Introduction

The medical management of inflammatory bowel disease (IBD) has been revolutionized over the past few decades by an expanding therapeutic armamentarium that now features several effective biologic agents and oral small molecules targeting different components of the immune response, including antagonists to cytokines such as tumor necrosis factor-α (TNF) and the p40 common subunit of interleukin (IL)-12 and IL-23, the α4β7 integrin on leukocytes, and the Janus kinase (JAK) intracellular signal transducers [1]. Although these advances have afforded patients with Crohn’s disease (CD) and ulcerative colitis (UC) more treatment options than ever before, the role of optimizing drug therapy has also become increasingly important [2]. A substantial proportion of patients will experience either primary non-response or secondary loss of response to biologic therapy [3]. The mechanisms of treatment failure in IBD are complex, including disease-related, drug-related, and patient-related factors [4]. Given this complexity, treatment decisions directed by symptom assessment alone are unlikely to achieve optimal outcomes.

Therapeutic drug monitoring (TDM) has emerged as a promising strategy to maximize treatment response in IBD. Using measurements of serum drug or active metabolite and anti-drug antibody (ADAb) concentrations to guide therapy is based on the premise that (1) there is an exposure-response relationship wherein higher drug concentrations are positively associated with the magnitude of therapeutic efficacy [5]; (2) non-response can be mediated by pharmacokinetic failure, defined by inadequate drug exposure secondary to immune (i.e., ADAb formation) or non-immune causes (i.e., body mass index (BMI), gender, disease phenotype, concomitant immunosuppression, degree of systemic inflammation) leading to accelerated drug clearance [6, 7]; or (3) non-response can be mediated by mechanistic failure due to alternative pathways of inflammation in disease pathogenesis [8].

TDM has been increasingly adopted in clinical practice [9] and recent American Gastroenterological Association (AGA) Institute guidelines suggest the use of reactive TDM in the context of secondary loss of response to thiopurines or biologic therapy [10••]. However, many questions remain unanswered. First, it is unclear if TDM performed during induction or proactive TDM for patients in symptomatic remission improves long-term outcomes. Second, the role of TDM for small-molecule therapies and biologic agents with a non-anti-TNF-driven mechanism of action is unclear. Third, thresholds for therapeutic drug concentrations above which further dose escalation would likely prove futile have not been fully validated. In this review, we summarize the most current evidence for incorporating TDM in treatment algorithms and propose potential future applications of TDM in clinical practice.

Therapeutic drug monitoring for small molecules

Therapeutic drug monitoring has routinely been used for small molecules in the treatment of IBD, to detect nonadherence to therapy or to guide dose adjustments because of lack of efficacy or adverse events. Azathioprine is a prodrug that is converted into 6-mercaptopurine, which in turn is converted into 6-methylmercaptopurine (through thiopurine methyltransferase [TPMT]), 6-thiouric acid (through xanthine oxidase), or active 6-thioguanine nucleotides [6-TGN] (through hypoxanthine-guanine phosphoribosyl-transferase). Clinical response is highly correlated with levels of 6-TGN [11], whereas patients with low or absent TPMT enzyme activity are at risk for excess production of drug-derived TGN metabolites potentially leading to life-threatening myelosuppression [12]. Current guidelines recommend routine TPMT testing (enzymatic activity or genotype) in patients starting thiopurine therapy to guide thiopurine dosing and in those patients who are treated with thiopurine therapy and who have active disease and/or adverse events thought to be due to thiopurine toxicity, to do reactive thiopurine metabolite monitoring to guide treatment changes [10••].

Tofacitinib, a recently approved oral small-molecule pan-JAK inhibitor, has shown to be effective for induction and maintenance therapy in patients with moderately-to-severely active ulcerative colitis. In the OCTAVE studies, a numerical higher proportion of patients randomized to 10 mg twice daily during maintenance therapy reached the primary endpoint of remission at week 52, compared to those randomized to 5 mg twice daily [13]. This is in line with an earlier phase 2 trial in patients with ulcerative colitis where a dose-dependent effect was observed during induction therapy [14]. These results indicate that some patients may benefit from higher doses of tofacitinib. Interestingly, data on the pharmacokinetics of tofacitinib in patients with psoriasis showed that heavier subjects and those with prior exposure to biologic therapies were predicted to require a higher dose to achieve benefit compared to lighter subjects [15]. A population pharmacokinetic analysis of tofacitinib was performed in patients with moderately-to-severely active UC showing that plasma tofacitinib concentrations increased proportionately with dose and estimated oral clearance, and the average steady-state concentrations were not significantly different between baseline and week 8 [16]. The estimated between-patient variability (% coefficient of variation) was 31.4% for clearance, which is similar to what has been observed for biologics. Clearance of tofacitinib did not significantly correlate with any of the covariates tested including baseline measurements of fecal calprotectin, C-reactive protein (CRP), albumin, and total Mayo Clinic score. Although these results indicate that baseline disease activity is likely not a determinant of tofacitinib clearance, a more extensive covariate analysis is warranted to determine factors that can explain part of the observed interindividual variability for clearance in patients with UC. An exposure-response analysis was also conducted to evaluate the association between different measures of exposure (dose, average steady-state concentration, and steady-state trough concentration) and important clinical outcomes. It was found that the baseline Mayo Clinic score was an important determinant of efficacy at week 8 and that plasma concentrations in individual patients did not provide additional predictive value for efficacy beyond that provided by tofacitinib dose [16].

Therapeutic drug monitoring for biologics

Advances in drug and anti-drug antibody detection assays

Several assays are commercially available for measuring biologic drug and ADAb concentrations. Drug-sensitive solid-phase enzyme-linked immunosorbent assays (ELISAs) are widely available and less expensive compared to drug-tolerant tests such as the homogeneous mobility shift assay (HMSA) and electrochemiluminescence immunoassay (ECLIA). However, drug-sensitive assays are unable to detect ADAbs in the presence of drug [17]. Van Stappen et al. [18] have recently shown that an acid-based pre-treatment protocol can be used to convert the traditional solid-phase bridging ELISA to a drug-tolerant assay, improving the detection of ADAbs. Fluid-phase drug-tolerant assays have a greater sensitivity for detecting low-level, low-affinity ADAbs compared to ELISA, though neither can determine ADAb functionality. In contrast, a reporter gene assay (RGA) allows differentiation of neutralizing antibodies using an erythrocyte cell-based test [19].

Serum concentrations of TNF antagonists are generally well correlated when comparing across different assays. Marini et al. [20] evaluated infliximab levels measured using four commercial ELISAs and demonstrated that the intraclass correlation coefficient exceeded 0.89 for all tests. Similarly, golimumab levels as measured by two different ELISAs are closely correlated (Spearman’s r = 0.98, p < 0.0001) [21]. When comparing different assay types, Steenholdt et al. [22] showed that ELISA, HMSA, radioimmunoassay, and functional RGA all accurately detected serum infliximab (Pearson’s r = 0.91–0.97, p < 0.0001), but all assays except radioimmunoassay and RGA significantly disagreed on sample IFX concentrations and with a mean difference from 0.64 (0.15–1.12) μg/mL in ELISA and HMSA to up to 3.44 (2.49–4.39) μg/mL in RGA and ELISA. A previous study demonstrated that while correlated (r = 0.69–0.82), adalimumab concentrations measured by HMSA were consistently higher compared to ELISA, highlighting potential limitations of cross-assay extrapolation of absolute concentrations [23]. There is emerging data for ustekinumab on the comparison of the KU Leuven ELISAs for measuring ustekinumab and ADAb concentrations with ECLIAs developed at Janssen R&D and used in clinical studies of IBD patients showing a strong agreement [24], although comparative studies with a wider range of assays are urgently needed. For vedolizumab, a comparison between an ELISA and ultra-performance liquid chromatography tandem mass spectrometry system showed a moderate-to-good correlation [25], but comparative studies including assays used commercially in the EU and US are also urgently needed.

Measurements of ADAb concentrations across assays are poorly correlated. ADAb titers are frequently reported in different units (μg/mL, μg/mL equivalents, U/mL, arbitrary units/mL) and quantitative results are variable depending on methodology [12]. Therefore, comparison across assays should be avoided. For TNF antagonists, the presence of ADAbs is associated with accelerated drug clearance and loss of response, although drug-tolerant assays may over-detect very-low-titer, clinically irrelevant ADAbs that have limited effects on drug pharmacokinetics. Over half of ADAbs detected by drug-tolerant HMSA may not have neutralizing potential when assessed by functional RGA [22] and approximately 30% of antibodies to infliximab may be transient [26].

Effective implementation of TDM into clinical practice requires timely and efficient drug level and ADAb quantification. Historically, the slow turnaround time for results has precluded using drug and ADAb concentrations measured at trough for adjusting the next biologic dose, or timely decision to switch treatments if antibodies are present. A point-of-care assay has been developed for infliximab and when compared to two ELISA-based methods, the accuracy of the rapid test was high with intraclass correlation coefficients of 0.889 and 0.939 [27]. Similarly, a lateral flow-based assay with a time to result of 20 min has also shown excellent agreement with ELISA for quantification of infliximab (Pearson r = 0.95 during induction and r = 0.93 during maintenance) [28]. Adoption of rapid assays may permit the use of TDM to make “on-demand” adjustments to therapy and improve uptake for optimizing induction regimens with TNF antagonists or facilitate decision to switch out of class.

Defining therapeutic trough concentrations for TNF antagonists

The concept of defining a therapeutic target range for serum TNF antagonist trough concentrations stems from several observations. First, there is an exposure-response relationship between serum drug concentrations and clinical efficacy, wherein higher levels of infliximab [29, 30•], adalimumab [31, 32], certolizumab pegol [33], and golimumab [34] are associated with higher rates of clinical remission. Second, low drug concentrations are associated with loss of response to both infliximab and adalimumab and increase the risk of developing ADAbs [35, 36]. Third, the therapeutic drug concentration can be defined based on correlation with efficacy rather than safety outcomes as higher TNF antagonist concentrations have not been shown to correlate with the risk of adverse events [37]. Trough drug concentration thresholds in the literature have been primarily derived from retrospective cross-sectional studies that validate a chosen cutoff, maximize the area under the receiver operating characteristic (ROC) curve, analyze incremental gains with higher drug levels, or evaluate differences in the proportion of patients achieving treatment endpoints by quartile of exposure [12]. The optimal therapeutic trough level is dependent on the clinical context in which TDM is applied and varies by treatment target (clinical versus objective outcomes, response versus remission), disease state (reactive versus proactive testing), and phase of therapy (induction versus maintenance). Suggested trough concentrations according to the AGA guidelines as well as the Australian Inflammatory Bowel Diseases Consensus Working Group are summarized in Table 1.

Table 1 Suggested target trough concentrations for therapeutic drug monitoring of TNF antagonists
Full size table

Treatment targets in both CD and UC have shifted from achieving symptomatic remission towards targeting objective endoscopic, biomarker, and histologic endpoints [38]. Higher drug concentrations may be required to achieve these more robust, objective outcomes [32, 39,40,41]. For example, Ungar et al. [39] have recently described that a therapeutic window of 6–10 μg/mL for infliximab and 8–12 μg/mL for adalimumab is associated with mucosal healing in 80–90% of patients with IBD. Juncadella et al. [42] found that threshold adalimumab concentrations of 11.8, 12.0, and 12.2 μg/mL in CD and 10.5, 16.2, and 16.2 μg/mL in UC stratified patients with and without biochemical (CRP ≤ 5 mg/L), endoscopic (absence of ulcerations/erosions, Rutgeerts score ≤ i1, or Mayo endoscopic subscore ≤ 1), and histologic remission (absence of active inflammation). Similarly, higher infliximab concentrations are associated with endoscopic (9.7 μg/mL) and histologic (9.8 μg/mL) healing in CD [43]. Infliximab serum concentrations ≥ 5.1 μg/mL at week 14 and ≥ 2.3 μg/mL at week 30 were associated with week 30 Mayo Clinic endoscopic subscore ≤ 1, whereas higher concentrations of ≥ 6.7 μg/mL and ≥ 3.8 μg/mL, respectively, were associated with a subscore of 0 [44]. Cumulatively, these results suggest that better outcomes can be achieved with higher drug trough concentrations, although determination of causality in cross-sectional studies is challenging because better disease control may result in higher trough levels secondary to reduced drug clearance.

Disease phenotype also influences the optimal therapeutic trough level associated with remission. Patients with highly aggressive fistulizing CD may require higher drug levels to achieve response. Yarur et al. [45] evaluated 117 CD patients with perianal fistulizing disease treated with infliximab for a minimum of 24 weeks; those achieving fistula healing had significantly higher median serum drug concentrations compared to those with persistently active disease (15.8 vs. 4.4 μg/mL, p < 0.0001). In quartile analysis, the highest rate of fistula healing (86%) was achieved by patients in the top quartile of infliximab exposure (trough level 20.2–50 μg/mL) and levels associated with fistula healing and closure were higher than those previously correlated with luminal mucosal healing.

Less is known about the optimal drug concentration during induction therapy prior to achievement of steady state. There is great interest in using early optimization to distinguish patients who are primary non-responders to TNF antagonists (due to mechanistic failure) from those patients who require more aggressive dosing (due to pharmacokinetic failure) [46]. Patients at risk for accelerated drug clearance could be identified before initiating therapy by using a population pharmacokinetic approach, as a linear relationship was found between baseline infliximab clearance and week 8 Mayo Clinic endoscopic subscore (p < 0.001) [44]. A threshold infliximab clearance of < 0.397 L/day was associated with week 8 Mayo Clinic endoscopic subscore ≤ 1 with a sensitivity, specificity, positive predictive value, and an area under ROC curve of 75%, 48%, 68%, and 0.64 (95% CI, 0.59–0.69) (p < 0.0001), respectively. Observational studies demonstrate that higher infliximab and adalimumab serum concentrations as early as 2 to 6 weeks after the first TNF antagonist dose are associated with improved rates of clinical response and remission [30•, 47], mucosal healing [28, 48], and long-term drug retention and surgery-free survival [49]. Infliximab concentration ≥ 15 μg/mL at week 6 is associated with a 4.6-fold increase in likelihood of achieving weeks 10–14 endoscopic mucosal healing [48]. Conversely, infliximab levels below 6.8 μg/mL or early antibodies to infliximab (> 4.3 μg/mL) before the second infusion are associated with primary non-response [50]. During the induction phase, the high rate of drug clearance, early development of ADAbs, and heavy inflammatory burden are important mediators of serum drug levels. Patients with moderate-to-severe UC for example have highly accelerated infliximab drug clearance from demonstrable infliximab fecal losses [51]. Furthermore, high inflammatory burden (defined by CRP > 50 mg/L) predicts those UC patients with lower total infliximab exposure (587 vs. 1361 mg/L/day, p = 0.001) [52]. Using an incremental gain analysis, therapeutic windows of 30–36 μg/mL at week 2 and 24–30 μg/mL at week 6 for infliximab have been proposed to maximize likelihood of early mucosal healing although these thresholds require prospective validation [46].

In maintenance treatment with TNF antagonists, based on a meta-analysis, the AGA guidelines suggest trough concentrations of infliximab ≥ 5 μg/mL, adalimumab ≥ 7.5 μg/mL, and certolizumab pegol ≥ 20 μg/mL to be associated with clinical remission [10]. Insufficient evidence was available to establish a target trough for golimumab. These cutoffs were chosen based on the proportion of patients not in remission for incremental increases in drug trough concentration: however, 8% of patients with an infliximab trough concentration ≥ 5 μg/mL, 10% of patients with an adalimumab trough concentration ≥ 7.5 μg/mL, and 26% of patients with a certolizumab pegol trough concentration ≥ 20 μg/mL will not be in clinical remission, and a subset of these patients may still respond by targeting higher concentrations.

Treatment algorithms incorporating reactive and/or proactive TDM

Prior to the adoption of TDM, patients experiencing secondary loss of response to TNF antagonists were typically managed by empiric dose escalation. Although this approach exhausts the therapeutic potential of each treatment and is sensible when limited options are available, it may delay initiation of effective therapy and increase potentially unnecessary drug exposure among patients with immune-mediated pharmacokinetic treatment failure or mechanistic treatment failure [53]. The use of reactive TDM can direct more personalized and efficient treatment decisions by distinguishing patients with pharmacokinetic failure due to inadequate drug levels from those whose disease is not driven by TNF-mediated pathways (Fig. 1). The use of reactive TDM is more cost-effective compared to empiric dose escalation [54, 55] and allows earlier implementation of effective treatment decisions. For example, in a retrospective cohort study of 247 IBD patients developing 330 loss-of-response events to infliximab or adalimumab, Yanai et al. [56] identified that the presence of either therapeutic trough levels (adalimumab > 4.5 μg/mL, infliximab >3.8 μg/mL) or high-titer ADAbs (anti-adalimumab > 4 μg/mL equivalent, anti-infliximab > 9 μg/mL equivalent) predicted failure to respond to dose escalation with 90% specificity and had longer duration of response when switched to a different class of treatment.

Fig. 1
figure 1

Algorithm for use of reactive therapeutic drug monitoring in IBD patients with secondary loss of response to TNF antagonists.

Full size image

While reactive TDM has an established role for managing secondary loss of response, integrating TDM into clinical practice proactively for patients in stable remission remains controversial. In a multicenter retrospective study of 264 consecutive IBD patients receiving infliximab maintenance therapy, Papamichael et al. [57] compared proactive versus reactive drug monitoring based on measurements of first infliximab concentration and ADAb. In multivariable Cox regression, proactive drug monitoring was associated with a reduced risk for treatment failure (hazard ratio HR 0.16 [95% CI, 0.09–0.27]), IBD-related surgery (HR 0.30 [95% CI, 0.11–0.80]), IBD-related hospitalization (HR 0.16 [95% CI, 0.07–0.33]), and serious infusion reactions (HR 0.17 [95% CI, 0.04–0.78]). However, this retrospective comparison is limited by potential differences in patient characteristics wherein patients undergoing proactive testing were asymptomatic compared to those patients potentially experiencing a symptomatic disease flare in the reactive group.

A second purported benefit to proactive drug optimization is the potential to circumvent the need for concomitant immunosuppression with azathioprine or methotrexate. Combination therapy with infliximab and azathioprine is superior to infliximab monotherapy in CD [58] and UC [59], mediated by a reduction in ADAb formation and higher trough infliximab levels. However, concomitant immunosuppression is associated with an increased risk of adverse events [60]. In a comparison of 16 patients managed with week 10 proactive infliximab TDM and 35 patients on combination infliximab and immunosuppressant therapy, Lega et al. [61] demonstrated that proactive “optimized monotherapy” achieved comparable endpoints to combination therapy with respect to median infliximab trough concentration (9.1 vs. 7.7 μg/mL, p = 0.24), probability of anti-infliximab antibody-free survival (p = 0.27), and frequency of infliximab discontinuation (0% vs. 3%, p = 1.0). Correspondingly, post hoc analysis of the SONIC (the Study of Biologic and Immunomodulator Naïve Patients in Crohn’s Disease) trial showed comparable outcomes are achieved regardless of concomitant azathioprine when patients are stratified by infliximab trough quartiles [62].

Although observational evidence supports the use of proactive TDM, two randomized controlled trials have been inconclusive. The TAXIT study (Trough Level Adapted Infliximab Treatment study) was a 1-year trial that evaluated 178 CD patients and 85 UC patients with stable response to infliximab maintenance therapy who were then randomized to receive infliximab dosing either based on clinical features or based on proactive TDM [63••]. Importantly, all patients initially underwent an optimization phase where infliximab dosing was escalated or reduced to reach a trough level range of 3–7 μg/mL prior to randomization. At 12 months, the proportion of patients with combined clinical and biochemical remission was similar between the clinically based and proactive TDM-based dosing groups (66%vs. 69%, p = 0.686). At the end of the study, similar rates of mucosal healing were observed in patients randomized to clinically based compared to proactive TDM-based dosing (91% vs. 90%) [64]. During long-term follow-up after TAXIT (median 41 months), there were no differences in IBD-related hospitalization (13% vs. 15%), abdominal surgery (6% vs. 7%), or corticosteroid use (13% vs. 8%) in those patients previously randomized to clinically based compared to proactive TDM-based dosing, respectively, although proactive TDM was continued approximately once per year in all patients. Interestingly, proactive TDM-based dose optimization in all patients before randomization, to achieve trough concentrations in the 3–7-μg/mL range, was associated with significant reductions in CRP and an increase in the proportion of CD patients in remission—potentially negating some of the benefits associated with TDM—and a 28% reduction in drug costs among patients with a trough level > 7 μg/mL who were able to undergo dose reduction. Although the trial did not meet its primary endpoint, important differences in secondary outcomes were achieved between the proactive TDM-based and clinically based dosing groups including a lower proportion of patients who required rescue therapy due to loss of response (7% vs. 17%, p = 0.018) and a higher proportion of patients who stayed within the target trough level range (74% vs. 57%, p = 0.001).

Treatment with infliximab incorporating proactive TDM was also evaluated in the TAILORIX (A Randomized Controlled Trial Investigating Tailored Treatment With Infliximab for Active Luminal Crohn’s Disease) randomized controlled trial [65••] that evaluated a treatment algorithm based on clinical symptoms, biomarkers, and infliximab TDM compared to symptom-based management alone. In the two dose intensification strategy (DIS) groups, dose escalation was prompted by the following criteria: (1) Crohn’s Disease Activity Index (CDAI) > 220 with a CRP > 5 mg/L and/or fecal calprotectin (FC) > 250 μg/g; (2) CDAI 150–220 for two consecutive weeks with an elevated CRP and/or FC; (3) infliximab serum concentration at trough < 1 μg/mL; (4) infliximab trough level 1–3 μg/mL; and (5) infliximab trough level 3–10 μg/mL with a drop by > 50% compared with the week 14 infliximab concentration. The control group received infliximab dose escalation based on clinical symptoms (CDAI > 220 or a CDAI 150–220 in the two prior visits) alone. A stringent primary endpoint of corticosteroid-free clinical remission without ulcers, need for surgery, or the development of fistulas between weeks 22 to 54 was used.

The primary endpoint was achieved in 33% (15/45), 27% (10/37), and 40% (16/40) of patients in the DIS1, DIS2, and the control groups, respectively (p = 0.50). Furthermore, no significant differences were observed in the proportion of patients achieving secondary endpoints of absence of ulcers, endoscopic remission, or endoscopic improvement at both weeks 12 and 54. Although outcomes did not differ between the DIS and control groups, dose escalation algorithms in TAILORIX were complex and incorporated symptoms, biomarkers, and TDM: separating the independent effects of each of these components is not possible. Second, only 47% (21/45) and 46% (17/37) of patients in the DIS1 and DIS2 groups sustained therapeutic infliximab trough concentrations > 3 mg/mL between weeks 12 and 54, respectively. Also, only 5 (25%) and 7 (30%) patients in DIS1 and DIS2 groups, respectively, underwent dose escalation because of TDM.

The AGA guidelines conditionally recommend the use of reactive TDM to guide therapeutic decisions in patients with active IBD treated with TNF antagonists, recognizing that the quality of evidence is very low [10]. However, no recommendation is made regarding the use of routine proactive TDM for patients with quiescent disease. Rather, this area is characterized as a knowledge gap. Additional concerns regarding proactive TDM were also raised, including (1) the potential for inappropriate treatment changes in the context of low-titer ADAbs that are of uncertain clinical significance, (2) the unclear frequency with which TDM should be repeated, and (3) the cost associated with both testing and downstream treatment changes.

These guidelines have come under scrutiny [66] and are contrasted with recent expert consensus statements that support using TDM reactively in secondary loss of response, in patients with primary induction non-response, and periodically in patients in clinical remission, with the caveat that proactive testing should only be performed if the results are likely to impact management [67••]. Furthermore, it is suggested that patients with supra-therapeutic drug trough levels be considered for dose reduction whereas high-risk patients with sub-therapeutic trough levels and undetectable or low ADAbs should have immunomodulators added/optimized and/or dose escalation. Eighty-six percent of panelists agreed that patients in clinical remission with high-risk features, undetectable trough drug levels, and persistently high titers of ADAbs be considered for switching within or out-of-class.

Differences in the AGA guidelines and expert consensus recommendations may in part reflect differences in methodology. The AGA guidelines were developed using standards set by the Institute of Medicine and the Grading of Recommendations Assessment, Development and Evaluation framework, whereas Mitrev et al. developed the consensus statements using a modified 3-iteriation Delphi to achieve agreement. Nevertheless, both groups reiterated the need for high-quality, controlled, prospective long-term studies to better clarify the role TDM in clinical practice.

Using TDM for non-TNF-antagonist biologics

The role of measuring drug and ADAb levels for novel biologic agents such as vedolizumab, an α4β7 integrin antagonist, and ustekinumab, a monoclonal antibody targeting the common p40 subunit of IL-12/-23, is less clear. Interindividual variability in drug clearance for both treatments has been demonstrated, with differences in serum albumin, body weight, and inflammatory burden affecting drug pharmacokinetics [68, 69]. Persistent antibody presence also increases drug clearance although, interestingly, immunogenicity to ustekinumab and vedolizumab appears attenuated compared to therapy with TNF antagonists. In CD, the incidence of ustekinumab antibody formation after 1 year of treatment in the IM-UNITI phase III trial program was only 2.3% using a purportedly drug-tolerant assay [70]. Approximately 12% of patients randomized to placebo in the maintenance arm of the GEMINI trials developed ADAb to vedolizumab after exposure in induction, and 10% developed antibodies in the active treatment arm at week 66 (14 weeks after the last dose of vedolizumab) [68].

An exposure-response relationship for vedolizumab has been demonstrated in both UC and CD. In the GEMINI-1 trial, UC patients with vedolizumab trough levels in the lowest quartile (< 17 μg/mL) had clinical remission rates of only 6% compared to 37% of patients in the highest quartile (> 35.7 μg/mL) [71]. A similar exposure-response relationship was demonstrated in GEMINI-2 among CD patients although this was less robust (difference in clinical remission rates of 22% vs. 6% comparing the highest quartile > 33.7 μg/mL and the lowest quartile < 16 μg/mL) [72]. In maintenance treatment, a dose-response relationship was evident in both CD and UC patients on every 8-week dosing; however, this response was less evident in patients on every 4-week dosing where the lowest trough concentration quartile overlapped with the highest quartile of the 8-week group in terms of serum concentrations.

Interpreting the exposure-response relationship for vedolizumab is further confounded by the fact that there is complete saturation of the α4β7 receptors on peripheral lymphocytes even at every 8-week maintenance dosing and at drug concentrations as low as 1 μg/mL [73]. This suggests that higher serum concentrations would not be associated with improved efficacy but are contrasted by the clinical observation that a substantial proportion of patients recapture response with vedolizumab dose escalation [74]. Therefore, receptor saturation may not be the only mechanism mediating vedolizumab efficacy. Furthermore, given that vedolizumab purportedly affects gut-specific leukocyte trafficking, it is unclear if serum levels are an accurate approximation of drug efficacy.

Exposure-response relationships have also been described with ustekinumab [75]. Clinical remission at week 8 after induction in the UNITI-1 and UNITI-2 trial programs was positively associated with serum drug concentrations [70]. Ustekinumab concentrations of 0.9–1.2 μg/mL in quartile analysis were associated with higher rates of clinical remission and the optimal cutoff determined in ROC analysis was a trough concentration of 1 μg/mL (area under the curve 0.64, p < 0.003). Higher trough concentrations were associated with increased rates of CRP normalization (52% vs. 25%, p < 0.0001 for trough concentration of above 1.1 μg/mL) and endoscopic response (40% vs. 8%, p < 0.003 for trough concentration of above 0.5 μg/mL) [75]. A higher serum trough concentration of > 4.5 μg/mL measured using a drug-tolerant HMSA after 26 weeks of treatment was reported to be associated with improved biomarker and endoscopic response in a real-world cohort, although this study did not incorporate intravenous induction and timing of assessment was not standardized [76].

In summary, although exposure-response relationships have been demonstrated with both ustekinumab and vedolizumab, the utility of TDM for optimizing treatment with these agents is still to be delineated, particularly given important differences in mode of action, immunogenicity, and drug pharmacokinetics of these novel agents compared to TNF antagonists.

Measuring drug at the site of action

Little is known about colonic mucosal concentrations of infliximab and TNF in IBD patients and whether this correlates with either (1) serum or stool drug concentrations or (2) clinically important outcomes. The recent proof of concept ATLAS (Anti-TNF Tissue Level and Antibodies in Serum) study demonstrated that serum TNF antagonist concentrations correlated with tissue concentrations in uninflamed, but not inflamed tissue [77]. Furthermore, TNF antagonist concentrations in tissue correlated with the degree of endoscopic inflammation, except for tissue with severe inflammation. Patients with active mucosal disease had high rates of serum-to-tissue drug concentration mismatch. This study demonstrated that in patients with active disease, serum concentrations may not accurately guide clinical management of IBD.

A recent study by Yoshihara et al. [78] confirmed the correlation of serum TNF antagonist with tissue TNF concentration in non-inflamed tissue. A total of 25 CD patients were treated with infliximab (n = 15) or adalimumab (n = 10). During maintenance therapy, tissue concentrations were measured 2 weeks after drug administration. Inflamed tissue had higher TNF antagonist concentrations and lower TNF concentrations than uninflamed tissue. No correlation between tissue concentrations and prospectively scored clinical or endoscopic outcomes was found. Drug concentrations only correlated between uninflamed tissue and serum. Using a non-conventional outcome (therapeutic intervention requirement after 6 months), the optimal cutoff concentration in non-inflamed tissue was 1.3 μg/g. In patients with a TNF antagonist concentration > 1.3 μg/g in non-inflamed tissue, the time to therapeutic intervention was longer compared to that in patients with lower concentrations.

However, further work on this subject is needed for several reasons. The ATLAS study did not provide quantitative data on TNF antagonist concentrations. Only 12 patients were on infliximab, with an unknown proportion of CD and UC within this group. Only 6 patients with UC were included in the study, of which an unspecified portion received infliximab or adalimumab. Furthermore, while 43 uninflamed biopsies were analyzed, only 17 inflamed biopsies were analyzed and an unknown proportion of these were from either infliximab- or adalimumab-treated patients. Lastly, neither objective endoscopic nor histologic scores were assessed. In the study by Yoshihara et al., tissue concentrations of individual therapies were not provided, and only one patient receiving adalimumab had measurable concentrations in non-inflamed tissue.

Drug development of locally acting agents

A better understanding of the disposition of systemically absorbed drugs into the mucosal tissue and the correlation with clinically important outcomes will be key for developing compounds with proven mechanism of action that are being designed to act locally in the gut and limit systemic absorption. Sandborn et al. [79] recently presented the results of a Phase 2b randomized, double-blind, placebo-controlled induction study in patients with moderately-to-severely active UC who were treated with PTG-100, an orally administered gut-restricted peptide α4β7 antagonist. PTG-100 showed a dose-dependent increase in clinical remission, endoscopic response, and histologic remission with maximal efficacy at the 900-mg dose. These efficacy results require confirmation in subsequent trials, including systemic and local exposure-response analyses to understand inter- and intra-individual variability in pharmacokinetics and pharmacodynamics. Panes et al. recently presented the results of a Phase 1b randomized, double-blind, placebo-controlled study in patients with moderately-to-severely active UC who were treated with TD-1473, an orally administered and intestinally restricted pan-JAK inhibitor [80]. TD-1473 was generally well tolerated over 4 weeks with evidence of signals for clinical and biomarker activity and colonic tissue concentrations of TD-1473 that were higher than plasma concentrations and in the range needed for JAK inhibition. Based on these early results, the local delivery of compounds (both peptides and small molecules) with proven systemic mechanism of action may prove to be a promising strategy leading to orally administered, effective, yet safer drugs because of limited systemic exposure. Understanding drug disposition after systemic or oral administration will be key for efficient dose finding and early drug development.

Conclusions

The adoption of TDM for patients with IBD undergoing treatment with thiopurines or TNF antagonists has offered a more personalized approach to optimizing therapy. The benefits of reactive TDM for defining mechanisms of loss of response or adverse events have been well established. Despite these advances, our review also highlights areas that require further investigation. First, although many clinicians employ proactive TDM for TNF antagonists, the evidence to support this practice is primarily observational. Second, the role of TDM for patients treated with biologics with alternative mechanisms of action other than TNF blockade is unclear, as these agents have different immunogenicity and pharmacokinetic profiles. Third, recent drug development of effective locally acting therapies may change our approach from measuring systemic drug concentrations to measuring drug at site of action. The development of these concepts will mark another important step forward in personalized IBD care.