were excluded. All identified citations were screened by two independent reviewers (JL and AM) in duplicate to identify relevant studies. Discrepancies were resolved through con‐ sensus and consultation with a third reviewer (CM) if needed. The full text of potentially eligible articles for inclusion was reviewed.
2.3 | Data Extraction and Outcome Measures
Data extraction was performed in duplicate by two independent re‐ viewers (CM and AT) using a standardised electronic data collection form. Discrepancies were resolved by consensus.
For CD, clinical outcomes were assessed using the Crohn’s Disease Activity Index (CDAI).21 We extracted clinical response de‐ fined by a CDAI reduction ≥100 points compared to baseline and clinical remission defined by a CDAI < 150. Endoscopic outcomes in CD trials were assessed using the Simple Endoscopic Score for Crohn's Disease (SES‐CD).22 Endoscopic response was defined as a >50% reduction in SES‐CD and endoscopic remission was defined by an SES‐CD ≤4 (with or without criteria for ≥2 point‐reduction compared to baseline and no segmental ulceration subscores >1). Mucosal healing was defined by an SES‐CD = 0.
For UC, clinical outcomes were evaluated using the Mayo Clinic Score (MCS)23 which is a composite index consisting of stool fre‐ quency, rectal bleeding subscore (RBS), physician global assessment, and Mayo endoscopic subscore (MCSe). Clinical response was de‐ fined either as: (a) an MCS reduction ≥3 points and 30% from base‐ line, with reduction in RBS ≥1 or an absolute RBS = 0 or 1; or 2) reduction in the Adapted Mayo Score ≥2 and 30% from baseline, with absolute RBS = 0 or 1. Clinical remission was defined by an MCS ≤2 with no individual component scores >1. Endoscopic response was
defined as a reduction in MCSe ≥1 compared to baseline. Endoscopic remission was defined as an MCSe = 0 or 1. Mucosal healing was defined as an MCSe = 0. This definition of mucosal healing does not incorporate histologic remission; however, most included trials were performed prior to this change in definition of mucosal healing. Given heterogeneity in outcome definitions in IBD trials,24,25 we ex‐ tracted data based on the above definitions rather than the original study definitions to facilitate comparisons between trials.
The proportion of patients experiencing any adverse event (AE), serious AEs (SAE), study withdrawal due to AE, infectious AEs and worsening IBD were extracted. Additionally, we captured the number of patients with herpes zoster (HZ), dyslipidaemia, bowel perforation, malignancy or death. AEs that occurred during open‐label mainte‐ nance or extension phases without a placebo comparator were not included.
Additional covariables extracted were trial characteristics (pub‐ lication date, first author, intervention drug and dosing, JAK selec‐ tivity, trial duration), patient characteristics (gender, age, disease duration, prior exposure to TNF‐antagonists, concurrent cortico‐ steroid use) and disease characteristics (baseline C‐reactive protein CRP and faecal calprotectin FCP, CDAI or MCS). An assessment of risk of bias in the included trials was performed using the Cochrane Risk of Bias Tool for Randomised Trials.26
2.4 | Data Synthesis
The relative risk (RR) was used to compare treatment effects for both efficacy and safety. Study level RRs with 95% confidence in‐ tervals (CI) were calculated in accordance with the intention‐to‐treat principle. For dose‐ranging studies, data from all treatment doses were pooled. Data were pooled using a random‐effects model based on the DerSimonian and Laird method to account for both within‐ and between‐study heterogeneity.27 We considered the study populations heterogeneous given the pooling of different therapies and doses. Between study heterogeneity was evaluated using the Cochran’s Q test and the I2 metric. The following thresh‐ olds for I2 interpretation were used in accordance with the Cochrane Handbook for Systematic Reviews of Interventions: 0%–40% (may not be important), 30%–60% (moderate heterogeneity), 50%–90% (substantial heterogeneity), 75%–100% (considerable heterogene‐ ity). Although we planned to pool induction and maintenance studies separately, there were an insufficient number of maintenance stud‐ ies to perform meta‐analysis. Accordingly, maintenance studies are described qualitatively for efficacy outcomes. However, for safety outcomes, we pooled both induction and maintenance studies. A priori, we decided to explore potential sources of heterogeneity by univariable meta‐regression (expressed as odds ratios [OR] with 95% CI), to evaluate the influence of JAK selectivity (JAK1‐selective in‐ hibitor vs pan‐JAK inhibitor) and previous TNF antagonist exposure (per 10% of the trial population) on efficacy and safety outcomes. Potential publication bias was not formally evaluated by funnel plots as fewer than 10 studies were identified for either CD or UC.All analyses were conducted in STATA version 14.2 (StataCorp LLC, College Station, TX). All P‐values are two‐tailed and P < 0.05 indicates statistical significance.
3 | RESULTS
3.1 | Search Results and Included Studies
We identified 491 records through the database search and four ad‐ ditional records from other sources. After removing duplicates, we screened 354 records and excluded 262 ineligible studies based on title and abstract review; another 80 records were excluded on full text as‐ sessment (Figure S1). A total of 12 placebo‐controlled RCTs were eligi‐ ble for inclusion in the systematic review and meta‐analysis.16‐18,28‐33 Five trials16,17,28,29 evaluated patients with CD, seven trials18,30‐33 evalu‐ ated patients with UC. These trials and their baseline patient character‐ istics are summarised in Table 1. A total of 813 CD patients and 2637 UC patients were randomised (2606 to active therapy with a JAK inhibi‐ tor, 844 to placebo). There were three trials of JAK1‐selective inhibitors (upadacitinib29,32 and filgotinib28) and nine trials of pan‐JAK inhibitors (tofacitinib,16,17,30,31 peficitinib18 and TDF‐147333).
3.2 | Efficacy of JAK Inhibitors as Induction Therapy
3.2.1 | Crohn's disease
Efficacy outcomes for all included studies are summarized in Table 2. Four studies assessed the efficacy of JAK inhibitors as induction therapy in CD. Treatment with a JAK inhibitor was as‐ sociated with both clinical remission (RR 1.38 [95% CI 1.04‐1.83] P = 0.025, χ2 = 3.47, I2 = 14%) and clinical response (RR 1.34 [95% CI 1.13‐1.58] P = 0.001, χ2 = 3.38, I2 = 0%) (Figure 1A,1) compared to placebo. Two trials assessed endoscopic outcomes in CD (FITZROY, CELEST). Treatment with a JAK inhibitor was not significantly as‐ sociated with endoscopic response (RR 3.04 [95% CI 0.68‐13.47] P = 0.114, χ2 = 2.22, I2 = 55%) or endoscopic remission (RR 2.83 [95% CI 0.76‐10.56] P = 0.122, χ2 = 1.18, I2 = 15%) (Figure 1C).
3.2.2 | Ulcerative Colitis
Six studies assessed the efficacy of JAK inhibitors as induction therapy in UC. Treatment with a JAK inhibitor was associated with both clinical remission (RR 3.07 [95% CI 2.03‐4.63] P < 0.001, χ2 = 2.45, I2 = 0%) and response (RR 1.72 [95% CI 1.38‐2.13] P < 0.001, χ2 = 8.05, I2 = 38%) (Figure 2A,2) compared to pla‐ cebo. Treatment with a JAK inhibitor was also associated with a greater likelihood of achieving endoscopic response (MCSe reduc‐ tion ≥1) (RR 1.44 [95% CI 1.04‐2.00] P = 0.028, χ2 = 0.74, I2 = 0%) endoscopic remission (MCSe = 0/1) (RR 2.43 [95% CI 1.64‐3.59] P < 0.001, χ2 = 6.81, I2 = 27%) (Figure 2C) and mucosal healing (MCSe = 0) (RR 5.50 [95% CI 2.46‐12.32] P < 0.001, χ2 = 0.82, I2 = 0%) (Figure 2D).
3.3 | Efficacy of JAK Inhibitors as Maintenance Therapy
3.3.1 | Crohn's Disease
One trial assessed the efficacy of tofacitinib as maintenance ther‐ apy. Panes et al reported that clinical response was maintained at 26 weeks in 46.5% (40/86) of CD patients treated with tofacitinib compared to 35.7% (15/42) of patients treated with placebo; clini‐ cal remission was maintained in 39.5% (34/86) of patients treated with tofacitinib compared to 28.5% (12/42) of patients treated with placebo.16 The proportion of patients maintaining clinical response or remission with either tofacitinib 5 mg twice daily (BID) or 10 mg BID were not significant compared to placebo.
3.3.2 | Ulcerative colitis
Maintenance outcomes were evaluated in the large scale OCTAVE‐ Sustain trial for UC; 593 patients who responded in induction therapy were followed for 52 weeks.31 A higher proportion of patients main‐ tained clinical response with tofacitinib (51.5% for 5 mg BID, 61.9% for 10 mg BID) compared to placebo (20.2%, P < 0.001 for both com‐ parisons). Likewise, a higher proportion of patients assigned to to‐ facitinib maintained clinical remission (34.3% for 5 mg BID, 41.1% for 10 mg BID) than those who received placebo (11.1%, P < 0.001 for both comparisons). Mucosal healing at 52 weeks occurred in 37.4% of patients in the 5 mg BID tofacitinib group (74/198) and 45.7% of pa‐ tients in the 10 mg BID group (90/197), compared to 13.1% (26/198) of patients in the placebo group (P < 0.001 for both comparisons).
3.4 | Safety Outcomes
The safety of JAK inhibitors compared top lacebo is summarised in Table3. IBD patients (CD and UC pooled) treated with a JAK inhibitor were not at signifi‐ cantly higher overall risk for AEs (RR 1.02 [95% 0.97‐1.09] P = 0.412, χ2 = 5.29, I2 = 0%) (Figure 3A). A total of 11 studies evaluated SAEs, with a pooled RR of 0.82 [95% CI 0.58‐1.16] (P = 0.266, χ2 = 12.51, I2 = 20%) (Figure 3B). For both SAEs and treatment withdrawals due to AEs, treatment with a JAK inhibitor resulted in a numerically but not statistically increased risk of adverse out‐ comes amongst CD patients (RR 1.30 [95% CI 0.74‐2.28] and RR 1.22 [95% CI 0.72‐2.10], respectively). In contrast, significant protection was observed in the UC studies (RR 0.62 [95% CI 0.42‐0.91] and RR 0.63 [0.40‐0.99], re‐ spectively). There were seven studies that evaluated infectious AEs: treat‐ ment with a JAK inhibitor was associated with a significantly increased risk of infections (RR 1.40 [95% CI 1.18‐1.67] P < 0.001, χ2 = 2.74, I2 = 0%). The most commonly reported infections were upper respiratory tract infections or na‐ sopharyngitis. However, treatment with a JAK inhibitor was also significantly protective against worsening IBD (RR 0.53 [95% CI 0.42‐0.66] P < 0.001, χ2 = 3.85, I2 = 0%), particularly for UC (RR 0.48 [95% CI 0.38‐0.67]).
Four cases of HZ were reported in CD trials amongst patients receiving a JAK inhibitor, whereas no cases of HZ occurred in pa‐ tients receiving placebo. A total of 19 cases of HZ were reported in UC trials amongst patients receiving a JAK inhibitor, whereas three cases were reported in patients receiving placebo. Most of these oc‐ curred in the OCTAVE trial with five cases in patients who received tofacitinib 10 mg BID during induction, three cases in patients who received tofacitinib 5 mg BID during maintenance and 10 cases in patients who received tofacitinib 10 mg BID during maintenance. Elevated serum lipid concentrations were primarily reported in in‐ duction studies: compared to baseline, the mean percent change in total cholesterol ranged from 8.4% to 19.2%, low‐density lipopro‐ tein 4.3%‐20.9% and high‐density lipoprotein (HDL) 11.0%‐25.5%. In maintenance studies, mean percent change in total cholesterol ranged from 0.2% to 4.3%. A total of four cases of bowel perforation were reported in patients who received a JAK inhibitor compared to two cases in patients assigned to placebo.
A total of 11 malignancies (nine cases in patients who received a JAK inhibitor, two cases in patients assigned to placebo) were re‐ ported across all trials (primarily nonmelanomatous skin cancers, two cases of malignant melanoma, one case of breast cancer). Two deaths were reported: one case occurred due to sepsis‐related ven‐ tricular fibrillation in a patient treated with placebo and the sec‐ ond due to an aortic dissection in a patient treated with tofacitinib 10 mg BID in the OCTAVE 1 trial.
3.5 | Influence of JAK Selectivity on Efficacy and Safety Outcomes
Compared to nonselective JAK inhibitors, treatment with a JAK1‐ selective inhibitor (filgotinib or upadacitinib) was not associated with significantly improved clinical remission (OR 1.06 [95% CI 0.39‐2.88], P = 0.90, n = 9 studies), clinical response (OR 1.12 [95% CI 0.71‐1.77], P = 0.57, n = 10 studies), endoscopic remission (OR 1.71 [95% CI 0.49‐5.93], P = 0.33, n = 8 studies) or mucosal healing (OR 0.33 [95% CI 0.01, 13.24], P = 0.41, n = 5 studies). Treatment with a JAK1 selective inhibitor was also not associated with reduced AEs (OR 1.11 [95% CI 0.92‐1.33], P = 0.24, n = 10 studies), SAEs (OR 1.40 [95% CI 0.51‐3.83], P = 0.46, n = 11 studies), treatment withdrawal (OR 1.84 [95% CI 0.62‐5.43], P = 0.23, n = 10 studies), infectious AEs (OR 0.98 [95% CI 0.57‐1.71], P = 0.94, n = 7 studies) or worsening IBD (OR 1.00 [95% CI 0.14‐7.31], P = 0.99, n = 9 studies) compared to a pan‐JAK inhibitor on meta‐regression.
3.6 | Influence of Previous TNF Antagonist Exposure on Efficacy and Safety Outcomes
Individual patient data on previous TNF antagonist exposure were not available. Therefore, we evaluated the effect of the proportion of patients previously exposed to TNF antagonists (per 10% increase) on efficacy and safety outcomes in each trial. In meta‐regression, previous TNF antagonist exposure was not associated with clinical remission in CD (OR 1.02 [95% CI 0.77‐1.36], P = 0.77, n = 4 stud‐ ies) or UC (OR 1.06 [95% CI 0.62‐1.80], P = 0.76, n = 5 studies) and there was no association with endoscopic remission (OR 1.12 [95% CI 0.65‐1.90], P = 0.56, n = 6 studies) or mucosal healing (OR 0.76 [95% CI 0.18‐3.30], P = 0.51, n = 4 studies) in UC. Additionally, previous TNF antagonist exposure was not associated with an increased risk of AEs (OR 1.02 [95% CI 0.98‐1.06], P = 0.29, n = 10 studies) or SAEs (OR 1.11 [95% CI 0.94‐1.31], P = 0.19, n = 11 studies). Data on whether effect modification by duration of induction period (ie 8 vs 16 weeks) were not available.
3.7 | Risk of Bias
Risk of bias assessment is summarized in Figure S2. All studies were deemed to be either low or unclear risk of bias. The risk of bias for incomplete outcome data was deemed to be unclear only for studies currently reported in abstract form; publication of the study protocol and full trial results are expected to reduce the risk of bias.
4 | DISCUSSION
Although biologic therapies have greatly improved IBD management and reduced morbidity from these debilitating diseases, a substantial proportion of patients do not respond, lose response or develop intol‐ erance or side effects.34 Accordingly, an unmet clinical need remains for novel agents that are effective, safe, and disease‐modifying. JAK inhibitors are the first new orally administered, non‐immunogenic small molecules to enter the treatment armamentarium for IBD. In this systematic review and meta‐analysis of all placebo‐controlled RCTs of JAK inhibitors in adult patients with IBD, we concluded that these drugs are effective induction therapy for patients with active CD or UC. We also demonstrated that treatment with a JAK inhibitor was not significantly associated with an overall increased risk of AEs or SAEs compared to placebo. However, an increased risk of infection was observed, and the development of HZ was a specific signal of concern.
The efficacy observed was clinically meaningful in the relatively refractory population of patients that was evaluated in these tri‐ als. In pooled analyses, treatment with a JAK inhibitor was associ‐ ated with a 38% increase in achieving clinical remission for CD and over threefold increase in clinical remission for UC. JAK inhibitor treatment was also associated with an over twofold increase in en‐ doscopic remission and 5.5‐fold increase in mucosal healing for pa‐ tients with UC. Higher rates of endoscopic remission rates in CD compared to placebo were observed, although this was not statisti‐ cally significant. Importantly, this analysis was informed exclusively by data from the short‐term FITZROY (10 weeks) and CELEST in‐ duction trials (16 weeks) and the rigorous outcome of endoscopic remission (SES‐CD ≤4) may be difficult to achieve in studies of this duration.28,35 Higher rates of endoscopic remission were reported in the CELEST 52‐week extension study, ranging as high as 37.5% in patients treated with upadacitinib 12 mg BID.36
The efficacy of JAK inhibitors has been demonstrated in both biologic‐naïve and biologic‐exposed patients. Individual patient‐level data on TNF antagonist exposure were not available for meta‐analy‐ sis. However, the overall proportion of patients previously exposed to TNF antagonists in each trial was not associated with efficacy or safety outcomes on meta‐regression, recognising that this type of analysis is subject to potential ecological bias. In subgroup analysis of the OCTAVE 1 and 2 trials, similar effect sizes for clinical remission were observed for patients naïve to TNF antagonists (∆9.4%–13.5%) compared to patients with previous exposure (∆11.1%–12.0%). Additionally, similar rates endoscopic remission were also observed irrespective of previous TNF antagonist failure (∆13.3%–17.3% versus ∆15.6%–17.9%).31 The efficacy of JAK inhibitors even after exposure to a TNF antagonist has important implications for their positioning in treatment algorithms relative to other therapies that show reduced efficacy in biologic‐experienced populations. In a net‐ work meta‐analysis of first‐ and second‐line pharmacotherapy for moderate‐to‐severe UC, Singh et al demonstrated that while inflix‐ imab and vedolizumab were ranked highest for induction of clinical remission amongst biologic‐naïve patients, tofacitinib ranked high‐ est for induction of clinical remission (OR 11.88 [95% CI 2.32‐60.89]) for patients with prior TNF antagonist exposure.37 Additional stud‐ ies should evaluate whether there is effect modification by duration of induction (eg 8 vs 16 weeks) in patients with previous TNF antag‐ onist exposure.
Concerns about the efficacy of JAK inhibitors for the treatment of IBD have primarily arisen from negative results obtained in phase II trials of tofacitinib in CD16,17 and peficitinib in UC.18 While drug dosing and JAK isoform selectivity may have influenced these find‐ ings, trial design flaws also need to be considered. All three phase II tofacitinib CD trials reported remarkably high placebo remission rates, ranging from 29.4% to 38.1%, which would have reduced trial sensitivity for detection of true treatment effects. In support of this notion, in all three trials significantly greater reductions in CRP and FCP concentrations were observed in patients who received tofaci‐ tinib compared to those assigned to placebo. The high placebo rate may have resulted from inclusion of patients with milder disease and lack of requirement for centrally read confirmation of endoscopic inflammation or elevated CRP/FCP at enrollment. In comparison, CD patients enrolled in the filgotinib FITZROY trial were required to have a total SES‐CD ≥7 at trial entry with ulceration scores of ≥1 in at least one ileocolonic segment: while the screen failure rate from this trial was high (44%), the clinical placebo remission rate in this positive trial was acceptably low (23%) and endoscopic placebo remission rates were 2%–7%.28 In the peficitinib trial, the primary outcome chosen was an unconventional dose‐response endpoint based on changes in MCS from baseline and not powered for con‐ ventional binary clinical or endoscopic response or remission. While continuous outcomes are likely to be more sensitive to change than binary endpoints, a validated composite disease activity index such as the recently developed UC100 score may have greater discrimina‐ tive performance for early phase trials.38
Our evaluation of safety outcomes was generally reassuring. Targeting JAK‐STAT has the potential for broad‐spectrum immuno‐ suppressive effects leading to concerns regarding risks of infection and malignancy.39,40 In this respect, it is reassuring that we iden‐ tified that the risk of AEs and SAEs overall was not significantly different from placebo. However, it should be recognised that most of the included studies were short‐term induction trials and the aggregate sample size was relatively small and disproportion‐ ally represented by the results of the OCTAVE trials that evaluated patients ≥65 years, Asians, those receiving tofacitinib 10 mg BID, and patients previously failing a TNF antagonist. Other agents that are used to treat IBD, including thiopurines and TNF an‐ tagonists also have an increased risk of HZ infection. Long et al evaluated the risk of HZ amongst 108 604 IBD patients using a nested case‐control design. In multivariable analysis, use of TNF antagonists was associated with a significantly increased risk of HZ (OR 1.81 [95% CI 1.48‐2.21]).45 The mechanisms underlying the HZ risk with JAK inhibitors likely relates to impaired innate immunity, with down‐regulation of type I and type II interferons and natural killer cell activity that is important for restricting var‐ icella zoster virus replication.46 The risk of HZ may be mitigated by vaccination, and an inactivated recombinant varicella zoster vaccine is now available, which demonstrated >90% protection in clinical trials.47,48
JAK isoform selectivity may be responsible for treatment‐related AEs. In our meta‐analysis, we did not identify JAK1 selectivity as associated with efficacy or safety in meta‐regression. This may reflect loss of JAK isoform selectivity at higher doses and variable selectivity based on tissue penetration, cell type, and individual patient genetics.49 However, it should be recognised that our anal‐ ysis lacked a sufficient number of trials to perform dose‐response analysis for each JAK inhibitor and thus unable to detect relevant differences in selectivity‐related AEs. Furthermore, small nominal although not statistical differences in the RR of efficacy and safety outcomes between different JAK inhibitors was observed (Figures 1‐3). Comparisons across different trials are challenging due to study population heterogeneity and differences in trial design. For ex‐ ample, the RR for achieving endoscopic remission in the OCTAVE induction trials ranged from 2.09 to 2.45, whereas the RR was 12.63 [95% CI 1.79, 88.87] in the upadacitinib U‐ACHIEVE study.
Comparing these RRs is difficult: the confidence interval for the smaller phase II U‐ACHIEVE trial is broad, and high RR reflects the very low placebo remission rate of 2% (compared to 11.6%‐15.5% in the OCTAVE trials). In contrast, the absolute rates of endoscopic remission in both studies were actually similar (Table 2).
Another potential strategy to maximise treatment efficacy while minimising safety concerns is to limit pharmacodynamic activity to the intestinal mucosa. This mechanism has recently been demon‐ strated in a phase I trial of TD‐1473, an intestinally restricted pan‐ JAK inhibitor that showed low plasma and high colonic tissue drug concentrations in the range required for JAK inhibition.33 Finally, while the current generation of JAK inhibitors primarily inhibit JAK1, 2 and 3, increasing attention is being focused on the development of TYK2 inhibitors. Single nucleotide polymorphisms in TYK2 have been shown to result in a loss of function mutation that is protec‐ tive against the development of UC50 and have also been associated with susceptibility to CD.51 Furthermore, TYK2 plays a key role in IL‐12/23 signalling that mediates intestinal inflammation.52 A selec‐ tive TYK2 inhibitor that has been shown to block IL12, IL23 and type I interferon in preclinical models is now in a phase II trial enrolling patients with moderate‐to‐severe CD (NCT03599622).53 Table 4 summarizes completed and currently enrolling JAK inhibitors for IBD indications by isoform selectivity.
Our study has some limitations. First, although we included all randomised trials in IBD, the number of placebo‐controlled stud‐ ies that have been fully reported remains small. This circumstance limited our ability to perform subgroup and sensitivity analyses. Second, although statistically significant heterogeneity was not observed for pooled efficacy or safety estimates, there was het‐ erogeneity in trial design, study duration and endpoint definitions between trials that may have influenced these assessments. Third, individual patient data are required to determine patient‐level co‐ variates that predict response to therapy or risk for AEs. This would allow evaluation of disease location, phenotype, smoking status and baseline CRP and FCP on treatment response. Although we evaluated the effect of previous TNF antagonist exposure and JAK isoform selectivity in meta‐regression, these analyses have limited therapies, it should not be assumed that these drugs are necessarily safer than parenterally administered biologics. Second, it is unclear that JAK inhibitors will be more cost‐effective than currently available biosimilars. Third, adherence to oral therapies in IBD has historically been poor and real‐world data are needed.55,56 Finally, there is a need to improve the efficiency of treatment by identifying individual patient predictors of response to JAK inhibitor therapy.
In conclusion, in this systematic review and meta‐analysis, we demonstrate that JAK inhibitors are effective for inducing clinical and endoscopic remission in patients with IBD, although treat‐ ment was associated with an increased risk of infections. Additional studies are required to understand the optimal positioning of these agents in management algorithms for CD and UC, and to refine methods of drug delivery, maximise treatment efficacy and minimise potential harms associated with JAK inhibition.