Dashboard-Driven Accelerated Infliximab Induction Dosing Increases Infliximab Durability and Reduces Immunogenicity

Abstract

Background and Aims

Accelerated infliximab (IFX) induction is often based on clinical parameters as opposed to pharmacokinetics (PK). We aimed to investigate the impact of dashboard-guided optimized induction dosing on IFX durability and immunogenicity in a real-world inflammatory bowel disease (IBD) setting.

Methods

Pediatric and adult IBD patients were enrolled in a prospective single arm intervention trial. Cumulative data from each infusion (INF), weight, albumin, C-reactive protein, IFX dose, IFX trough level, and antidrug antibody presence were used to inform subsequent INF dosing. Forecasts driven by adaptive Bayesian modeling were generated to maintain trough levels for the third (INF3) and fourth (INF4) infusions of 17 μg/mL and 10 μg/mL, respectively. The primary outcome was proportion of patients prescribed accelerated dosing (AD) intervals by INF3 (<22 days) or INF4 (<49 days). Secondary outcomes included week 52 clinical and PK outcomes. Multivariate analyses and Kaplan-Meier curves compared outcomes based on adherence to dashboard forecasts.

Results

Of the 180 per-protocol population, AD was forecast for 41% (INF3) and 69% (INF4) of patients with median intervals of 17 (INF3) and 39 (INF4) days. Baseline age >18 years, albumin >3.5 g/L, and 10-mg/kg dose were independently associated with lower rates of AD by INF4. Nonadherence with the INF4 forecast (n = 39) was an independent predictor of antidrug antibody (P < .0001) and IFX discontinuation (P = .0006). A total of 119 of 123 patients on IFX at week 52 were in steroid-free remission.

Conclusions

The application of a PK dashboard during induction can optimize dosing early to improve IFX durability and immunogenicity.

Introduction

Infliximab (IFX), an anti-tumor necrosis factor (anti-TNF) monoclonal antibody, is effective for the treatment of inflammatory bowel disease (IBD) patients.^1,2 IFX is approved for Crohn’s disease (CD) and ulcerative colitis (UC) at 5 mg/kg with induction doses at 0, 2, and 6 weeks followed by maintenance every 8 weeks. The label allows escalation to 10 mg/kg every 8 weeks in adult CD patients experiencing loss of response (LOR). Despite robust efficacy data from clinical trials, approximately 30% of patients fail to respond to induction, and approximately 50% lose response during maintenance.³ Primary nonresponse was thought to be mechanistic; however, we have gained better understanding of the role of IFX pharmacokinetics (PK) in patient response.^4,5

Secondary LOR is often attributed to low trough levels (TLs), frequently in the presence of antidrug antibodies (ADA).⁶ ADA at levels >10 U/mL often requires IFX discontinuation.⁷ Studies have demonstrated relationships between TLs, ADA, and efficacy.^8,9 Preventing low TLs and subsequent ADA is critical to maintain a durable response. Therapeutic drug monitoring (TDM) in IBD practice is predominately reactive, with TL and ADA measured when LOR occurs.^10,11 Proactive TDM facilitates early dose optimization, preventing LOR.^12,13 Studies reported proactive dose optimizing IBD patients early or later in maintenance improved outcomes, including fewer surgeries and hospitalizations.^14,15 The PANTS (personalized anti-TNF therapy in CD) study reported that the only factor that was independently associated with primary nonresponse, and LOR by week 54 was low drug concentration at week 14.¹⁶ A follow-up study found a genome-wide significant association between HLA-DQA1∗05 and the development of antibodies against anti-TNF agents.¹⁷ We reported that proactive dose optimization based on TDM before the first maintenance infusion, at week 10, improved IFX durability and decreased ADA.¹⁸ However, true proactive TDM should be initiated during induction when the inflammatory burden and drug clearance is highest in order to optimize durability. Papamichael et al^19,20 reported that IFX TLs >28.3 μg/mL at week 2 and ≥15 μg/mL at week 6 predicted short-term mucosal healing, suggesting that optimization should start during induction. A recent pediatric study reported similar target concentrations for infusion 2 (INF2) and INF3.²¹

Currently, accelerated IFX induction dosing is most often based on clinical parameters as opposed to PK. The recently published NOR-DRUM (NORwegian DRUg Monitoring) Trial, which included multiple different immune conditions with 35% being IBD, reported that proactive TDM was indeed more effective than treatment without TDM after 52 weeks.²² Given the complexity of IFX clearance in IBD patients as compared with other immune conditions, tools that can help identify those individuals who do need early optimization to minimize immunogenicity and increase IFX durability are needed. Dashboards integrate individual clinical and PK data to generate dosing recommendations to achieve prespecified target TLs using adaptive Bayesian forecasting.²³ Our initial exploration of one such dashboard (iDose; Projections Research, Phoenixville, PA, USA) demonstrated that clinical intuition alone was not as accurate at predicting the need for dose escalation.²⁴ Strik et al²⁵ recently demonstrated the superiority of Bayesian dashboard guided dosing over standard dosing of IFX during maintenance but did not assess the use of a dashboard during induction. A recent pediatric study demonstrated that a population PK model utilizing weight, albumin, ordinal drug-tolerant ADA, and erythrocyte sedimentation rate accurately predicts IFX trough concentrations during maintenance therapy in real-practice pediatric patients with IBD.²⁶

We hypothesized that adaptive dosing dashboards that integrate both clinical and PK data during induction would more precisely dose-optimize IFX and improve therapeutic outcomes and IFX durability. We report herein the dosing prescribed by the dashboard, PK parameters, and clinical outcomes from the real-world application of dashboard-guided proactive IFX optimization starting in induction in IBD. Until now, proactive IFX dose optimization based on induction TDM using a Bayesian dashboard has never been tested.

Methods

Study Population

IBD patients 6-45 years of age initiating IFX at the Mount Sinai Therapeutic Infusion Center were eligible to enroll in the Precision IFX (Precision Infliximab) study. The starting IFX dose and whether patient received concomitant immunomodulators (IMMs) was at the discretion of the treating physician as part of standard of care. To be included in the primary per protocol analysis, patients must have had their first 2 infusions at weeks 0 (INF1) and 2 (INF2) as part of standard of care and a dashboard prescribed interval for their third induction infusion (INF3). Total duration of the study was 52 weeks. The study was approved by Icahn School of Medicine at Mount Sinai Institutional Review Board (HS#: 15-00702) (https://clinicaltrials.gov/ct2/show/NCT02624037).

Data Collection

At enrollment, demographic and baseline characteristics were recorded, including sex, age, IBD subtype, disease duration, medication history, prior anti-TNF exposure, disease location and behavior per Montreal classification.²⁷ At baseline and subsequent infusions up to 1 year, weight, IFX dose, disease activity, recent steroid and IMM use, and standard-of-care labs such as albumin and C-reactive protein (CRP) were recorded. Disease activity was calculated using the Harvey-Bradshaw index (HBI) for CD and partial Mayo (pMayo) scores for UC.

Serum IFX Concentrations and ADA

Blood for IFX concentrations and ADA were collected prior to INF2, INF3, and prior to the first maintenance infusion (INF4). After INF4, patients had protocolized draws at INF 6, 9, 12, or week 52. Serum was shipped and analyzed using a homogeneous mobility shift assay (Prometheus Laboratories, San Diego, CA, USA) described previously.^13,28

Description of PK Dashboard

A Bayesian dashboard (iDose Projections Research) developed to facilitate IFX dosing was utilized to determine dose regimens that would meet or exceed physician-selected TLs for each dose.²⁴ Bayesian inference links PK models with observed data. Interpretation of observed concentrations takes information from underlying pharmacology (described by model), variability (arising from different patient factors), and prior knowledge.²⁹ In this system, forecasts can be made without TLs, although forecast is imprecise.²⁴ Dashboard performance improves as individual concentration data accumulates, allowing the model parameters to be updated and more reflective of individual patients. Dashboards allow interpretation of time-dependent patient factors (eg, weight and albumin), which is important because these factors are expected to change as response to therapy changes, and are particularly relevant for children, who are growing. iDose has undergone substantial evaluations including testing of multiple assays in the software.^30,31

IFX Dosing Protocol

Patients received physician prescribed dose (5 mg/kg or 10 mg/kg) per protocol at INF1 and INF2. Patient factors such as IFX dose, weight (kg), albumin, CRP, concomitant IMM use from the first 2 infusions, IFX concentrations, and presence of ADA (yes vs no) from INF2 were inputted to inform INF3 dosing. The same variables from the first 3 infusions along with IFX concentrations, and presence of ADA from INF2 and INF3 were used to inform INF4 dosing (first maintenance infusion).

TLs of 17 µg/mL and 10 µg/mL were targeted for INF3 and INF4, respectively. These TLs were derived using the Xu model to simulate 5 mg/kg during induction and maintenance for a 70-kg adult male patient, with no ADA, albumin of 4.0 g/dL, low CRP.²⁴ IFX PK data from pediatric IBD patients, juvenile rheumatoid arthritis, and Kawasaki disease were combined with adult IBD patient data to generate this model. Simulations demonstrated that 5-mg/kg dosing during induction results in TLs ≥3 μg/mL at week 14, although reports suggest that IFX TLs should approximate 10 μg/mL for optimal response.^13,19,32

Treating physicians were provided the model-prescribed date at which patients would reach a target of 17 µg/mL for INF3 and dosing interval was accelerated accordingly (Supplementary Figure 1). Of note, decision on mg/kg dosing changes during induction was at the discretion of the treating physician. Similarly, the date to reach a TL of 10 μg/mL for INF4 was provided for interval adjustments and patients were dosed accordingly. In addition to the INF4 prescribed interval, clinicians were also provided a table to guide remaining maintenance dosing (interval and mg/kg dosing) aimed at maintaining TLs at 10 μg/mL (Supplementary Table 1). This could lead to mg/kg dosing to be escalated as early as INF4 based on desired interval for INF5 and beyond. Forecasted intervals >8 weeks were not provided to the clinicians for this study.

Outcomes Measures

Primary outcome was proportion of patients prescribed accelerated dosing (AD) intervals by INF3 (<22 days interval between INF2 and INF3) or INF4 (<49 days between INF3 and INF4) based on a priori TL targets described previously per protocol. Non-AD patients were those patients not prescribed accelerated dosing and thus received INF3 >22 days from INF2 or prescribed to receive INF4 >49 days from INF3. Secondary outcomes included durability of IFX defined as proportion of patients still on IFX at week 52 in the per-protocol population, frequency of steroid-free clinical remission (HBI <5 for CD or pMayo <2 and off all corticosteroids for at least 4 weeks) and frequency of CRP normalization at INF4 and at week 52 in those who received INF4, PK profiles at week 52 or at termination, and rates of ADA. Reason for termination from the study was classified as either protocol deviation or treatment failure due to primary nonresponse if during induction, ADA development, infusion reaction or persistent inflammation or loss of response during maintenance.

The impact of nonadherence with dashboard dosing predictions for INF3 and INF4 on IFX durability and ADA frequency and ADA-free survival was also analyzed for the per-protocol population. Nonadherence with dashboard-prescribed induction interval defined patients who were late in receiving a dose at the prescribed interval of INF3 and INF4 (>3 days of INF3 and >7 days of INF4). The reason for nonadherence was not collected but could have been related to any patient or physician factor.

Statistical Analysis

Clinical and laboratory data were summarized by descriptive statistics. Median and interquartile range (IQR) were calculated for continuous variables and compared using Mann-Whitney U test or Wilcoxon signed rank test. Frequencies were computed for categorical variables and compared using McNemar’s test or chi-square test when appropriate. Kaplan-Meier curves for IFX durability and ADA development were constructed based on adherence to the INF3 or INF4 forecast. Survival curves and associated comparisons between groups and their corresponding P values were derived using the log-rank test. Cox proportional hazards (CPH) models then investigated the associations between IFX durability and ADA-free survival for multiple variables including patient type (UC vs CD, pediatric vs adult), and adherence to recommended dose intervals. A separate CPH evaluation was conducted looking at factors that were associated with the need for AD by IFN4. Two-sided P values ≤.05 were considered statistically significant.

Results

Patient Population

A total of 189 patients enrolled, but 9 patients did not complete INF1 and INF2 per protocol nor did they receive a prescribed interval for INF3. Of the 180 patients (per-protocol population) completing INF1 and INF2, 53% initiated IFX 5 mg/kg and 47% 10 mg/kg as prescribed by their treating physician (Table 1). The median age of all patients was 14.7 (IQR, 12.1-17.2) years with 78% pediatric onset and median disease duration of 0.4 (IQR, 0.10-1.87) years. At baseline, 69% (n = 125) of patients had CD and most had ileocolonic (L3) involvement, and of the 55 UC patients, 75% had extensive colitis (E3). Almost 80% of patients were bio-naïve, and <10% were on IMM at IFX initiation. Steroid use, however, was noted in 49% of patients. Median baseline HBI and pMayo scores were 2 (IQR, 0-7) and 6 (IQR, 4-8), respectively. Of the 47% (n = 85) of patients in clinical remission, 81% (n = 69) were on steroids or had abnormal CRP.

Table 1 also highlights the baseline characteristics that were statistically different between those who were prescribed 5 mg/kg or 10 mg/kg. Albumin was the only biomarker that showed significance (3.4 [IQR, 3.1-3.9] g/L vs 2.9 [IQR, 2.5-3.5] g/L; P < .001). The other most notable difference was that more CD patients were initiated on 5 mg/kg and more UC patients were initiated on 10 mg/kg at the discretion of their provider (P = .003), and 5-mg/kg patients were more likely to be bio-naïve (P = .02).

Induction Dosing and Prescribed Dosing Intervals

Dose interval predictions were performed on all 180 patients who received both INF1 and INF2. Although prescribed a dosing interval, 3 patients never received INF3, as they had either a protocol deviation (moved to home infusion = 1) or early termination (ADA and infusion reaction = 1, surgery = 1). Of the 177 patients who received INF3 and were prescribed a dosing interval for INF4, 8 patients were either protocol deviations as not infused at MSH (n = 4) or early termination (primary nonresponse = 2, ADA positive = 1, and infusion reaction = 1), resulting in 169 total patients who received a dose at INF4. Of these 169, 131 (78%) were dosed at 10 mg/kg. As Figure 1 illustrates, 12 patients were dose-escalated between INF2 and INF3, and 40 patients were dose-escalated between INF3 and INF4 at the discretion of their treating physician. The increased number of patients receiving 10 mg/kg at INF4 could largely be explained by the fact that clinicians were provided with a maintenance dosing table (Supplemental Figure 1B) so that they could choose the ideal dose and interval that would be most feasible for their patient to maintain a trough of 10 µg/mL (Supplementary Table 1).

AD was forecasted for 41% (n = 73 of 180) and 69% (n = 122 of 177) of patients by INF3 (target trough 17 µg/mL) and INF4 (target trough 10 µg/mL), respectively, increasing to 80% by INF4 if induced with 5 mg/kg for the first 3 infusions (Figure 1). Importantly, 11 of the 12 who escalated to 10 mg/kg for INF3 still required interval shortening for INF4. Baseline age >18 years, albumin >3.5 g/L, and starting IFX dose of 10 mg/kg were independently associated with lower rates of AD by INF4 (Table 2). Two-thirds (n = 49) of the 73 AD by INF3 were pediatric patients and 24 (33%) were adult patients. A similar breakdown was seen for INF4 with 64% (n = 78) and 36% (n = 44) of those 122 prescribed AD for INF4 being pediatric and adult patients, respectively. The PK dashboard inputs from INF2 and INF3 that characterized the prescribed dosing intervals for INF3 and 4 respectively are noted in Supplemental Tables 2 and 3. Of the 169 who had at minimum an INF3 and INF 4 forecast; 5% (n = 9 of 169) needed escalation for INF3 only and 32.5% (n = 55 of 169) needed escalation for INF4 only. A total of 32.5% (n = 55 of 169) needed escalation for both INF3 and INF4 and 30% (n = 50 of 169) needed escalation for either INF3 or INF4.

The median forecasted interval for the AD group was 17 (IQR, 14-19) vs 30 (IQR, 26-37) days for the non-AD group for INF3 and was 33 (IQR, 26-39) vs 59 (IQR, 53-64) days for INF4 (Figure 2). INF3-observed IFX TL was higher than the target of 17 µg/mL in the non-AD group, and the AD group TL was just below target (25.3 [IQR, 15. 2-40.3] μg/mL and 15.7 [IQR, 9.8-23.3] μg/mL, respectively) (Figure 3A). Similarly, INF4-observed IFX TL was higher than the 10-µg/mL target in the non-AD group and was on target for the AD group (15.4 [IQR, 10.9-22.9] μg/mL and 10.9 [IQR, 6.9-17.6] μg/mL, respectively) (Figure 4A).

Of the 177 patients who had a forecast and received INF 3, 42 (23.7%) did not adhere and were late to INF 3 with a median TL of 13.6 (IQR, 8.7-19.4) μg/mL as compared with 23.3 (14.9-36.7) μg/mL for the 135 patients who were on time (P < .0001) (Figure 3B). Of the 42, 15 (35.7%) reached the target of 17 µg/mL or above even if they were late. Of the 135 adherent to their INF3 forecast, 44 were in the AD group with a median TL right on target (17.9 [IQR, 12.6-29.5] μg/mL). Of the 169 patients who had a forecast and received INF4, 39 (23%) did not adhere and were late to INF4 with a median TL of 7.0 (IQR, 4.5-10.8) μg/mL as compared with 14.0 (IQR, 10.1-20.2) μg/mL for the 130 patients who were on time (P < .0001) (Figure 4B). Of the 39, 11 (28%) reached the target of 10 µg/mL or above. Of the 130 adherent to their INF4 forecast, 82 were in the AD group with a median TL close to target at 12.9 (IQR, 9.5-19.5) μg/mL (Supplemental Figure 2).

Maintenance Dosing Interval

Continuing to target a TL of 10 µg/mL throughout maintenance starting with INF4, only 5 of the 123 (4%) patients were receiving 5 mg/kg every 8 weeks at week 52 with all having received 5 mg/kg throughout induction. The remaining 119 patients were receiving a median dose of 10.0 (IQR, 9.9-10.1) mg/kg at a median interval of 6.1 (IQR, 5.1-8.0) weeks. Eleven (9%) of the 123 were receiving a dose of >12.5 mg/kg. Overall, compared with dose/interval at INF4, 95 (77%) patients remained on the same dose or interval (median 10.0 [IQR, 9.9-10.0] mg/kg, median 6.6 [IQR, 5.1-7.9] weeks), and 26 (21%) experienced additional dose optimization with a median dose of 10.0 (IQR, 10.0-14.9) mg/kg at a median interval of 6 (IQR, 5.1-7.4) weeks. The remaining 2 (2%) were able to de-escalate by week 52 to 5 mg/kg every 8 weeks. Median week 52 TL for all 123 patients was 13.6 (IQR, 10.1-19.2) μg/mL. Although baseline albumin was not associated with week 52 interval or dose, a diagnosis of UC was associated with a shorter interval, not dose, at week 52 as compared with CD (5.57 [IQR, 4.27-6.57] g/L vs 6.28 [IQR, 5.09-7.99] g/L; P = .03).

Clinical Outcomes and IFX Durability

At the time of INF4, 67% (n = 113) of the 169 patients were in steroid-free clinical remission, 78% (n = 128) had normal CRP, and 55% (n = 93) had both steroid-free remission and normal CRP. No differences were observed for steroid-free remission or the combined endpoint of steroid-free remission and normal CRP between the 2 groups at INF4 (AD vs non-AD: 77% vs 58% and 75% vs 54%, respectively). However, CRP normalization rates were significantly lower when patients did not adhere with their prescribed dosing interval (92% vs 58%; P = .003 for adherent vs nonadherent).

Overall, 123 (68%) of the 180 per-protocol population, 73% of the 169 who got INF4, were still on IFX at week 52, with 57 terminating after a median of 23.9 (IQR, 12.3-34.0) weeks (range, 0.6-47 weeks): 31 (54%) were due to treatment failure (5 during induction) and 26 (6 during induction) were due to a protocol deviation, with 12 (46%) moved to home infusions, 11 (42%) moved to another geographic region for care, and 3 (12%) moved to an unknown location of care. Of the 31 total treatment failures; 24 (77%) failed after IFN4 at a median 25.3 (IQR, 16.4-32.5) weeks at a dose of 10.1 (IQR, 10.0-15.0) mg/kg every 4.2 (IQR, 4.0-5.3) weeks.

All but 4 of the 123 patients still on drug at week 52 were in steroid-free clinical remission. Only 1 of these 4 was in the nonadherent induction group. Removing those 22 lost to follow-up increases the rate of steroid-free remission to 81% (n = 119 of 147). Of the 119 in steroid-free clinical remission at week 52, 96 (81%) were also in steroid-free clinical remission at INF4. No patient had an elevated CRP at week 52 compared with 23 (17.8%) patients at baseline (P < .0001). Median week 52 albumin was 4.1 (IQR, 3.2-4.8) mg/L, as compared with 3.2 (IQR, 1.9-4.4) mg/L (P < .0001) at baseline.

Adherence significantly impacted IFX durability. Patients who did not adhere to at least one or both forecasted intervals during induction progressed faster to treatment discontinuation compared with those who were adherent for all induction forecasts (log-rank P = .0006) (Figure 5). The CPH analysis showed that adherence to INF4 interval forecast regardless of adherence to INF3 forecast, a diagnosis of UC and baseline prednisone use were associated with lower IFX durability (Table 3). Those patients who did not undergo AD at INF3 and or INF4, despite not being forecasted to do so, did have lower rates of both clinical (log-rank P = .04) and steroid-free remission plus CRP normalization (log-rank P = .04).

ADA Frequency

Of the 23 of 180 individuals who developed ADA throughout the duration of the study including at INF2, 8 terminated the study before week 52 and 5 were lost to follow-up. Of the 23, 14 (61%) developed ADA by INF4. Four were first detected at INF2 (10 [IQR, 7.9-10.1] U/mL); 2 cleared ADA before INF3, 1 cleared ADA before INF4, and 1 discontinued IFX due to an infusion reaction at INF2. Seven patients had ADA first detected at INF3 (7.2 [IQR, 5.2-11.3] U/mL), with 3 clearing their ADA by INF4, 1 discontinuing IFX after due to an infusion reaction at INF3 and 1 lost to follow-up. Of the 5 patients with detectable ADA at INF4 (6.7 [IQR, 4.8-8.3] U/mL), 3 were de novo while 2 had detectable ADA at INF3. The median TL for the 5 patients with ADA was 8.2 (IQR, 4.8-9.5) µg/mL at INF4, and 3 of them cleared ADA by week 52, with 1 clearing by INF5 and 2 clearing by INF6.

The 9 patients who developed de novo ADA after INF4 after a median of 25.5 (IQR, 21-34) weeks had an ADA level of 5.2 (4.5-8) U/mL at first occurrence and corresponding TL of 6.1 (IQR, 4.9-15.6) μg/mL. Of these 9 patients, 4 discontinued IFX at a median of 34.3 (IQR, 28.2-36.9) weeks and corresponding TL of 1.4 (IQR, 0.9-5.5) μg/mg and ADA level of 11.5 (IQR, 9.5-12.3) U/mL. The remaining 5 completed 52 weeks of IFX in steroid-free remission, despite 3 of 5 having persistent ADA through week 52 (median ADA level 3.59 [IQR, 2.3-3.6] U/mL and median TL of 6.7 [IQR, 6.4-24.3] μg/mL).

Like IFX durability, patients who did not adhere to at least one or both forecasted intervals during induction had lower ADA-free survival compared with those who were adherent for all induction forecasts (log-rank P < .0001) (Figure 6). Based on CPH, nonadherence to INF4 interval forecast regardless of adherence to INF3 forecasted was found to be a statistically significant predictor (Table 4). Those patients who did not undergo AD at INF3 and or INF4, despite not being forecasted to do so, did have lower rates of IFX durability (log-rank P = .02) and ADA-free survival (log-rank P = .01) at week 52.

Discussion

In this real-world study, the majority of IBD patients induced with standard-of-care dosing of IFX were prescribed an AD regimen by INF4. Needing a dose or interval escalation due to LOR with IFX is not a unique concept in the routine management of IBD patients³³; however, this is the first study to proactively optimize IFX based on induction TDM guided by an adaptive dosing dashboard. Median dosing intervals were accelerated by approximately 2 weeks and 4 weeks for INF3 and INF4, respectively. Additionally, <8% of patients were ADA positive by end of induction; most ADAs were transient and did not impact treatment durability despite minimal concomitant IMM use. Moreover, the largest driver of ADA was nonadherence with the prescribed dosing interval. Overall, 23 (13%) of 180 patients developed ADA during the study, and in some cases, ADA was transient. Treatment success rates for induction and maintenance were substantially higher than commonly reported. Of the 180 enrolled patients, 169 total patients received a dose at INF4, which is only a 6.11% early treatment failure rate, and 123 of the 180 remained on IFX at week 52, suggesting a 31.7% failure rate for the first year. However, these numbers should be considered in light of the fact that 26 of the 180 patients initially enrolled changed to home infusions, moved to other clinical care or were lost to follow-up and would not be considered treatment failures. Most patients remaining on drug were in steroid-free remission and the use of IMM was minimal.

The exposure-response relationship for IFX during maintenance has long been evaluated, giving rise to reactive TDM algorithms to guide dosing recommendations. The principal concern about reactive TDM is that low concentrations typically precede ADA development, and waiting until symptoms re-emerge fosters ADA development, and treatment discontinuation. These results advocate for proactive TDM in which dose is optimized based on individual concentrations instead of symptoms. To date, there are limited studies that have proactively optimized IFX dosing even at the start of or during maintenance.^14,34 The Tailored Treatment With Infliximab for Active Crohn\'s Disease (TAILORIX) Study, which targeted a low TL of ≥3 μg/mL pre-INF4 (week 14), showed that the majority of patients require dose escalation by week 14, whether driven by symptoms or due to low trough levels. Based on prior literature, a TL of 3 µg/mL was the prespecified target for the PRECISION study published by Strik et al.²⁵ In this randomized study, dashboard-guided IFX dosing resulted in a significantly higher proportion of patients maintaining clinical remission during 1 year of follow-up compared with patients who continued IFX treatment without proactive dose adjustments. The dashboard advised dose intensification in 35% of patients, no change in 15%, and 50% had dose de-escalation in the PRECISION study. Moreover, the first 2 dose recommendations were the most significant. Interestingly, we similarly showed that once a patient is optimized and achieved their ideal dose regimen by INF4, most patients remained on the same regimen out to week 52. The Trough Concentration Adapted Infliximab Treatment (TAXIT) Study also demonstrated that once optimized, there was no advantage to TL-driven dose adjustments over using clinical factors to drive escalation.¹⁴

As compared with the many studies retrospectively evaluating the exposure response association in maintenance, there are few studies evaluating this relationship during induction. The recent NOR-DRUM trial demonstrated that proactive TDM, applied to management of multiple immune diseases, was more effective than treatment without TDM in sustaining disease control without disease worsening.²² It is intuitive that when the disease is most active (eg, during induction), higher doses or more frequent dosing may be needed compared with once remission is achieved given the more rapid initial clearance. For this study, a TL of 17 µg/mL for INF3 was derived using the Xu model and described in detail by Mould et al.^29,35 Additional PK data and models support this target TL, most recently in pediatrics.^19-21 In the post hoc analysis of the TAILORIX study was that IFX TLs >23.1 mg/L at week 2 and >10.0 mg/L at week 6 were associated with endoscopic remission at week 12 (positive predictive values, 72% and 76%; negative predictive values, 65% and 59%, respectively).³⁶ Although there is no consensus on the ideal pre-INF4 TL, our earlier study showed that a TL of at least 7 was most predictive of persistent remission and subsequently proactive dose optimization targeting a TL of 10 μg/mL was associated with better IFX durability.^13,18 The PANTS study similarly showed that week 14 TL of at least 7 µg/mL was associated with both primary nonresponse at week 14 and treatment failure at week 54.¹⁶ Additional findings from the post hoc analysis of TAILORIX study, however, reported that a week 14 cutoff TL of 7.8 µg/mL was associated with radiologic remission at week 54, and TL >10.6 μg/mL upon dose escalation were associated with the absence of ulcers at week 54.³⁷ Our study demonstrates that targeting a TL of 10 µg/mL was associated with both INF4 and week 52 biomarker and steroid-free clinical remission and was achievable within standard dosing regimens deployed in current practice.

Our study has several limitations. This was a nonrandomized, single-arm, single-referral center pragmatic study with more pediatric patients than adults. However, weight, the biggest difference between pediatric and adult patients, is accounted for in the model. The lack of an active control arm makes interpretation of safety and efficacy more difficult, although the number of patients studied is sufficient to show positive trends when compared with historical data. Owing to the lack of a control arm for the intervention, this study was designed to use a “per-protocol” rather than an intention-to-treat analysis. However, within the analyzed population, the outcomes for patients who did vs did not adhere to the prescribed interval were compared. In addition to enrolling both adults and pediatrics, the study enrolled both CD and UC patients, and the starting dose of IFX was not controlled, but rather was at the discretion of the treating physician. Additionally, there was no protocol-specified interval or dose that all patients or treating physicians were required to follow in maintenance. Instead, physicians were given various dosing permutations to achieve a target maintenance TL of 10 µg/mL. The median dose and interval at the end of follow up was 10 mg/kg every 6 weeks; for a minority of payers, this would not be an approved dosing regimen. Despite this, approvals were deemed more justified by payers upon review of dosing forecasts. These factors introduce more variability into the results but as a real-world study, the results support several conclusions. Protocol forecasts did not include dosing intervals >4 weeks for INF3 or >8 weeks for INF4 and remaining maintenance infusions. Thus, we did not adjust for the non-AD patients whose clearance slowed during induction if exposure is adequate. As our study shows, the principal issue is underdosing in induction, and attempts to de-escalate in maintenance perhaps by targeting a TL closer to 7 µg/mL may be more cost-efficient and merits further exploration. Despite the strong clinical outcomes in the real-world setting, the lack of investigated stringent therapeutic outcomes such as endoscopic and histologic remission is a limitation of the study. Another limitation is that the reason for nonadherence to the forecast was not collected.

Conclusions

This is the largest proactive IFX dose optimization study using personalized PK profiles to proactively dose optimize during induction. The iDose dashboard is designed as a clinical decision support tool, and its ability to adapt to patient schedules while still improving outcomes is a testament to its effectiveness. Our study demonstrated that the standard 8-week interval dosing between end of induction and first maintenance is too long for many patients, and precision dosing can minimize ADA development and maintain effective troughs without needing to introduce IMMs, which will improve safety and adoption of anti-TNF therapy. Based on our data, the stage is set to formally compare effectiveness and safety of proactive dashboard-optimized IFX monotherapy vs current standard-of-care approach of combination therapy with reactive (symptom-based) TDM. Automation of adaptive dosing dashboards and point-of-care testing will move model-based dosing forward.

Acknowledgment

This work would not have been possible without the generous support of Jeff and Tessie Nedelman and the Kaufman Family Fund. The authors would like to acknowledge Prometheus Biosciences for performing the drug assays at no cost. The data underlying this article will be shared on reasonable request to the corresponding author.

Author Contributions

Guarantor of article: M.C.D. Specific author contributions: M.C.D. was the principal investigator of the study. B.L.P., S.S.T., H.R.M., and M.M. contributed to the data collection and statistical analysis of the data. D.R.M. contributed to the design and statistical analysis of the software intervention. All authors contributed to the drafting and revising of the work. M.C.D., as corresponding author, confirms that all authors approve and have seen the final text.

All authors approved the final version of the article, including the authorship list.

Conflicts of Interest

M.C.D. is a consultant for Prometheus Labs, Janssen, AbbVie, Takeda, Celgene, Gilead, UCB, Pfizer, Arena, Eli Lilly and co-Founder Mi Test Health. And Trellus Health DRM is the President of Projections Research Inc., a consulting company for the Pharmaceutical Industry. B.L.P., S.S.T., H.R.M., and M.M. have nothing to disclose.

References