Clinical and economic impact of pharmacist interventions to identify drug-related problems in multidisciplinary cancer care: a prospective trial

Abstract

Background

The prescription of antitumor drugs has often been associated with drug-related problems. Pretherapeutic multidisciplinary risk assessment programs including pharmaceutical care have been established to secure the initiation of injectable and oral antitumor therapies. This prospective cross-sectional double-center study evaluated the clinical and economic impact of the pharmacist in detecting drug-related problems in patients initiating antitumor therapies.

Materials and Methods

Following pharmaceutical consultations, pharmaceutical interventions were validated by a multidisciplinary team. A committee of independent clinical experts assessed the potential clinical impact of drug-drug interactions. The association of clinical variables with pharmaceutical interventions was tested using a multivariate logistic regression model. Pharmacist cost of the program was assessed by valuing pharmacists’ time at their salaries and compared with potentially avoided costs.

Results

Four hundred thirty-eight patients with solid tumors were included: 62% males, mean age of 65 ± 13 years, and average of 6 medications. Half of the patients required at least one pharmaceutical intervention and independent factors associated with pharmaceutical interventions were the number of medications (5-9 vs <5: OR = 2.91 [95% CI 1.82-4.65], P < .001) and the type of antitumor treatment (immunotherapy vs intravenous chemotherapy: OR = 0.35 [95% CI 0.18-0.68], P = .002). One hundred seventy-four out of 266 pharmaceutical interventions (130 patients) involved clinically significant drug-drug interactions. Pharmacist costs were estimated to range between €4899 and €6125. Average costs were estimated at €11.4-14.3 per patient. Avoided hospitalization costs were estimated to be €180 633.

Conclusion

Clinical pharmacists contribute to the cost-effective reduction of drug-related problems in pre-therapeutic assessment programs for patients with cancer.

Pre-therapeutic multidisciplinary risk assessment programs including pharmaceutical care have been established to secure the initiation of injectable and oral antitumor therapies. This study evaluated the clinical and economic impact of the pharmacist in detecting drug-related problems in patients initiating antitumor therapies.

Graphical Abstract

Graphical Abstract.

Implications for practice.

Clinical pharmacists contribute to cost-effective reduction of drug-related problems in pre-therapeutic assessment programs for patients with cancer before initiation of oral and injectable treatment. In this study, half of the patients initiating an antitumor therapy required a pharmaceutical intervention and 65% of these interventions were due to clinically significant drug-drug interactions. The estimated costs per patient ranged from €11.4 to €14.3 and the avoided hospitalization costs were €180 633. Further development of multidisciplinary risk assessment programs with ongoing pharmacist involvement is needed to address the increasing therapeutic complexity of patients with cancer.

Introduction

Polymedication in oncology has been a widely recognized international problem.^1,2 Multiple studies have shown a high rate of drug-related problems (DRP) in patients with cancer,^3-5 defined as “an event or circumstance involving drug therapy that actually or potentially interferes with desired health outcomes.”⁶ Among these DRPs, drug-drug interactions (DDI) occur when the effects of one drug are changed by the presence of another drug, herbal medicine, food, or drink.⁷ DDIs harm around 5 million inpatients per year and cause up to 220 000 emergency department visits per year.⁸ Cancer drugs are indeed associated with major safety issues due to their narrow therapeutic index and their toxicities. Cancer patients are at high risk of DRPs due to their advanced age, comorbidities, renal impairment, and polymedication. Furthermore, cancer patients frequently use over-the-counter drugs or other complementary and alternative medicines (CAM)—but do not discuss commonly about it with their oncologist (with approximately 77% nondisclosure).^9-11 Therefore, DRPs are frequent in patients with cancer and may be associated with negative clinical outcomes: fatal adverse drug events have been observed in 4% of patients with cancer.¹² Clinical pharmacists may contribute to the prevention of DRPs in patients with cancer.^13,14

To best tailor the prescribed antitumor treatment to the individual patient with cancer, some oncology departments have set up multidisciplinary pre-therapeutic evaluation programs.^14-16 The main care providers involved in these programs are oncologists, nurses, dietitians, and pharmacists. Psychologists, social workers, geriatricians, cardiologists, and diabetologists can also participate in the risk assessment of the patient. Our study included 2 French university hospitals that organize this type of pre-therapeutic risk assessment in cancer day hospitals. The strength of these programs is the collegial therapeutic optimization. During this weekly multidisciplinary meeting: oncologists, nurses, dieticians, and pharmacists discuss the specific risks involved with antitumor treatment initiation such as the management of toxicities and DDIs, the complex medication schedule for patients with cognitive deficits or large tablets in elderly patients with dysphagia. The multidisciplinary team suggest changes in the management of each patient to reduce the risk of DRPs before the antitumor treatment initiation.

These evaluation programs aim to optimize antitumor treatment and ensure a good quality of life for patients with cancer. Currently, active pre-therapeutic risk assessment programs mainly concern oral therapies and do not focus on injectable antitumor drugs. Indeed, a single pharmaceutical validation is generally performed before the reconstitution of injectable antitumor drugs but rarely includes patient consultations. Despite an increase in the number of clinical pharmacists working in healthcare services,¹⁷ no literature describes the clinical impact of pharmacists integrated as permanent care providers in these multidisciplinary risk assessment programs.¹³ In addition, in the context of limited resource allocation, economic evaluations are essential to study the benefit of clinical pharmacists and the means to prevent DRPs. Little is known about the cost-effectiveness of these programs, mainly due to the limited ability to randomize a routinely performed activity. Our aim was to contribute to the understanding of the potential cost savings resulting from clinical pharmacist interventions by studying a large cohort of patients from 2 hospitals.

Our study therefore aimed to assess the clinical and economic impact of the pharmacist—in a multidisciplinary risk assessment program—on the detection of DRPs in patients initiating injectable and oral antitumor therapies.

Materials and methods

Results of the study are reported according to the STROBE guidelines.

Pharmacist involvement in the multidisciplinary pretherapeutic evaluation program

Following the consultation with the oncologist, the patient was offered a half-day appointment at the hospital (Figure 1), for oral or injectable antitumor agents (Cochin hospital) or oral antitumor therapies (“hôpital européen Georges-Pompidou” [HEGP]). Clinical pharmacists visited all patients who agreed to participate in the pretherapeutic evaluation during a day hospital. The pharmaceutical care intervention involved a medication reconciliation (including self-medication and CAM).¹⁸ The pharmacist identified DRPs and draft pharmaceutical interventions (PIs), defined as “any pharmacist action that directly results in a change in patient management or therapy.”^19,20

Figure 1.

Multidisciplinary pre-therapeutic assessments for the initiation of antitumor therapy in cancer outpatients and pharmacist’s involvement. (A) Panel shows the organization of multidisciplinary pre-therapeutic assessments for the initiation of antitumor therapy in cancer outpatients. The pharmacist’s involvement is shown in gray: during consultations in the day hospital and at the multidisciplinary meeting for pre-therapeutic optimization. (B) Panel shows the several steps of the pharmaceutical consultation and analysis.

Patients

Patients were included in this observational study between February 2020 and April 2021 during the pharmaceutical consultations of the multidisciplinary pre-therapeutic evaluation program. Eligible patients were at least 18 years old and treated for solid tumors. Both day hospitals included patients treated with oral-targeted therapies. Cochin Hospital also enrolled patients before the initiation of any i.v. antitumor drugs (ie, chemotherapies and immunotherapies).

Study design, outcomes, and ethical considerations

This French prospective observational cross-sectional double-center study was launched in February 2020. The study protocol was approved by the Committee for the Protection of Persons “CPP Ouest 6” on July 18, 2019 (Approval #PP-1170-RIPH3, IDRCB number: #2018-A03512-53). This study was in accordance with the Declaration of Helsinki. All eligible patients agreed to participate in the study and signed a non-opposition consent form.

The primary outcome measures were the number, the type, and the clinical impact of PIs. The secondary endpoint was the economic evaluation of the impact of PIs on the detection of DDIs.

Data collection and pharmaceutical care

At the end of the multidisciplinary evaluation in the cancer day hospital, each investigator completed a report in the patient’s electronic health record system (DxCare , Dedalus, France for HEGP and Orbis Dedalus, France for Cochin hospital). Clinical, demographic, and biological data were collected at baseline: age, gender, body-mass index (BMI), metastatic status, Eastern Cooperative Oncology Group (ECOG) Performance status, Charlson Comorbidity Index Score,²¹ number of regular medications, the tumor type and the type of antitumor treatment. The collected biological data included albumine, creatinin, cystatin, C-reactive protein (CRP), alanine transaminase (ALT), aspartate transaminase (AST), alkaline phosphatase (ALP), and bilirubin. Dieticians at Cochin Hospital determined the patients’ nutritional status.

Following the guidelines of the French Society of Clinical Pharmacy,²² pharmaceutical care included (i) a medication reconciliation, (ii) information on correct drug use and potential side effects and (iii) community–hospital coordination if necessary (Figure 1). DRPs—including DDIs—were assessed by the pharmacist. Consultations were timed.

Pharmaceutical analysis

DRPs were categorized according to the PCNE classification (PCNE Classification Scheme for Drug-Related Problems V9.1).⁶ Drugs involved in DRPs were classified according to the WHO’s Anatomical Therapeutic Chemical (ATC) classification. DDIs were assessed according to several DDI software databases: Micromedex (www.micromedexsolutions.com/home/dispatch), Thériaque (www.theriaque.org/, a drug database certified by the French National Agency for the Safety of Medicines and Health Products), and DDI-predictor (www.ddi-predictor.org) in case of suspected interactions with a cytochrome P450 (CYP) inhibitor or inducer. For interactions with CAM, we consulted Memorial Sloan Kettering Cancer Center “About Herbs” (www.aboutherbs.com) and Hedrine (www.theriaque.org/). For dosage adjustments related to renal insufficiency, “GPR” (sitegpr.com/fr/) was consulted. Pharmaceutical analyses were timed.

Following the analysis of these data, PIs were presented to several care providers during a weekly multidisciplinary staff, leading to potential treatment optimization.

DRPs and PIs were classified according to the French Society of Clinical Pharmacy.¹⁹

Independent evaluation of drug-drug interactions

To evaluate the potential clinical impact of these PIs, an independent clinical expert committee was formed, consisting of 2 medical oncologists, 2 oncology clinical pharmacists and one pharmacologist. PIs previously validated by local medical staff were presented to the committee. For each avoided DDI, these experts classified the potential clinical impact and the level of evidence using previously validated scales.^23,24 This panel of experts graded the level of severity of the potential clinical consequences from minor (minor injury or illness requiring minor intervention) to major (major injury leading to long term incapacity/disability). Finally, the experts classified the level of evidence from very low to high. Two of them independently evaluated each DDI. If they disagreed, a third independent expert was consulted to ensure that a consensus was reached.

Medico-economic evaluation

The economic analysis was conducted from the hospital’s perspective, as one of the possible hurdles to implementing these multidisciplinary programs is optimizing the allocation of human resources. The valuation year used was 2022. Costs were not discounted, given the duration of the study (14 months).

First, we calculated the pharmaceutical time based on consultations and analysis times. The time spent has been valued (i) to an average annual full-time equivalent based on a maximum working week of 48 hours and (ii) by the grade of the contributor. Two scenarios have been established with low- and high-salary grades. The cost of the 14 months’ study time has been reduced to 12 months for estimates applied on an annual basis.

Second, we selected PIs regarding clinically significant DDIs related to drug toxicity, as measuring the consequences on survival of avoided DDIs seemed more hazardous. We sought to estimate the avoided cost based on the avoided clinical consequences of the most severe levels (moderate and major levels) that could have caused hospitalization or an emergency room (ER) stay. We valued the likely “diagnosis-related groups” (DRG) of the avoided event thanks to the 2019 national survey on hospital costs.²⁵ In the base-case, the moderate and major levels assessed by the panel were linked to DRG levels 2 and 3 (on a 4-level scale). The average overall cost of ER has been estimated at €227 (according to a parliamentary report²⁶). Costs were weighted by the probability of occurrence of the adverse drug event in the absence of PIs based on the level of evidence of the DDIs avoided found in the literature. Using the methodology described by Nesbit et al²⁷ and applied recently,^28,29 each avoided clinical consequence was assigned to a corresponding probability of occurrence category of 0.01, 0.1, 0.4, 0.6, ie, a very low, low, moderate and high level of evidence, respectively. The additional or avoided costs induced by the interventions in themselves or forming part of basic clinical monitoring are deemed marginal and have not been valued.

Finally, to explore the robustness of our results, we conducted scenario-based analysis and sensitive analysis by varying the cost-driver parameters of our base-case (number of outpatients, grade of the contributors, DRG’s severity levels, ± 20% variation for the mean avoided cost).

Statistical analysis

Descriptive statistics used mean and standard deviation (SD) or median and interquartile range (IQR) for quantitative variables, and numbers and percentages for qualitative variables. Normality distribution assumption was tested for each quantitative variable. Variables not normally distributed were dichotomized according to clinically relevant cutoffs or on the median. Groups were compared using Student’s t test or ANOVA for quantitative variables, and with Chi-square test for qualitative variables. Logistic regression models were used to test the association of clinical and biological variables—collected to assess the risks associated with the introduction of antitumor therapy—with PIs. Variables significantly associated with PIs in univariate models were then combined into the multivariable logistic regression model.

All P-values were 2-sided, and the level of significance was set at P < 0.05. Statistical analyses were performed with R statistical software (version 4.2.2).

Results

Description of the population

Over the 14-month period, 438 patients with solid tumors were included in the study (Figure 2, A panel). The demographic profile showed a majority of males (62%) and a mean age of 64.8 (13.3) years. Their characteristics are described in Table 1. The average number of regular medications was 6 (IQR: 3-8) per patient.

Figure 2.

Flowchart and pharmaceutical interventions in the cohort. (A) Panel shows the flowchart of the study. (B) Panel shows the classification of pharmaceutical interventions. The potential clinical impact of drug-drug interactions was evaluated by a multidisciplinary committee of independent clinical experts. They graded the level of severity of the potential clinical consequences that could have occurred from minor (minor injury or illness requiring minor intervention) to major (major injury leading to long term incapacity/disability) and the experts classified the level of evidence from very low to high (PI = Pharmaceutical Interventions).

Table 1.

Baseline characteristics ofthe population.

Variable	Value	Number of evaluable data
Age, years old: mean (SD) or n (%)	64.8 (13.3)	438
Low < 75 years old	112 (26%)
High ≥ 75 years old	326 (74%)
Gender, n (%)		438
Male	272 (62%)
Female	166 (38%)
BMI, kg.m-²: mean (SD) or n (%)	25.7 (4.6)	437
< 18	10 (2%)
18-25	199 (46%)
> 25	228 (52%)
Charlson Comorbidity Index Score: mean (SD) or n (%)	4 (3)	429
< 3	148 (34%)
≥ 3	281 (66%)
ECOG Performance Status, n (%)		416
≥ 2	151 (36%)
0-1	265 (64%)
Undernutrition, n (%)		376
Not undernourished	138 (37%)
Moderate undernutrition	74 (20%)
Severe undernutrition	48 (13%)
At risk of undernutrition	116 (31%)
Metastasis, n (%)		435
No	107 (25%)
Yes	328 (75%)
Antitumor agents, n (%)		438
Immunochemotherapy	37 (8%)
Immunotherapy	58 (13%)
Intravenous chemotherapy	204 (47%)
Oral therapy	139 (32%)
Line of treatment: mean (SD)	1.6 (1.0)	434
Tumor type, n (%)		438
Urological	148 (34%)
Thoracic	100 (23%)
Sarcoma	89 (20%)
Digestive tract	47 (11%)
Gynecological	30 (7%)
Breast	12 (3%)
Other	12 (3%)
Baseline biological data: mean (SD)
ALT (UI L−1)	33.8 (45.3)	420
AST (UI L−1)	31.6 (28.2)	419
Bilirubin (μmol L−1)	8.4 (33.2)	409
ALP (UI L−1)	122.6 (223.2)	416
Albumin (g L−1)	39.4 (5.7)	405
Serum creatinine (μmol L−1)	85.3 (31.6)	427
Serum cystatin (mg L−1)	1.15 (0.3)	361
C-reactive protein (mg L−1)	18.1 (24.6)	400
Number of regular medications: mean (SD) or n (%)	6 (4)	434
<5	180 (41%)
5-9	174 (40%)
> 10	80 (18%)
Self-medication, n (%)		436
No	324 (74%)
Yes	112 (26%)

Abbreviations: ALT, alanine amino transferase; AST, aspartate amino transferase; ALP, alkaline phosphatase.

Patients were mainly treated for urothelial cancer (34%), thoracic cancer (23%) then sarcoma (20%). Most patients (60%) received a first line of treatment. Half of the patients were evaluated before intravenous chemotherapy and a third of them before oral antitumor therapy (Table 1).

Pharmaceutical interventions

Half of the patients in our study (51%, n = 223) required at least one PI, ie, 374 interventions were recorded over the period (Figure 2, B panel). These patients had an average of 1.7 PIs (IQR: 1-2).

These PIs involved 578 drugs and concerned mostly alimentary tract and metabolism drugs (32%, n = 184), antineoplastic and immunomodulator agents (29%, n = 169), and cardiovascular system drugs (19%, n = 108; Supplementary Table S1). The most common drugs involved in DDIs were proton pump inhibitors (48/578), aprepitant (43/578), and racecadotril (36/578).

The main proposed therapeutic optimizations were: biological or clinical monitoring (33%) then pharmacological switch (26%) or drug discontinuation (25%) (Supplementary Figure S1).

In multivariable models, the type of antitumoral treatment (immunotherapy vs intravenous chemotherapy: OR = 0.35 [95% CI 0.18-0.68], P = 0.002) and the number of medications (5-9 vs <5: OR = 2.91 [95% CI 1.82-4.65], P < 0.001; ≥10 vs <5: OR = 4.29 [95% CI 2.24-8.21], P < 0.001) were identified as independent risk factors of PIs (Table 2).

Table 2.

Uni- and multivariable logistic regression models for clinical factors associated with pharmaceutical interventions.

Variable	No pharmaceutical interventions	Pharmaceutical interventions	Univariate model: OR (95% CI)	Univariate model: P-value	Multivariate model: OR (95% CI)	Multivariate model: P-value
Age, years old: mean (SD)	63.2 (13.3)	66.3 (13.3)
Low < 75 years old	171 (80%)	164 (74%)	1
High ≥ 75 years old	44 (20%)	59 (26%)	1.40 (0.90-2.18)	.14
Gender
Male	132 (61%)	140 (63%)	1
Female	83 (39%)	83 (37%)	0.94 (0.64-1.39)	.77
BMI, kg/m²: mean (SD)	25.6 (4.7)	25.9 (4.5)
<18	4 (2%)	6 (3%)	1.58 (0.43-5.76)	.49
18-25	102 (47%)	97 (44%)	1
>25	109 (51%)	119 (54%)	1.15 (0.79-1.68)	.48
Charlson Comorbidity Index Score: mean (SD)	3.7 (3.1)	4.2 (2.8)
< 3	87 (42%)	61 (28%)	1
≥3	121 (66%)	160 (78%)	1.89 (1.26-2.82)	.002	1.57 (0.89-2.77)	.12
ECOG Performance Status
0-1	138 (69%)	127 (59%)	1
≥2	61 (31%)	90 (41%)	1.60 (1.07-2.40)	.022	0.99 (0.62-1.60)	.98
Undernutrition
No	126 (69%)	128 (66%)	1
Yes	56 (31%)	66 (34%)	1.16 (0.75-1.79)	.50
Metastasis
No	55 (26%)	52 (23%)	1
Yes	158 (74%)	170 (77%)	1.14 (0.74-1.76)	.56
Type of antitumor treatment
Intravenous chemotherapy	99 (46%)	105 (47%)	1
Immunochemotherapy	12 (6%)	25 (11%)	1.96 (0.94-4.12)	.074	2.18 (0.97-4.87)	.06
Immunotherapy	40 (19%)	18 (8%)	0.42 (0.23-0.79)	.007	0.35 (0.18-0.68)	.002
Oral therapy	64 (30%)	75 (34%)	1.11 (0.72-1.70)	.65	1.35 (0.81-2.22)	.25
Tumor type				.92
Breast	6 (3%)	6 (3%)	1.11 (0.34-3.61)	.86
Digestive tract	22 (10%)	25 (11%)	1.27 (0.82-2.27)	.24
Gynaecological	14 (7%)	16 (7%)	1.27 (0.58-2.80)	.55
Other	5 (2%)	7 (3%)	1.56 (0.47-5.14)	.47
Sarcoma	45 (21%)	44 (20%)	1.09 (0.64-1.84)	.75
Thoracic	45 (21%)	55 (25%)	1.36 (0.34-3.61)	.86
Urological	78 (36%)	70 (31%)	1
Number of regular medications	4.9 (3.5)	7.2 (4.1)
<5	117 (55%)	63 (28%)	1
5-9	72 (34%)	102 (46%)	2.63 (1.71-4.04)	<10e-04	2.91 (1.82-4.65)	<10e-04
≥10	22 (10%)	58 (26%)	4.90 (2.75-8.73)	<10e-07	4.29 (2.24-8.21)	<10e-04
Self-medication
No	157 (74%)	167 (75%)	1
Yes	56 (26%)	56 (25%)	0.94 (0.61-1.45)	.78
Use of complementary and alternative medicines
No	148 (69%)	147 (66%)	1
Yes	66 (31%)	76 (34%)	1.16 (0.78-1.73)	.47

Abbreviations: BMI, body mass index; ECOG, Eastern Cooperative Oncology Group; OR, odds ratio.

PIs were mainly due to 3 causes (Figure 2, B panel): DDI (n = 266, 71%), other DRP (n = 62, 17%), and interactions with CAM (n = 46, 12%).

From the 62 PIs classified in “other DRP,” the main problem retrieved was supratherapeutic dosage (n = 26, 42%) due to poor adaptation of treatments to renal function (Supplementary Figure S2).

Of the patients in our cohort, 112 (26%) were self-medication users. In addition, 32% of them (n = 141) used complementary and alternative medicines (CAM) including herbs infusion and capsules (53%), vitamins and minerals (30%), homeopathy (8%), aromatherapy (4%), and other CAM (5%). We found that 46/374 PIs concerned CAM that may lead to DDI (12%): due to an antioxidant effect during chemotherapy or due to pharmacokinetic interactions (mainly by inhibition of CYP 3A4). The top 3 possibly interacting CAM used by patients were vitamin C (16/59), green tea (8/59), and turmeric (7/59). 87% of the PIs recommended a discontinuation of the CAM.

Clinical impact evaluation of drug-drug interactions

The expert committee reviewed 266 PIs that concerned DDIs. Their clinical impact was mostly related to drug toxicity (79%, 211/266) but some could impact patients’ survival (21%, 54/266; Figure 2, B panel). 65% of these PIs (174/266) were considered clinically significant (major or moderate clinical severity) and concerned 130/175 patients (74%). However, the level of evidence was mostly weak: low for 39% (104/266) and very low for 24% (65/266) of the PIs.

The main DDIs potentially impacting toxicity are presented in Table 3 (exhaustive data in Supplementary Table S2). The most frequent DDIs were increased risks of (i) antitumor overdose due to drugs affecting their renal elimination (non-steroidal anti-inflammatory drug for example) or pharmacokinetic interactions inhibiting enzymatic metabolism, (ii) morphine overdose due to pharmacokinetic interactions inhibiting enzymatic metabolism that increased the risk of respiratory depression, and (iii) angioedema due to the concomitant use of racecadotril (premedication in case of diarrhea under chemotherapy) and angiotensin-converting enzyme inhibitors. PIs were adapted according to the drugs involved and the patient’s clinical and biological situation.

Table 3.

Frequent clinically significant drug-drug interactions impacting toxicity.

PK/PD	Potential clinical effect	Potential drug-drug interaction	Severity	Evidence	Intervention
PK	Aplasia due to chemotherapy overdose	Pemetrexed & PPI (n = 7)	Moderate	Moderate	PPI discontinuation or switch
		Pemetrexed and NSAID (n = 3)	Major	Moderate	NSAID discontinuation or switch
		Methotrexate and PPI or NSAID (n = 3)	Major	Low	PPI discontinuation
	Respiratory depression due to morphine overdose	Aprepitant and oxycodone (n = 8)	Major	Moderate	Aprepitant discontinuation or reinforced monitoring of oxycodone toxicity
		Aprepitant and fentanyl (n = 3)	Moderate	Low	Aprepitant discontinuation or reinforced monitoring of fentanyl toxicity
	Emergency room consultation due to high-grade TKI toxicity	Amlodipine and TKI: axitinib, dabrafenib, sunitinib (n = 6)	Moderate	Low	TKI therapeutic drug monitoring
		Axitinib and fluconazole (n = 3)	Moderate	Low	Substitution by a amphotericin B oral suspension
		Aprepitant and nilotinib (n = 2)	Moderate	Low	Nilotinib therapeutic drug monitoring
		Aprepitant and erlotinib (n = 1)	Major	Moderate	Substitution with osimertinib
	Anticoagulant side effects: risk of bleeding	Apixaban and several drugs (n = 5)	Moderate	Very low	Three substitutions by low-molecular-weight heparin. Two clinical and therapeutic drug monitoring.
PD	Angioedema	Racecadotril and angiotensin-converting enzyme (ACE) inhibitor or angiotensin II receptor antagonists (n = 36)	Major	Moderate	Substitution of racecadotril with loperamide
	QT prolongation: Torsade de pointe	Ondansetron and several drugs (n = 5)	Moderate	Very low	Electrocardiogram monitoring

We first presented pharmacokinetic (PK) drug-drug interactions then pharmacodynamic (PD) drug-drug interactions. The exhaustive table is in Supplementary Table S2.

Abbreviations: PPI, proton pump inhibitor; NSAID, non-steroidal anti-inflammatory drug; TKI, tyrosine kinase inhibitor.

In multivariable models, the type of antitumor treatment (oral therapy vs intravenous chemotherapy: OR = 0.31 [95%CI 0.14-0.70], P = 0.004) was the only independent risk factor of moderate or major DDI (Table 4).

Table 4.

Uni- and multivariable logistic regression models for the clinical severity among the 175 patients with DDIs.

Variable	Patients with minor DDI	Patients with moderate or major DDI	Univariate model: OR (95% CI)	Univariate model: P-value	Multivariate model: OR (95% CI)	Multivariate model: P-value
Age
Low < 75 years old	35 (78%)	93 (72%)	1
High ≥ 75 years old	10 (22%)	37 (28%)	1.39 (0.63-3.10)	.42
Gender
Male	28 (62%)	83 (64%)	1
Female	17 (38%)	47 (36%)	0.93 (0.46-1.88)	.85
BMI				.14
<18	1 (2%)	4 (3%)	2.04 (0.22-19.30)	.53
18-25	24 (53%)	47 (36%)	1
>25	20 (44%)	78 (60%)	1.99 (0.99-3.99)	.052
Charlson Comorbidity Index Score
<3	9 (21%)	29 (22%)
≥3	34 (79%)	101 (78%)	0.92 (0.40-2.14)	.85
ECOG Performance Status
0-1	24 (57%)	70 (55%)	1
>2	18 (43%)	57 (45%)	1.09 (0.54-2.20)	.82
Undernutrition
No	20 (65%)	80 (69%)	1
Yes	11 (35%)	36 (31%)	0.818 (0.355-1.884)	.64
Metastasis
No	9 (20%)	26 (20%)	1
Yes	36 (80%)	103 (80%)	0.99 (0.42-2.31)	.98
Type of antitumor treatment
Intravenous chemotherapy	13 (29%)	62 (48%)	1
Immunochemotherapy	5 (11%)	16 (12%)	0.67 (0.21-2.16)	.50	0.67 (0.20-2.20)	.51
Immunotherapy	1 (2%)	15 (12%)	3.14 (0.38-25.96)	.29	3.74 (0.44-31.53)	.22
Oral therapy	26 (58%)	37 (28%)	0.30 (0.13-0.65)	.002	0.31 (0.14-0.70)	.004
Tumor type
Breast	1 (2%)	4 (3%)	1.8 (0.19-17.26)	.61
Digestive tract	4 (9%)	16 (12%)	1.8 (0.53-6.15)	.34
Gynecological	4 (9%)	7 (5%)	0.79 (0.20-3.03)	.73
Other	1 (2%)	6 (5%)	2.7 (0.30-24.10)	.37
Sarcoma	7 (16%)	21 (16%)	1.35 (0.49-3.75)	.56
Thoracic	10 (22%)	36 (28%)	1.62 (0.66-3.96)	.29
Urological	18 (40%)	40 (31%)	1
Number of regular medications
<5	14 (31%)	26 (20%)	1
5-9	22 (49%)	60 (46%)	1.47 (0.65-3.31)	.35	1.82 (0.76-4.34)	.18
≥10	9 (20%)	44 (34%)	2.63 (1.00-6.93)	.0499	2.6 (0.94-7.17)	.07
Self-medication
No	30 (67%)	99 (76%)	1
Yes	15 (33%)	31 (24%)	0.63 (0.30-1.31)	.22
Use of complementary and alternative medicines
No	32 (71%)	92 (71%)	1
Yes	13 (29%)	38 (29%)	1.02 (0.48-2.15)	.97

Abbreviation: DDI, drug-drug interaction.

Pharmaceutical time and costs

Pharmacist time was 283 hours over the 14-month study period, including 438 patients. The estimated annual full-time equivalent was 0.13. Per patient, the average time is 39.2 (15.4) minutes: 22.8 (7.36) minutes of pharmaceutical consultation, and 16.4 (11.0) minutes of pharmaceutical analysis. The number of regular medications and the type of antitumor treatment seemed to significantly influence pharmaceutical time, whether it be consultation time (Supplementary Table S3) or analysis time (Supplementary Table S4).

The total cost for the pharmaceutical time was estimated between €4899 (low salaries) and €6125 (high salaries), which amounts to €4199 (low salaries) and €5250 (high salaries) per year. The cost was estimated between €11.4 and €14.3 per patient and between €18.42 and €23.02 per DDI on average. The base-case was calculated on an annual basis according to the total pharmaceutical time spent and the different grades of the contributors. Analysis in scenarios (Figure 3) showed the different cost ranges and how they differ from the base-case in terms of internal organization (involvement of a senior pharmacist or a junior pharmacist) and the number of patients included.

Figure 3.

Scenario-based analysis on the pharmacist cost in this multidisciplinary pre-therapeutic evaluation program (in euros). The gray and black bars represent the estimated cost of the program according to the level of the salary grid (two scenarios: high and low). The left part of the graph lists 2 various conditions: the grade of the pharmacist and the number of patients.

Economic impact of the pharmaceutical interventions about drug-drug interactions that could have caused a hospitalization

174/266 PIs related to toxicity or survival risk were evaluated to be clinically significant DDIs (major or moderate clinical severity). We focused on the 122 DDIs related to drug toxicity that could have caused a hospitalization or emergency room (ER) stay. The various events avoided could have led to different types of hospitalization (DRG) or ER (17/122) stays, which, depending on the level of severity, represent a mean cost of €4869 (apart from ER costs, hospitalization costs are fairly homogeneous with a median cost of €4764; min: €2649; max: €10 807) (Supplementary Figure S3). By multiplying each avoided event by its probability of occurrence category, total avoided hospitalization costs were estimated at €180 633. By subtracting the total cost for the pharmaceutical time, the estimated total net benefit could reach around €175 000. In other words, for €1 invested, between €29 and €37 could be avoided. The sensitivity analysis—that considers the mean avoided cost ± 20% (whatever the event or the severity)—showed that total avoided hospitalization costs could reach between €124 690 and €187 035. In an exploratory analysis, when grading DRG’s severity by one level up (eg, levels 3 and 4), the cost of managing an adverse event reached €8816, and total avoided hospitalization costs were estimated at €334 184.

Discussion

This prospective double-center observational study suggests that clinical pharmacists, integrated into pre-therapeutic multidisciplinary risk assessment programs, may reduce the risk of DRPs in outpatients with cancer. Thus, half of the patients in our study required at least one PI. Furthermore, according to an independent committee of experts, 30% of patients with cancer experienced a clinically significant DDI. Pre-therapeutic multidisciplinary assessment programs for patients with cancer to discuss complex clinical situations are essential in this context. This program should include a clinical pharmacist and may result in substantial cost savings (for €1 invested, between €29 and €37 could be avoided).

The deployment of clinical pharmacists has increased over the years.³⁰ Limited pharmacist resources may require prioritization of patients requiring medication reconciliation. Clinical and demographic factors associated with PIs were evaluated in a multivariate analysis. First, polymedication was associated with PIs rendering, as described in several studies.^4,31 In our study, approximately 58% of patients with cancer were taking at least 5 regular medications, compared to 85% in a previous study.² This difference could be explained by the study setting: day hospital in our study, compared to palliative care, general oncology wards or outpatient clinics in Kotlinska-Lemieszek et al’s study. The second factor associated with PIs was the type of antitumor therapy, identified for the first time to our knowledge. The number of PIs was significantly higher for patients treated with immunochemotherapy then intravenous chemotherapy or oral treatment then immunotherapy. When intravenous chemotherapy was chosen as the standard, the odds ratio for PIs was 2.18 (0.97-4.87) for immuno-chemotherapy, 0.35 (0.18-0.68) for immunotherapy and 1.3 (0.81-2.22) for oral therapy. This result first showed that there were fewer pharmacist interventions for immunotherapy, compared to chemotherapy. This may be explained by the fact that immunotherapies do not have pharmacokinetic interactions. With immune checkpoint inhibitors for example, only pharmacodynamic interactions with antibiotics or corticoids have been described.^32,33 Second, there were more pharmacist interventions with immuno-chemotherapy combinations than chemotherapy alone. This finding was directly related to the number of antitumoral treatments, which increases when immunotherapy and chemotherapy are combined, and is likely to cause more DDIs than chemotherapy alone.

According to the independent experts committee, one in 4 PIs implicated a major clinical impact in our cohort. Statistical analysis revealed that there were more clinically significant DDIs in the case of intravenous chemotherapy than with oral therapy (P = .004). This result may be biased since (i) only one of the 2 centers included injectable antitumor treatments in the pre-therapeutic assessment, and (ii) intravenous treatments were frequently polychemotherapy and were combined with several supportive treatments. Similarly, Walsh DJ. et al have shown that monochemotherapy was associated with unplanned hospitalization due to adverse drug events, but to a lesser extent degree than polychemotherapy.³⁴ However, the clinical impact of DDIs with injectable antitumor drugs seems to be underestimated in clinical practice. To optimize oral antitumor treatment, pharmaceutic consultations have been set up^15,35,36 and have proven to be effective in enhancing patient safety.^15,37 However, this study has shown that the risks of significant clinical impact seem to be at least as high in the case of DDIs with oral therapies as with intravenous treatments. Some hospitals have tried a multidisciplinary approach for elderly patients receiving their first chemotherapy, without systematic collegial therapeutic validation.^38,39 For example, Herledan et al found that prescribers did not follow approximately 30% of the pharmacist’s recommendations.²⁹ These findings highlight the importance of integrating pharmacists into a multidisciplinary team.

The pharmacist can reduce patient risk, through case-by-case analysis, especially in situations where recommendations are non-existent. The independent evaluating committee identified a major issue with the level of scientific evidence, which was frequently considered as low (39%) or very low (24%) for the DDI studied. The clinical relevance of the literature data is limited due to the preponderance of preclinical studies, a scarcity of controlled clinical studies, and case reports.⁸ This presents a challenge for pharmacists,⁴⁰ who must cross-reference multiple sources, a time-consuming process. Variability in DDI alerting performance across online databases and pharmacy software systems has been demonstrated.⁴¹ This highlights the importance of multidisciplinary assessments and the subjectivity of decision making.

The use of CAM is a regularly observed practice, in approximately 40% of cancer patients,⁹ to boost their immune defences, reduce adverse events or increase their chances of recovery.⁴² In the present study, 32% of cancer patients used CAM. A comprehensive American study on 3118 patients with cancer reported the use of CAM in 1023 (33.3%) participants, including 288 (29.3%) patients who did not disclose it to their physician.⁴³ A multidisciplinary approach allows the detection of CAM use, which may expose patients to risks of pharmacokinetic or pharmacodynamic interactions. In fact, 12% of PIs were related to drug interactions with CAM. The 3 most commonly used CAM by patients were vitamin C, green tea, and turmeric. Therefore, a pharmacodynamic interaction was detected between these antioxidants and cancer chemotherapy, since antioxidants may protect tumor cells from the killing effects of chemotherapy, as suggested by Lawenda et al.⁴⁴ In addition, green tea and turmeric have also been reported to have pharmacokinetic inhibitory properties although in vivo / in vitro data are sometimes conflicting.^45–49 The risk of combined pharmacodynamic and pharmacokinetic interactions with these CAMs often lead the pharmacist to discontinue them. A guideline recently offered directions on how to support evidence-informed decision-making about the use of CAM among patients with cancer.¹¹

Our study confirms the cost-effectiveness of integrating clinical pharmacists in pre-therapeutic multidisciplinary risk assessment programs. However, as in previously published studies,^{27–29,35,50} the valuation of avoided events must be treated with caution because of all the assumptions made (how dedicated pharmacist time was accounted for, level of evidence as a proxy of the probability of occurrence, category of the probability of occurrence used, and choice of the possible DRG and its level of severity). We did not work on a full-cost basis, considering that certain follow-up costs were negligible, and this was confirmed by the results, as therapeutic drug monitoring was performed in only 33 patients. Uniquely, this study accurately measures pharmacy time, differentiating between analysis and consultation time, and evaluates events avoided based on French hospital production costs. The time spent is reflected in the comparison, rather than an annual salary, making it difficult to compare. The results on cost avoidance are robust to sensitivity analysis, despite the strong calculation assumptions. Furthermore, we have been conservative in our assessment of avoided costs, including only the potentially most severe events. Managing adverse drug events can be time-consuming without generating coded activity for the hospital. On the opposite, revenues received by the hospital in return for hospitalizations and consultations could have been taken into account, but we wanted to stay as close as possible to production costs, as pricing grids are far removed from them. Finally, while the efficient allocation of human resources is crucial from a hospital’s point of view, the potential seriousness of the events avoided and their frequency justify the resources devoted to them, from both the payer’s and society’s point of view. The extension of integrated clinical pharmacists also enables the development of shared analysis tools, guidelines, and recommendations, as demonstrated by the collaboration between the 2 centers in this study, thus improving the quality and effectiveness of these highly time-sensitive programs.

Our study has several strengths. It is a cross-sectional study with prospective data collection that covers both DDIs with oral and injectable treatments. Additionally, the study is highly representative as it includes multiple cancers. Methodologically, the study underwent independent expert committee assessment and multidisciplinary PIs validation. Our results concerning PIs and DRPs support previous studies: around 1.5 PIs per patient,²⁰ 40% of patients at risk of DDIs,^30,51,52 and 65% of clinically significant DDIs.^53,54

Our study has limitations. Patients were not followed over time to monitor the outcome of the PIs or the emergence of new DDIs. Furthermore, although the Committee estimated the clinical significance of DDIs, this approach did not allow clinically manifested DDIs to be studied. This observational study showed the benefits of a multidisciplinary risk assessment program including a pharmacist in French patients with cancer. Other multidisciplinary interventions have been described worldwide, particularly in older adults with cancer.³⁹ For example, in a US randomized controlled trial of 60 outpatients receiving intravenous chemotherapy, Nipp et al showed the feasibility and effectiveness of an intervention involving pharmacists in the care of older adults with cancer.⁵⁵ More controlled trials of this type are needed to provide evidence on clinical and economic outcomes. To evaluate specifically the pharmacist role, a randomized trial should be conducted with a multidisciplinary team using an electronic prescribing system or a drug-drug interaction database, with one arm including a pharmacist and the second arm without a pharmacist. Nevertheless, pharmacist expertise seems essential in some complex situations where no recommendations exist and has been recommended by the International Society of Geriatric Oncology (SIOG) in its recent guidelines on quality of life in older people with cancer.⁵⁶

Conclusion

In conclusion, multidisciplinary risk analysis before initiation of antitumor treatments, whether oral or intravenous, is a process that needs to be further developed given the growing clinical and therapeutic complexity of oncology. Clinical pharmacists are essential and legitimate members of these risk assessment programs. They help to reduce patient risk and drug-related problems cost-effectively.

Supplementary material

Supplementary material is available at The Oncologist online.

Author contributions

Writing (original draft): J.-S.G., A.T.S., V.K.S., G.P., A.D.T., F.G., B.B., A.J., B.S., R.B. Conceptualization: A.T.S., V.K.S., A.D.T., F.G., B.B., A.J. Investigation: J.-S.G., A.T.S., V.K.S., G.P., S.D.P., C.V., M.V., F.G. Formal analysis: J.-S.G., A.T.S., V.K.S., A.J., A.D.T., M.R.

Funding

CHOPIN clinical study was financed by the French Ministry of Health (Grant « Direction Generale de l’Offre de Soins » for the implementation of clinical pharmacy activities in French hospitals).

Conflicts of Interest

The authors declare no conflict of interest.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.