This retrospective cross-sectional study considered patients with a diagnosis of CD or UC according to the Porto criteria (14), aged between 6 and 18 years and treated with adalimumab, enrolled at three Italian pediatric facilities (Gastroenterology, Digestive Endoscopy and Nutrition Unit, Institute for Maternal and Child Health-IRCCS “Burlo Garofolo,” Trieste; Pediatric Gastroenterology Unit, Maggiore Hospital, AUSL Bologna; Cà Foncello Hospital, Treviso) between January 2013 and September 2019. Adalimumab was started in case of treatment failure or intolerance to standard therapies (steroids, exclusive enteral nutrition and immunomodulators) or infliximab. In selected patients, adalimumab was used as first-line treatment (15). Concomitant therapy with immunomodulators was permitted. Exclusion criteria were disease needing urgent surgery, presence of an ileostomy or colostomy, infectious complications, or other contemporary uncontrolled medical conditions. Children were treated according to a therapeutic protocol consisting of subcutaneous administration of adalimumab on day 1 at 160 and 80 mg, and on day 15 at 80 and 40 mg, for body weight ≥40 or <40 kg, respectively (induction phase), followed by every other week at 40 or 20 mg for body weight ≥40 or <40 kg, respectively (maintenance phase). In case of clinical evidence of loss of response, therapy with adalimumab could be escalated by shortening intervals between injections to once a week.

Demographic and clinical data, including age at diagnosis, localization and behavior of the disease according to the Paris classification (16), and therapies previous to adalimumab, were collected. Details on adalimumab treatment included age at adalimumab start, initial dose and concomitant therapies, and need for interval shortening during maintenance. Biochemical tests during adalimumab treatment were recorded and included C reactive protein (CRP), albumin, and fecal calprotectin. Serum samples for adalimumab level measurement were collected during routine visits between adalimumab administrations and therefore not necessarily at trough, both during the induction (week 4 ± 4) and maintenance phases (week 22 ± 4, week 52 ± 4, and week 82 ± 4). Samples were stored at −20°C and were analyzed retrospectively. All patients with complete clinical response (disease activity score ≤10) to treatment after induction continued therapy with adalimumab and were followed until week 82.

Clinical disease activity was assessed using pediatric Crohn's disease activity index (PCDAI) and pediatric ulcerative colitis activity index (PUCAI) for CD (17) and UC patients (18), respectively, at baseline, at the end of induction (4 weeks), and during maintenance phase at 22, 52, and 82 weeks. The disease was defined in remission if the disease activity index was ≤10. Loss of response was considered either as clinical worsening (PCDAI/PUCAI > 10) in a patient who had previously attained clinical response or as the need for treatment intensification.

Adalimumab levels were measured using the commercial Promonitor ELISA assay (Proteomika S.L., subsidiary of Progenika Biopharma S.A., Spain) on sera from peripheral blood samples collected at the time points specified above. An ELISA kit was used according to the manufacturers' instructions, and the lower and upper limits of quantification were 0.024 and 12 μg/ml, respectively. For sera above the limit of the kit (12 μg/ml), a further dilution was required. ADAs were measured when adalimumab serum levels were <1.5 μg/ml (Promonitor Anti-Adalimumab ELISA assay; Proteomika S.L., subsidiary of Progenika Biopharma S.A., Spain). ADA levels higher than 10 arbitrary units per milliliter (AU/ml) were considered positive.

Statistical analysis was performed using the R software (version 3.4.2). The association between adalimumab concentrations and the therapeutic response was evaluated in a univariate analysis by linear mixed effect (LME) models, which allow to account for repeated measures, using patient's response to adalimumab as the dependent variable and adalimumab serum levels as the independent variable. To find the optimal drug levels for predicting efficacy, the association between adalimumab concentrations and response at the various time points was identified using logistic regression analysis. The potential confounding effect of adalimumab dose was tested by considering the difference in adalimumab weight normalized dose at different time points by LME logistic regression analysis. Receiver operating characteristic (ROC) curves were then constructed for adalimumab concentrations, to determine the optimal cutoff, using the Youden index to predict clinical response. Sensitivity, specificity, and positive and negative predictive values of the cutoff point were analyzed. An analysis of the association between adalimumab concentrations considering all time points and the clinical and biochemical parameters [clinical scores, CRP, and fecal calprotectin] was carried out by the non-parametric Spearman's test. The potential effect of demographic and clinical patients' characteristics on clinical response was tested by LME analysis. For the clinical laboratory parameters, the normality of the distribution was evaluated by Shapiro's test. The association between ADAs and drug concentrations was determined by Spearman's test. Continuous data are presented as medians with interquartile ranges (IQRs) (represented as the interval between the first and the third quartile), and categorical data are presented as absolute numbers and percentages. P < 0.05 was considered statistically significant.

Thirty-two pediatric patients with IBD were enrolled. The demographic characteristics of the population are summarized in Table 1. Twenty-eight patients (87.5%) had stopped infliximab before adalimumab, and 4 patients (12.5%) were anti-TNF naive. Infliximab was stopped because of adverse events in 13 patients (46.4%), loss of response due to anti-infliximab antibodies in 8 patients (28.6%), and loss of response during induction phase in 7 patients (25%). At adalimumab start, 8 patients (28.6%) were in clinical remission according to the PCDAI/PUCAI activity score, but had biochemical alterations (elevated CRP and fecal calprotectin levels); however, the pre-adalimumab therapy score was not associated with response or adalimumab concentration at later time points (pre-adalimumab therapy score vs. clinical response at 4, 22, 52, and 82 weeks: P = 0.47, P = 0.11, P = 0.56, and P = 0.67, respectively; pre-adalimumab therapy score vs. adalimumab concentration at 4 and 22 weeks: P = 0.88 and P = 0.86, respectively). Sixteen, 15, 14, and 12 patients were in remission at weeks 4, 22, 52, and 82, respectively. All patients were steroid free. Five (15.6%) patients lost the initial clinical response: 2 recovered spontaneously, while 3 shortened intervals between injections without the need for additional treatments (Supplementary Figure 1). Three patients (9.4%) were lost at follow-up before week 52 and 7 patients (21.9%) before week 82. No patient discontinued the treatment because of adverse event (anaphylactoid reaction). No statistical association between clinical response and demographic (age and gender) and clinical (type of IBD) characteristics of the population was found after logistic regression analysis (P > 0.05).

Serum adalimumab levels at 4 weeks were correlated with clinical remission during maintenance. Higher serum adalimumab levels at week 4 were found for patients in clinical remission compared with patients with clinically active disease at week 22 (22.72 μg/ml, IQR: 8.4 vs. 8 μg/ml, IQR: 5.08 μg/ml, P = 0.0029), at week 52 (22 μg/ml, IQR: 8.4 vs. 8 μg/ml, IQR: 4.66 μg/ml, P = 0.003), and at week 82 (22.72 μg/ml, IQR: 8.39 vs. 8 μg/ml, IQR: 4.5 μg/ml, P = 0.003) (Figure 1).

The same trend was observed also considering shorter timeframes in the induction interval and long term response (Supplementary Figure 2). However, median adalimumab levels measured at week 4 did not differ between patients in clinical remission and patients with clinically active disease at the end of induction (13.7 μg/ml, IQR: 17.13 vs. 8.26 μg/ml, IQR: 2.73 μg/ml, P = 0.71), and the same trend was observed considering median adalimumab levels at week 22 and the clinical response at week 22 (7.04 μg/ml, IQR: 9.6 vs. 7.14 μg/ml, IQR: 10.34 μg/ml, P = 0.85). Patients in remission at weeks 52 and 82 had higher adalimumab levels at week 22 compared with patients with clinically active disease (12.44 μg/ml, IQR: 9.62 vs. 6.38 μg/ml, IQR: 3.82 μg/ml, P = 0.09; and 10.8 μg/ml, IQR: 9.72 vs. 5.4 μg/ml, IQR: 5.09 μg/ml, P = 0.016, respectively) (Figure 2). Quartile analysis comparing the lowest vs. highest quartile exposure (Q1 vs. Q4) resulted in higher percentage of responder in Q4 at the different time point (Supplementary Figure 3). Patients' enrolment was done cross-sectionally and only for 7 patients; measurements were obtained both at 4 and 22 weeks.

No significant difference in adalimumab dose, normalized on patients' weight, was observed at week 4, while at week 22, patients with clinically active disease were treated with higher adalimumab weight normalized doses (Supplementary Figure 4). ROC curves were constructed to assign optimal cutoff values for adalimumab levels at the end of induction therapy and after early maintenance phase to predict clinical response at 52 and 82 weeks (Figure 3).

Optimal cutoff values were 13.85 and 7.54 μg/ml at weeks 4 and 22, respectively, to predict clinical response at week 52. The cutoff values of 13.85 and 10.51 μg/ml were obtained at weeks 4 and 22, respectively, to predict clinical response at week 82. Logistic regression analysis confirmed that patients who reached the cutoff point of 13.85 μg/ml at the end of induction (6 patients, 5 in sustained remission both at weeks 52 and 82) had a higher probability of maintaining remission at weeks 52 and 82 compared with those who did not (nine patients, only one in sustained remission at week 52; eight patients, none in remission at week 82), with an odds ratio (OR) of 40.0 [95% confidence interval (CI) 2.0–794.3, P = 0.015] and OR of 62.33 (95% CI 2.13–1822.7, P = 0.016), respectively, at weeks 52 and 82. Patients above the cutoff point of 7.54 μg/ml during week 22 (8 patients, 5 reached remission at week 52) had a higher probability of maintaining remission at week 52 compared with those who did not (13 patients, only one in sustained remission at week 52) with an OR of 20 (95% CI 1.65–241.7, P = 0.018). Patients with a cut-off point of 10.51 μg/ml during week 22 reached remission at 82 weeks (5 patients, 4 in sustained remission) compared with those who did not (14 patients, only 3 in sustained remission) with an OR of 14.66 (95% CI 1.16–185.24, P = 0.037) (Supplementary Table 1).

The area under the ROC curve (AUC) was 88.9% at week 4 and 78.9% at week 22, considering the response at week 52, and 95.6 and 73.8%, respectively, at weeks 4 and 22, considering the response at week 82. The test had a sensitivity of 83.3% and specificity of 88.8% considering week 4 [positive predictive value (PPV) 83.3%; negative predictive value, NPV 88.8%] and a sensitivity of 83.3% and specificity of 80% at week 22 (PPV 62.5%; NPV 92.3%) for obtaining sustained remission at week 52. A sensitivity of 100% and a specificity of 88.9% were obtained considering week 4 (PPV 83.3%; NPV 100%) and a sensitivity of 57.14% and a specificity of 91.66% considering week 22 (PPV 80%; NPV 78.57%) for obtaining sustained remission at week 82.

Sex, age, and IBD type were not significantly associated with adalimumab concentration. The cause of infliximab failure and the concomitant treatment with immunomodulators did not affect adalimumab levels (Supplementary Figure 5). A statistically inverse correlation (P < 0.05, Spearman's test) between serum adalimumab levels and clinical laboratory variables, including disease activity score, CRP, and fecal calprotectin was observed (Supplementary Figure 6). The adalimumab cutoff value for sustained remission (i.e., 13.85 μg/ml at the end of induction) significantly correlated with fecal calprotectin measured during the induction: patients not achieving the cutoff for adalimumab concentrations had significantly higher fecal calprotectin compared with others, and marginally significant results were observed for CRP (Supplementary Figure 7).

ADA concentrations were measured in 9 samples from 6 patients (18.8%) that showed serum adalimumab levels below 1.5 μg/ml. Only 4 samples from 2 patients (6.3%) resulted positive to ADAs. None had previously developed anti-infliximab antibodies. ADA concentrations were inversely correlated with adalimumab levels, but the correlation was marginally significant (Spearman test, P = 0.073).