Proactive Drug Monitoring Is Associated With Higher Persistence to Infliximab and Adalimumab Treatment and Lower Healthcare Utilization Compared With Reactive and Clinical Monitoring

Abstract

Background

Serum drug-level assays for infliximab (IFX) and adalimumab (ADA) are widely available and are most often obtained reactively, to determine the next steps in patients with loss of response. Studies done thus far on the use of these assays proactively, or during symptom remission, have had mixed results. Here we investigate persistence on therapy and healthcare utilization in patients on 3 drug-level monitoring strategies.

Methods

We conducted a retrospective chart review of 235 patients treated for both Crohn disease and ulcerative colitis on either IFX or ADA. Monitoring strategy was defined as proactive if patients underwent testing at predefined time points regardless of symptoms or signs of disease, reactive if done during relapse, or control if no drug levels were obtained. Groups were compared on persistence on original therapeutic at 1 and 2 years as well as on various measures of healthcare utilization during the 2-year follow-up period.

Results

Proactive drug monitoring was associated with a higher likelihood of persistence on therapy at 1 year when compared with the control (odds ratio [OR] = 4.76, 95% confidence interval [CI] = 1.65, 13.67) and reactive groups (OR = 6.10, CI = 2.19, 17.02). Similarly, at 2 years, proactive monitoring was superior to the control (OR = 5.41, CI = 2.26, 12.94) and reactive groups (OR = 4.51, CI = 1.88, 10.80). Proactive monitoring was also associated with lower healthcare utilization across almost all measures related to inflammatory bowel disease.

Conclusions

Proactive drug monitoring increases persistence on IFX and ADA in patients with ulcerative colitis or Crohn disease and decreases overall healthcare utilization in these patients.

INTRODUCTION

Two antitumor necrosis factor (TNF) agents, infliximab (IFX) and adalimumab (ADA), have emerged as effective options in the treatment of inflammatory bowel disease (IBD). However, up to a third of patients do not initially respond to induction therapy.^{1, 2} Additionally, another 40% of patients experience relapsing symptoms despite the initial success, termed loss of response.³ Several factors probably underlie this process, including pharmacokinetic issues causing rapid clearance of drug, change in mechanism of inflammation to one less reliant on TNF, nonimmune-mediated mechanisms (e.g., progressive stricture formation), and development of antidrug antibodies.^{3, 4}

The development of drug assays for these agents led to studies that showed a positive correlation between serum drug concentrations and rates of clinical response and mucosal healing.^5–7 The findings from these studies support the utility of assessment of drug levels to help guide decision-making, termed therapeutic drug monitoring (TDM). Further studies supported this concept^{8, 9} and TDM began to be applied reactively, or during loss of response, to guide interventions. There is uncertainty about the usefulness of proactive TDM, which is carried out during remission with the goal of preventing loss of response.

A few retrospective studies have shown that proactive TDM compared with reactive TDM may increase persistence on therapy¹⁰ and decrease risk of treatment failure (discontinuation of drug or need for surgery) as well as risk of hospitalization, antibody development, and infusion reaction.¹¹ As of this time, 2 prospective studies have examined proactive TDM. The Trough Level Adapted Infliximab Treatment trial (TAXIT) did not show that trough concentration-based dosing was superior to clinically based dosing, although all patients received one time trough concentration-based dose optimization prior to randomization, which did significantly increase the number of patients with CD that were in remission.¹² A randomized controlled trial investigating tailored treatment with IFX for active luminal CD (TAILORIX) assigned patients to 3 groups: 1 group with dosing based on clinical symptoms and 2 groups with dosing based on biomarker or serum IFX levels in addition to symptoms.¹³ This trial also did not meet its primary end point. Here, we conducted a retrospective study that compares patients receiving proactive and reactive TDM approaches as well as those undergoing changes in dosing based on clinical parameters only. We hypothesized that proactive TDM compared with reactive TDM or empiric changes in treatment (control) will increase rates of persistence on therapy as well as decrease healthcare utilization. Persistence on therapy was assessed at 1 and 2 years since the start of therapy, and healthcare utilization during this period was also compared.

METHODS

Design and Participants

Study participants were retrospectively identified by chart review from a single outpatient center and were at least 18 years of age, carried a diagnosis of UC or CD, and were started on either IFX or ADA at our center. Patients in whom therapy was started between 2010 and 2016 were considered for inclusion to allow for enough follow-up time to meet end points. For those who underwent trough concentration testing, only patients who obtained their levels through LabCorp (2020 Laboratory Corporation of America) were included to improve comparability. Data were collected on each participant and de-identified to be included in the database. This study was approved by the Human Subjects Research Protections Office at the University of Maryland, Baltimore.

Groups and Outcomes

Patients included in the proactive TDM all had trough concentrations checked during induction therapy and intermittently thereafter for purposes of dose optimization. For example, patients treated with IFX had a trough level assessed at week 6 and before the first maintenance dose. An example of proactive monitoring during maintenance therapy would be an assessment of ADA levels before and after the withdrawal of immunomodulator therapy. Due to the evolving understanding of IFX and ADA pharmacokinetics during the study period, there was no single definitive target trough concentration. Generally, an IFX trough of between 3 and 10 μg/mL was considered therapeutic during the period studied, and an ADA level of greater than 5 μg/ mL was similarly desired. Patients were included in the reactive TDM group if they had IFX or ADA concentrations drawn due to symptomatic relapse, evidence of ongoing inflammation, or adverse drug event. Change in dose, interval, or therapeutic agent based on the results of testing was at the provider’s discretion. Patients in the control group had no drug concentration testing; changes in therapy, if any, were made based solely on clinical evaluation.

The primary outcome was persistence on original therapy at 1 and 2 years as a marker for therapeutic success. The secondary outcome was healthcare utilization, tracked as number of hospitalizations (both related and unrelated to IBD), total length of stay (LOS) of hospitalizations, number of surgeries (related and unrelated to IBD), number of invasive nonsurgical procedures (such as endoscopy or imaging-guided procedures), number of emergency room and office visits, and number of diagnostic imaging studies. Hospitalization and surgery were considered in relation to IBD based on diagnostic code and/or presenting symptom(s). Clinical response was also assessed. Complete clinical response was defined as complete cessation of symptoms related to IBD ascertained at an office visit within 1 month of the time point, either at 1 year or at 2 years. Finally, mean doses of each anti-TNF were calculated for all groups. Doses were standardized to dose received every 2 weeks for ADA and every 4 weeks for IFX by averaging the dose received over the time interval for each subject.

Statistical Analysis

Baseline characteristics are described by mean and SD for continuous variables and by number and percentage within group for categorical variables. Comparisons of these characteristics were made using Student t-test, chi-square test for homogeneity, or Fisher exact test based on type of variable and normality of data. A P-value of 0.05 was considered significant for differences between groups. Selected covariates previously identified in the literature were examined for potential effect modification by stratified analysis, and any covariates that were different between TDM groups were assessed for confounding using a 10% change in effect cutoff for inclusion in the final multivariable model. Covariates examined included age at diagnosis and start of therapy, sex, race, IBD type (Crohn’s disease or ulcerative colitis), concomitant immunomodulation (thiopurines or methotrexate), choice of anti-TNF agent (IFX vs ADA), and smoking status. A final logistic regression was performed to examine the relationship between exposure group and persistence on therapy; similarly, Poisson regression was utilized for the relationship between exposure group and each measure of healthcare utilization except for LOS of hospitalizations, which was compared using Wilcoxon rank-sum testing.

RESULTS

Study Population

The final study population included 235 patients of which 77% had CD. Eighty-one patients (34.5%) were included in the reactive group, 82 (34.9%) in the proactive group, and 72 (30.6%) in the control group. For patients in the reactive group, levels were checked due to partial or complete loss of response (56, 69.1%), primary nonresponse (4, 4.9%), persistent endoscopic or radiographic disease activity (18, 22.2%), and medication reaction (3, 3.7%). In the control group, 32 patients (44.4%) did not have any change in therapy during the study period. Of the remaining 40 patients who did have a change in therapy, their first change in therapy was recorded. Twenty (50%) had either an increase in dose or shortening of interval, 5 (12.5%) added a second medication, 1 (2.5%) changed anti-TNF and added a second medication, 15 (32.5%) increased dose/shortened interval and added a second medication, and 1 (2.5%) removed a second medication.

Baseline characteristics of the participants are listed in Table 1. Groups were similar in these characteristics except for duration of disease, use of prednisone, budesonide, and antibiotics as well as the presence of extraintestinal manifestations and choice of anti-TNF therapy.

TABLE 1.

Demographic and Clinical Characteristics of Patients With Inflammatory Bowel Disease Treated With Infliximab or Adalimumab Between 2010 and 2016 at the University of Maryland Inflammatory Bowel Disease Program Categorized Based on Type of Drug Monitoring

Variable	Total (235)	Reactive (81)	Proactive (82)	Control (72)	P
Age at diagnosis, mean (SD)	26.1 (12.8)	26.1 (12.4)	24.7 (13.1)	27.5 (12.8)	0.42†
Duration of disease, mean (SD)	11.8 (9.9)	13.1 (11.1)	13.6 (9.9)	8.2 (7.29)	0.0002¥
Sex					0.37‡
Male	108 (46)	38 (47)	33 (40)	37 (51)
Female	127 (54)	43 (53)	49 (60)	35 (49)
Race					0.24§
Caucasian	174 (74)	61 (75)	59 (72)	54 (75)
African-American	52 (22)	18 (22)	20 (24)	14 (22)
Hispanic	2 (0.9)	0 (0)	0 (0)	2 (3)
Asian	2 (0.9)	2 (2)	0 (0)	0 (0)
Other	5 (2)	0 (0)	3 (4)	2 (3)
Smoking					0.86‡
Current	28 (12)	11 (14)	10 (12)	7(10)
Former	65 (28)	24 (30)	20 (24)	21 (29)
Never	142 (60)	46 (57)	52 (63)	44 (61)
IBD type					0.96§
Crohn disease	181 (77)	64 (79)	62 (76)	55 (76)
Ulcerative colitis	49 (21)	16 (20)	18 (22)	15 (21)
Indeterminate	5 (2)	1 (1)	2 (2)	2 (3)
Crohn disease location (n = 181)					0.62‡
Ileal	53 (29)	15 (23)	22 (35)	16 (29)
Colonic	36 (20)	14 (22)	13 (20)	9 (16)
Ileocolonic	92 (51)	34 (53)	28 (44)	30 (55)
Crohn disease phenotype (n = 181)					0.19‡
Inflammatory	55 (30)	17 (27)	18 (29)	20 (36)
Stricturing	48 (27)	23 (36)	12 (19)	13 (24)
Perforating	78 (43)	24 (38)	32 (52)	22 (40)
Medical therapy
Thiopurine use	83 (35)	30 (37)	24 (29)	29 (40)	0.33‡
Methotrexate use	24 (10)	16 (20)	6 (7)	2 (3)	0.0014‡
Prednisone use	27 (12)	8 (10)	2 (2)	17 (24)	0.0002‡
Budesonide use	13 (6)	7 (9)	2 (2)	4 (6)	0.016§
5-Aminosalicylate use	45 (19)	15 (19)	14 (17)	16 (22)	0.71‡
Antibiotic use	30 (13)	8 (10)	3 (4)	19 (26)	<0.0001‡
Extraintestinal manifestations	57 (24)	12 (15)	16 (20)	29 (40)	0.0006‡
Anti-TNF
Infliximab	162 (69)	59 (73)	68 (83)	35 (49)	<0.0001‡
Adalimumab	73 (31)	22 (27)	14 (17)	37 (51)

Values are n (%) or mean with SD. MTX, methotrexate.

^†ANOVA.

^‡Chi-square test.

^§Fisher exact test.

Persistence on Therapy

Persistence on original anti-TNF therapy at 1 and 2 years is shown in Figure 1. At 1 year, 77 (94%) of the proactive group, 58 (72%) of the reactive group, and 55 (76%) of the control group patients remained on their original anti-TNF. The odds of remaining on therapy was not different (P = 0.50) between reactive and control groups; however, the odds ratio (OR) between proactive and control groups was 4.76 (confidence interval [CI] = 1.66–13.68, P = 0.0019) and between proactive and reactive groups was 6.11 (CI = 2.19–17.03, P = 0.0002).

FIGURE 1.

A, Increased persistence on therapy at 1 y in patients receiving proactive TDM. Patients undergoing proactive TDM were more likely to be on their original therapeutic agent at 1 y when compared with the reactive TDM and control group. B, Increased persistence on therapy at 2 y in patients receiving proactive TDM. Patients undergoing proactive TDM were more likely to be on their original therapeutic agent at 2 y when compared with the reactive TDM and control group.

At 2 years, 68 (89%) of the proactive group, 49 (65%) of the reactive group, and 44 (61%) of the control group patients remained on their original therapy. Similar to rates at 1 year, the rates between reactive and control groups were not different (P = 0.596). Compared with control group, the OR of the proactive group for persisting on original therapy was 5.41 (CI = 2.26–12.94, P < 0.0001) and the reactive group was 4.51 (CI = 1.88–10.80, P = 0.0004).

Twelve patients were missing outcome data at 2 years; these patients were compared with the overall group with regard to baseline characteristics to determine whether there were any significant differences (data not shown). These patients were more likely to be current or former smokers (P = 0.009) and more likely to have UC (P = 0.0021).

Clinical response rates were also compared between groups (Table 2). Regarding clinical response 1 and 2 years after initiation of therapy, the direction and significance mirrored the persistence findings except that the control group had more patients in complete clinical response at 2 years when compared with the reactive group (P = 0.014).

TABLE 2.

Improved Clinical Response Rates in Patients Undergoing Proactive TDM

Measure	Reactive	Reactive to Proactive*	Proactive	Proactive to Control*	Control	Reactive to Control*
Response at 1 y
Number of patients	81		82		72
Complete response	28 (35%)	<0.0001	63 (77%)	0.006	40 (56%)	0.0142
Response at 2 y
Number of patients	75		76		72
Complete response	29 (39%)	<0.0001	55 (72%)	0.002	34 (47%)	0.32

Values represent number of patients in each group with complete clinical response at each time point with percentages in parentheses. P-values are shown between group columns.

*Wilcoxon rank-sum test.

Healthcare Utilization

Differences in the amount of healthcare utilization between groups are shown in Table 3. The proactive group had fewer events when compared with the reactive group in every category except for IBD-unrelated surgeries and overall LOS of IBD-unrelated hospitalizations. The proactive group also had fewer events compared with the control group in the above 2 categories as well as in the number of IBD-unrelated hospitalizations and emergency room visits. Compared with the control group, the reactive group had more IBD-unrelated hospitalizations, more IBD-related surgeries, more office visits, and more diagnostic imaging studies. There was a significant increase in dose of IFX of the proactive group when compared with the control group; however, there was no difference in IFX and ADA dose between the proactive and the reactive groups. The reactive group utilized higher mean doses than the control group of both anti-TNFs.

TABLE 3.

Decreased Healthcare Utilization in Patients Receiving Proactive TDM

Measure	Reactive (n = 75)	Proactive (n = 76)	Control (n = 72)
Hospitalizations
IBD related	33§	5.3†	42
IBD unrelated	18§	7.2‡	6.3
Length of stay (d)*
IBD related	155	20†	285
IBD unrelated	72	24	39
Surgeries
IBD related	19§	3.3†	40
IBD unrelated	9.3	3.9	3.5
Procedures	83	56†	80
ER visits	38	17‡	25
Office visits	657§	418†	528
Diagnostic imaging studies	235§	104†	249
Maximum dose reached: mean (SD)*
Infliximab (mg/kg/mo)	6.07 (2.63)§	5.51 (2.31)¥	4.58 (1.92)
Adalimumab (mg/2 wk)	68.2 (27.1) §	52.4 (18.5)	54.5 (19.2)

Values represent total number of events per group per 100 person-years with comparisons by Poisson regression except as otherwise noted. Only patients with 2 years of follow-up were included.

*Wilcoxon rank-sum test. ER, emergency room.

^†Proactive group significantly different from both reactive and control groups (P < 0.05).

^‡Proactive group significantly different from the reactive group (P < 0.05).

^¥Proactive group significantly different from the control group (P < 0.05).

^§Reactive group significantly different from the control group (P < 0.05).

Effect Modification and Confounding

Four variables were identified prior to data analysis for assessment for effect modification. These 4 variables were not found to be effect modifiers by the Breslow–Day test for homogeneity: IBD type (P = 0.23), sex (P = 0.84), extraintestinal manifestations of disease (P = 0.054), and choice of anti-TNF agent (P = 0.17). Of the covariates studied, two were found to be associated with both exposure and outcome, namely use of prednisone and budesonide. Of those, accounting for budesonide use in the model changed the estimate of the OR by greater than 10%, which we used as criteria for inclusion in the final model as confounders. However, budesonide use is likely to be colinear with our primary outcome and thus inappropriate for inclusion as a confounder.

Finally, as mentioned above, 44.4% of the patients in the control group had no change in therapy. We conducted a repeat analysis comparing the reactive group with control group excluding these patients. In terms of persistence on original anti-TNF therapy, rates were similar between these 2 groups at 1 year (P = 0.22); however, the reactive group had higher rates of persistence at 2 years (65% vs 45%, P = 0.047). We also found that the patients in the control group who required a change in therapy had significantly more IBD-related hospitalizations (P < 0.001) and surgeries related to IBD (P < 0.001). This group had less hospitalizations unrelated to IBD (P = 0.025), but a similar LOS for both IBD-related and unrelated hospitalizations (P = 0.06 and 0.17, respectively). Groups had similar rates of surgery that was unrelated to IBD (P = 0.15) and had similar numbers of diagnostic/therapeutic procedures (P = 0.97), emergency room visits (P = 0.08), and office visits (P = 0.2); however. the control group patients required more imaging (P < 0.001).

DISCUSSION

Our study showed that patients receiving proactive TDM were more likely to be persistent on their original anti-TNF therapy at both 1 and 2 years. Furthermore, less overall healthcare utilization was found in the proactive group when compared with both the reactive and control groups in almost all measures, including IBD-related hospitalizations and surgeries as well as procedures, emergency room and office visits, and diagnostic imaging. Finally, regardless of the monitoring technique, the mean doses of IFX and ADA used were similar except that when compared with controls, proactive patients treated with IFX used ~1 mg/kg more. In an average 70-kg patient, this would account for one more vial of IFX every 4 weeks. Our data suggest that optimizing the dose of either IFX or ADA using drug levels obtained during induction therapy and periodically thereafter may prevent loss of response and associated time and cost of rescue therapy.

Unfortunately, the literature on the utility of proactive TDM has been mixed. As noted above, retrospective studies have noted that proactive TDM decreases the risk of treatment failure and increases persistence on therapy,¹⁰ in line with our findings. Similarly, decreased are the risk of hospitalization, infusion reaction, and development of antidrug antibodies that can lead to loss of a therapeutic option.¹¹ To our knowledge, there have been 2 prospective studies in adults examining the role of proactive monitoring. The TAXIT trial examined clinically based versus level-based dosing of IFX in both UC and CD and did not find a difference in remission at 1 year.¹² However, all patients underwent a one-time dose optimization based on serum trough prior to randomization, aiming for a level between 3 and 7 μg/mL. Of the 263 patients assessed for this optimization, 76 (28.9%) required dose escalation and 72 (27.4%) required dose reduction. This level-based optimization probably blurs any benefit gained from proactive monitoring done post-randomization, and the numbers of patients requiring dose adjustments are not trivial. Additionally, the authors found that the number of CD patients in remission increased after this dose optimization phase; relapse rates were also lower in the proactive group. The TAILORIX trial tested level-based versus symptom-based dosing of IFX in CD patients and similarly did not find a difference in sustained steroid-free clinical remission from weeks 22 to 54 of their study.¹³ Notably, the remission rates in all study groups were lower than expected, and the rate of dose escalation in the symptom-based dosing group was significant, probably due to aggressive dose escalation based on symptoms in the control group.

The strengths of our study include sample size, inclusion of both CD and UC and both IFX and ADA, use of a single lab assay for drug levels for comparability, and 2 time points for assessment of persistence. Additionally, assessment of healthcare utilization is important as critics argue that proactive dose optimization is more costly than reactive testing. Our data conflict with that hypothesis as we found lower healthcare utilization in the proactive group without a clear increase in dose required compared with the reactive and control groups. Our study was limited by the retrospective nature of its design. Additionally, some of the patients in the control group did not require any changes in therapy. Inclusion of patients without dose escalation probably introduces some bias in favor of the control group when compared with the reactive group as all patients in the reactive group had recurrent symptoms or persistent inflammation resulting in TDM. This probably explains the lack of difference between the control and reactive groups, which is unlike prior studies. Our sensitivity analysis supports this notion as control patients who underwent changes in treatment generally had outcomes worse than those in the reactive group. Some patients in the proactive group also did not undergo any change in dosing despite TDM. Nevertheless, this alone would not account for the superiority of proactive monitoring to reactive monitoring. The study period used for identification of patient records was long in order to increase sample size; however, the supplementary resources for patients have increased during that time period at our institution. As the control patients were more likely to be from earlier in the study period whereas the proactive patients were more likely to be later in the period, the availability of support staff may confound the relationship between TDM type and healthcare utilization. We do not believe that this should affect persistence on therapy at either time point, however. Finally, the lack of serum markers of inflammation at the end of the follow-up period prevents us from comparing another important objective measure of effectiveness of proactive monitoring.

CONCLUSION

In summary, we found that proactive monitoring of IFX and ADA levels with the LabCorp assay is associated with a significant increase in persistence on treatment at 2 years and a decrease in healthcare utilization. These positive effects were achieved without a significant increase in ADA dose; however, the dose of IFX was increased 1 mg/kg every 4 weeks in the proactive group compared with controls. Our study adds to the literature supporting proactive drug monitoring of IFX and ADA in patients with IBD. Proactive TDM may also be used to decrease the need for concurrent immune suppression as we did not find that use of thiopurine or methotrexate affected our results. Providers should consider proactive TDM in management of patients with IBD treated with IFX or ADA. Additional studies are needed to address the benefit or proactive TDM with other biologic therapies.

Funding: M.T. was supported by a T32 Research Grant (DK067872-11) from the National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases.

Conflict of Interest: R.K.C. has received income for consulting and participation in advisory boards for Abbvie, LabCorp, Janssen, Pfizer, Samsung Bioepis, and Takeda.

Data Availability: The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to information contained within that could compromise the privacy of research participants.