A case–control study to assess the ability of the thymine challenge test to predict patients with severe to life threatening fluoropyrimidine‐induced gastrointestinal toxicity

Abstract

Aims

A previous study suggested that a thymine (THY) challenge dose could detect aberrant pharmacokinetics in known cases of fluoropyrimidine toxicity compared with healthy volunteers. The preliminary data suggested that urine sampling also could detect this aberrant disposition. The aim of this case–control study was to assess the ability of the urinary THY challenge test to discriminate cases of severe gastrointestinal toxicity in a cohort of patients treated with 5‐fluorouracil or capecitabine.

Methods

Patients (n = 37) received a 250 mg (per os) dose of THY and a cumulative urine sample was collected for 0–4 h. The urinary amounts of THY and metabolite dihydrothymine (DHT) were determined by liquid chromatography/mass spectrometry. Genomic DNA was analysed for DPYD gene variants. Renal function was estimated from blood creatinine levels. Cases (n = 9) and noncases (n = 23) of severe (grade ≥ 3) gastrointestinal toxicity were defined based on Common Terminology Criteria for Adverse Events.

Results

The median THY/DHT ratios were 6.2 (interquartile range 2.9–6.4) in cases, including the 2 patients who were DPYD heterozygous carriers. However, this was not significantly different (P = .07) from the THY/DHT in noncases (median 2.6, interquartile range 2.8–4.2). Although creatinine clearance was lower (P = .001) in cases, renal function could not discriminate cases from noncases. However, logistic regression analysis using both of these explanatory variables could discriminate most cases (receiver operating characteristic area 0.8792, 95% confidence interval 0.72–1.00).

Conclusions

The THY challenge test combined with a patient's renal function may be useful as a phenotypic diagnostic test to detect risk of life‐threatening fluoropyrimidine gastrointestinal toxicity.

What is already known about this subject

• Most patients with severe fluoropyrimidine toxicity do not have inherited dihydropyrimidine dehydrogenase deficiency; phenotypic tests may be more helpful than genotyping

• A thymine challenge detected aberrant pyrimidine pharmacokinetics in toxicity cases compared with healthy volunteers

• It is not known if this test can discriminate tolerant patients from toxicity cases

What this study adds

• Patients with a high thymine/dihydrothymine urinary ratio after a thymine challenge test may predict those vulnerable to life‐threatening treatment‐related toxicity

• The thymine challenge test combined with information on renal function appears to discriminate the majority of patients who cannot tolerate standard doses of fluoropyrimidine drugs

1. INTRODUCTION

A key component of treatment of gastrointestinal (GI) and breast cancer, 5‐fluorouracil (5‐FU) is administered either intravenously (IV) or as the oral prodrug, capecitabine. These fluoropyrimidine drugs cause a range of normal tissue toxicities that can lead to dose decreases or delay successive chemotherapy cycles. Of particular clinical concern are GI toxicity (diarrhoea and mucositis), hand–foot syndrome and neutropenia. The prevalence of severe to life threatening GI toxicity (Grade 3 or greater) ranges between 10 and 20%1, 2, 3, 4 and lethal toxicity also occurs.5 Fluoropyrimidine treatment is discontinued because of these safety concerns in about 8% of patients.1, 2

The enzyme dihydropyrimidine dehydrogenase (DPD), catalyses the conversion of 5‐FU to dihydrofluorouracil, and metabolic clearance eliminates around 80% of a dose of 5‐FU. Hence a deficiency in the DPD enzyme leads to elevated drug plasma concentrations, greatly increasing the risk of toxicity.6, 7

Deleterious variants in the DPYD gene, which codes for this enzyme, have been shown to associate with cases of toxicity. The most consistently associated variants are c.1905 + 1G > A (DPYD*2A; rs3918290), c.2846A > T (rs6737698), c.1679 T > G (DPYD*13; rs55886062) and, more recently, c.1236G > A/HapB3 (rs56038477/rs75017182/rs53676561).8, 9, 10 A recent study11 reported that genotype‐guided dose adjustment following prospective testing for these variants decreased the relative risk of treatment‐related toxicity. However, toxicity was still common with 23% of DPYD wild‐type patients receiving standard dosage experiencing severe Grade 3 or greater adverse events. Moreover, carriers of these variants are relatively uncommon, comprising only 1–3% of populations.12 As a result, DPYD genotyping has low sensitivity, since only about 1/4 of patients with severe fluoropyrimidine toxicity have 1 or more DPYD variant.9, 10, 12, 13, 14 Recent comprehensive DPYD gene sequencing of a large cohort of patients has not substantially improved the sensitivity for severe GI toxicity.15 Epigenetic factors may also regulate expression of the enzyme.16, 17 Thus, whilst there is a clear intrinsic risk of toxicity for an individual with a deleterious DPYD gene variant, any other cause of a decrease in the activity of the enzyme (perhaps due to epigenetic factors) could also play a role in the risk of toxicity. Theoretically, assessing DPD activity (i.e. phenotyping) in an individual prior to drug exposure should be more effective than genotyping.

Since DPD also catalyses conversion of endogenous pyrimidine bases (uracil and thymine), a number of studies have investigated the ability of the ratio of basal plasma uracil to its metabolite dihydrouracil to detect low DPD activity (reviewed in18). However, endogenous uracil concentrations vary considerably throughout the day and these fluctuations may influence the sensitivity of endogenous uracil as a phenotypic probe.

Recently we have developed a challenge test using thymine (THY).19, 20 Challenge tests may be particularly useful as the pyrimidine dose temporarily saturates the DPD enzyme resulting in zero order kinetics in a similar way to 5‐FU dosing. Moreover, THY may be a more sensitive test than uracil, since the basal levels of endogenous THY and its metabolite dihydrothymine (DHT) are much lower and less variable.

We previously assessed the THY challenge test in a small retrospective cohort study of 6 cases of fluoropyrimidine toxicity.20 The plasma pharmacokinetic profile of THY elimination was able to detect undiagnosed inherited DPYD deficiency in 2 patients. However, 3 other patients exhibited substantial elevations of THY plasma concentrations very soon after dosing, as well as raised urinary THY excretion, which was not attributable to decreased THY clearance. We suggested that this may be the result of an unrecognised phenomenon of enhanced THY absorption (ETA), which may provide an explanation for fluoropyrimidine toxicity in addition to DPD deficiency.

However, our previous study did not assess the THY challenge test in cancer patients who tolerated fluoropyrimidine treatment. We now report the results of the THY challenge test combined with noninvasive urine sampling in a small case–control study of a cohort of patients treated with 5‐FU or capecitabine. The aim of this study was to establish whether the THY challenge test using a cumulative 0–4 h urine collection could discriminate cases of severe toxicity. Only patients in receipt of fluoropyrimidine monotherapy were recruited into this study to ensure that cases of toxicity were most likely to be attributable to fluoropyrimidine.

2. METHODS

This was an observational study to assess the ability of the THY challenge test to detect patients with severe to life‐threatening fluoropyrimidine‐induced GI toxicity (Australian New Zealand Clinical Trials Registry: ACTRN12615000586516). Ethical approval for the study was given by Northern A Health and Disability Ethics Committee of NZ (14/NTA/186). Eligibility criteria were as follows: able to provide written informed consent, age >18 years, with histologically confirmed GI cancer or metastatic breast cancer. Patients who were about to receive, or had recently completed, treatment with 5‐FU or capecitabine monotherapy as part of standard care were recruited into the study. In addition, 6 patients with a history of fluoropyrimidine toxicity were also recruited into the study. Exclusion criteria were pregnant or breast‐feeding, or receipt of concurrent abdomino‐pelvic radiation therapy. All laboratory analyses were undertaken using deidentified patient samples by researchers blinded to the clinical data, while adverse event data were collected by clinical staff blinded to the research laboratory data.

Based on the earlier study in healthy volunteers (n = 12) and a case series (n = 6) we aimed to recruit a minimum of n = 15 (noncases; i.e. patients who tolerated treatment with fluoropyrimidines) and n = 15 patients with severe GI toxicity (cases). This sample size was calculated based on the log THY/DHT ratios previously observed in the healthy volunteers (mean 0.05, SD 0.32) and the case series (mean 0.86, SD 0.13) using the sample SD of the healthy volunteers and a minimum important difference of 0.4 and a power of 90% at the 2‐sided 0.05 level. Sample size calculations were conservative to allow for the planned use of the Wilcoxon rank sum test.

2.1. THY loading test

THY was purchased from MP Biomedicals, Australasia Pty Limited, analytical and microbiological purity was confirmed (ANQual, University of Auckland), and gelatine capsules containing 250 mg THY prepared by a registered pharmacist. THY was compounded under Section 29 of the NZ Medicines Act 1981. Challenge testing was undertaken in the morning after a light breakfast avoiding meat.21 Participants voided their bladder and then were administered a single gelatine capsule of THY (250 mg, per os) with a drink of water. A cumulative (total) urine sample was then collected for 4 hours. The volume and pH were measured, and aliquots of the sample stored at −80°C until analysis.

2.2. Preparation of urine samples

THY was purchased from Sigma‐Aldrich, DHT was purchased from Toronto Research Chemicals and the internal standard, THY‐d₄, purchased from Cambridge Isotope Laboratories Inc. Urine samples were diluted 1 in 10 with deionised water before analysis. The internal standard working solution (200 μL, 4.41 μM of THY‐d₄) and 800 μL of acetonitrile were added to the diluted urine sample (200 μL). Proteins were precipitated (12 h, −20°C), then centrifuged (10 min, 8200 × g) and the supernatant (1 mL) transferred to a new tube and evaporated to dryness under nitrogen gas. The samples were reconstituted in 0.1% formic acid in liquid chromatography (LC)‐grade water (150 μL) prior to analysis.

2.3. Quantification of THY and DHT

Aliquots of the urine samples were analysed using high‐performance LC–atmospheric pressure chemical ionisation/mass spectrometry. Comprising an Agilent 1260 Infinity LC series coupled with a single quadrupole mass spectrometer, with atmospheric pressure chemical ionization mode (Agilent 6150 Single Quadrupole LC/mass spectrometry) was used. The previously validated method21 was adapted with minor modifications. Nitrogen drying gas flow was 12.0 L/min and nebuliser pressure was 2.4 bar. The drying gas temperature was 250°C, vaporiser temperature was 250°C, capillary voltage 400 V, corona current 4.0 μA and dwell time was 390 ms. Detection by selective ion monitoring (SIM; positive ion mode) for each mass ion was used: m/z 131 (THY‐d₄), 127 (THY) and 129 (DHT). The limit of quantification (LOQ) for THY and DHT was 25 ng/mL. Calibration curves were linear (r² > 0.99) between 25 and 1000 ng/mL. The analytes have previously been shown to be stable in urine for 365 days.21 If the analyte in a sample was outside of these values the sample was reanalysed by reconstitution in a different volume of mobile phase. Intra assay and inter assay CV were all below 15% and accuracy was within 9%. The amount of THY and DHT was then determined from the analyte concentration and the known volume of urine and the data reported as the ratio (THY/DHT) with values shown as the average of 2 replicate analyses. The cumulative amount excreted (Ae 0–4 h) as THY as a % of the dose was also calculated.

2.4. Fluoropyrimidine treatment

All patients received normal standard of care based on the institutional guidelines for fluoropyrimidine treatment. For GI cancer capecitabine was administered at 1250 mg/m² (per os, twice a day) for 14 days every 3 weeks, with 25% dose reduction if over 70 years or if estimated creatinine clearance 30–50 mL/min. If capecitabine was contraindicated, for example due to frailty, 5‐FU was administered as 370 mg/m² IV bolus, weekly. For metastatic breast cancer, capecitabine was administered at a fixed oral dose of 2 g (twice a day).

2.5. Identification of cases

Adverse event data following fluoropyrimidine treatment was collected onto a case record form and were graded (G) from 0 (no toxicity) to 5 (death) according to the Common Terminology Criteria for Adverse Events scale (version 4.0). Fluoropyrimidine‐related adverse events collected were diarrhoea, oral mucositis, hand–foot syndrome, neutropenia, hyperbilirubinaemia, cardiac toxicity and any other toxicities. Information on dose reduction and dose delays were also collected. Cases were defined as G3 (or greater) GI toxicity (diarrhoea and/or mucositis) that resulted in hospitalisation and/or treatment withdrawal. Noncases (controls) were defined as G2 (or less) GI toxicity and absence of other severe (G3) toxicities.

2.6. Other patient information

Age, sex, height, weight and self‐identified ancestry were recorded. Serum creatinine was determined by the hospital clinical laboratory and the estimated creatinine clearance (CrCl) calculated using the Cockcroft–Gault formula22 using the software, https://www.mdcalc.com/creatinine-clearance-cockcroft-gault-equation. Estimated CrCl at the time closest to the initiation of fluoropyrimidine treatment was reported for each patient.

2.7. Gene variant analysis

A baseline blood sample was collected into a 10 mL PAXgene DNA tube (Qiagen). Genomic DNA was extracted according to the manufacturer's instructions (PAXgene Blood DNA kit, Qiagen). De‐identified genomic DNA was then double coded and aliquots screened for following deleterious DPYD variants: rs3918290 (c.1905 + 1G > A; IVS14 + 1G), rs6737698 (c.2846A > T; D949V), rs55886062 (c.1679 T > G; I560S, DPYD*13), rs56038477/rs75017182/rs56276561 (HapB3) using custom designed Sequenom MassARRAY iPLEX assays (primers, Table S1) at the Grafton Clinical Genomics Centre, Auckland, New Zealand.

2.8. Statistical analyses

Graphs were prepared and statistical analyses undertaken using GraphPad Prism 7.03. For continuous variables, median and interquartile range (IQR) are presented and groups were compared using the Mann–Whitney U test. Proportions were compared using Fisher's exact test. All tests were 2‐sided and not corrected for multiple testing; P values <.05 were considered statistically significant. Logistic regression was used to obtain predicted probabilities of being a case. Receiver operating characteristic (ROC) curves were used to compare continuous values of the diagnostic test. Confidence intervals for the area under the curve and a P‐value comparing the curves were calculated using the method described in Hanley and McNeil.23

3. RESULTS

Thirty‐seven patients were assessed in this observational cohort study, comprising 29 GI cancer patients, 6 metastatic breast cancer and 1 patient with concurrent metastatic breast cancer and GI cancer (Table 1). Patients received fluoropyrimidine monotherapy either as capecitabine (n = 31) or as 5‐FU, IV bolus (n = 5), 1 patient (4016) began capecitabine treatment and was transferred to IV bolus 5‐FU (Table 1).

Table 1.

Characteristics of cases and noncases

	Cases n = 9	Noncases n = 23	P‐value	Total n = 32
Cancer type
Gastrointestinal	7	19		26
Metastatic breast	1	4	.5	5
Metastatic breast and GI	1	0		1
Age (median, range)	73 (50, 75)	67 (35, 80)	.3
Sex
Female	6	12		18
Male	3	11	.6	14
Ancestry
European	9	15		24
Chinese	0	4		4
Indian	0	2	.6	2
Other	0	2		2
Drug
5‐fluorouracil	2	2		4
Capecitabine	6	21	.1	27
Both	1	0		1
Percentage dose reduction
Median (range)	100 (25, 100)	15 (0, 40)	<.0001
Timing of thymine test
Before fluoropyrimidine treatment	3	19		22
After fluoropyrimidine treatment	2	3	NA	5
After toxicity	4	1		5

Of the 37 patients, 9 met the case definition of G3 (or greater) GI toxicity (diarrhoea and/or mucositis) which resulted in hospitalisation and/or treatment withdrawal (Table 2). Twenty‐three patients met the definition for controls, i.e. G2 (or less) GI toxicity and absence of other G3 toxicities (Table 2). There were 5 patients classified as other toxicity, 4 experienced G3 hand–foot syndrome and 1 experienced G3 neutropenia. There was no statistically significant difference between cases and noncases for age, sex, ancestry, cancer type or fluoropyrimidine therapy (Table 1). Of the 9 cases, treatment was withdrawn for 6 and the remaining 3 had dose reductions. Dose reduction was significantly greater in cases than noncases (P < .0001; Table 1).

Table 2.

Fluoropyrimidine‐related adverse events in the patient cohort

	Treatment cycle
ID	Cycle 1	Cycle 2	Cycle 3	Cycle 4	Cycle 5	Category
4001	G1D	G1D	G0	G0	G0	Noncase
4002	G4D [plus others listed in footnote#] hospitalised 17 days	Treatment withdrawn				Case
4003	G0	G1HF	G1HF	G1HF	G1HF	Noncase
4004	G0	G1D	G1D, G1B	G1D, G2B	G1HF, G2B	Noncase
4005	G1 HF	G2 HF	G2HF	G2HF	G2HF	Noncase
4006 a	G1D, G2HF	G1D, G3HF	G0	G1N	G0	Other
4007	G3D, G2N, G1CR, hospitalised 7 days	G1D	G1D	G0	G0	Case
4008	G1DM	G1D	G2D, G1M, G2HF	G1D, G1HF	G1D, G1HF	Noncase
4009	G1D, G1NAU	G0	G0	G0	G0	Noncase
4010	G0	G1HF	G2HF	G2HF, G1D	G1HF, G1N	Noncase
4011	G2, NAU	G3N	G1N	G1HF, G2N	G1HF, G1N	Other
4012	G1D, G1CR	G1D	G1D, N			Noncase
4013	G1N	G0	G0	G1N	G0	Noncase
4014 b	G1D	G0	G1D	G3D, G2N, hospitalised 6 days	Treatment withdrawn	Case
4015 b	G1NAU	G3D, G2HF, G4N, hospitalised 12 days	Treatment withdrawn			Case
4016	G2D, G1M	G2D G2CR	G3D, hospitalised 5 days	Treatment withdrawnc		Case
4017 b	G2D, G2M, G3HF (cycle not known)					Other
4018 b	G1D	G0	G3D, hospitalised 10 days	Treatment withdrawn		Case
4019	G1HF	G2M, G1HF	G1HF	G1 HF	G2 HF	Noncase
4020	G0	G0	G2HF	G1D, M, HF	G1D, M, HF	Noncase
4021	G1M	G2HF	G2HF	G0	G2HF, G1DM	Noncase
4022 b	G1 D	G1HF	G0	G2HF	G2HF	Noncase
4023 a	G3D, hospitalised 6 days	Treatment withdrawn				Case
4024	G0	G1M	G1M, G2HF	G1HF	G1D, HF	Noncase
4025	G3HF	G1HF, D	G2HF, G1D	G2D, HF	G1D, M, G2HF	Other
4026	G0	G1HF	G2HF, G1D	G2HF, G1D	G2HF, G1D	Noncase
4027	G1M, HF	G2HF	G1HF	G1HF		Noncase
4028	G1D	G1D, M, HF	G1M	G0	G2D	Noncase
4029 b	G3D, G1M, G1HF, G2V Hospitalised 2 days	G1D, G2HF	G2 HF	G3D hospitalised 1 day	G0	Case
4030 a	G1D	G1D	G1D	G1D		Noncase
4031 a	G0	G2N	G0	G0		Noncase
4032	G0	G0	G0			Noncase
4033 a	G1HF	G2HF	G3D, G2HF Hospitalised 7 days	Treatment withdrawn		Case
4034	G2D	G1HF	G2HF, D	G2HF, D		Noncase
4035	G2M	G1M	G3HF	G2HF		Other
4036	G0	G0	G1HF	G1HF		Noncase
4037 a	G1D	G1D	G1D	G1D	G2HF, G1B	Noncase

G = grade (0–4); D = diarrhoea, M = oral mucositis, N = neutropenia; HF = Hand‐Foot syndrome; NAU = nausea; V = vomiting; CR = creatinine; B = hyperbilirubinemia

#Other toxicities for patient 4002: G4 acute kidney injury, G4 aspiration, G3 hyperbilirubinemia, G3 anorexia, G3 atrial fibrillation with rapid ventricular response and G2 gastric haemorrhage

^aPatients 4006, 4030, 4031, 4037 were tested after completion of scheduled treatment or after recovery from toxicity (4023, 4033).

^bHistorical patients recruited into the study to enrich grade 3 toxicity cases.

^cDue to fatigue.

Severe GI toxicity was early onset (cycle 1–2) in 55% of the cases (Table 2). One of these patients (4007) restarted treatment following a 25% dose reduction and then tolerated 5‐FU (IV, bolus). One patient (4029) restarted treatment with a 50% dose decrease but did not tolerate this dose and had a second episode of G3 diarrhoea at cycle 4. Later onset (after cycle 3) toxicity was observed in 4 patients. Of note, 1 of these patients (4014) had received an initial 30% dose reduction due to previous adverse events on neoadjuvant therapy. The length of hospitalisation for cases ranged from 1–17 days (median 7 days, Table 2).

Two patients (one prospective and 1 historical) were heterozygous carriers of DPYD variants that have validated deleterious effects on DPD function (rs3918290, IVS14 + 1G and rs6737698, D949V). Both of these heterozygous carriers (4002, 4014) had severe toxicity after treatment with capecitabine (Table 2) and were hospitalised for 17 and 6 days respectively.

The THY challenge was carried out prior to scheduled treatment in 2/3 of the participants (22 of 32); 5 were tested after completion of their scheduled fluoropyrimidine treatment and 5 were tested after experiencing fluoropyrimidine toxicity. All patients tolerated the 250 mg THY dose, with no adverse effects observed or reported. The percentage of dose recovered in the 0–4 h urine ranged between 0.001 and 6.19 and was highly skewed. The median Ae(%) value for the cohort was 0.38% (IQR 0.04–1.0). Eleven of the patients had values above the interquartile range and this may suggest ETA phenotype in these subjects. Indeed, the median Ae(%) in this upper quartile was almost 10‐fold higher than the values below this this upper quartile (1.1%, IQR 1.0–2.4 vs 0.08%, IQR 0.01–0.1; P < .0001).

The 0–4 h THY/DHT urinary ratio of cases and noncases following the THY dose, these ranged from 0.26 to 19.6 (Figure 1). The ratios observed in the cases (median 6.2, IQR: 2.9–6.4) were similar to the THY/DHT ratios in the previously reported case series (median 7.1, IQR: 4.8–9.2).20 The 2 patients with confirmed DPYD deleterious variants had elevated THY/DHT ratios (6.24 and 6.20), confirming that this test can detect inherited partial DPD deficiency. However, the THY/DHT urinary ratio in cases was not significantly different (P = 0.07) to that observed in the noncases (median 2.6, IQR 0.8–4.24).

Figure 1.

The thymine challenge test result (THY/DHT urinary ratio) observed in noncases (n = 23) and cases (n = 9) of severe gastrointestinal toxicity. Open circles are the patients identified as DPYD variant carriers. The dotted line is the previously proposed cutpoint THY/DHT >4.0 for discrimination of cases.20 Median ratio (interquartile range) was 2.6, 0.8–4.24 in noncases and was 6.2, 2.9–6.4 in cases. There was no significant difference between cases and noncases (P > 0.05).

Fluoropyrimidine drugs undergo renal elimination and adjustments to capecitabine dosing are recommended when estimated CrCl is below 50 mL/min. One patient (4023) required a starting dose adjustment due to a low CrCl of 48 mL/min and did receive an adjusted dose. Figure 2 compares the kidney function measured by estimated CrCl at the time of fluoropyrimidine therapy in noncases vs cases. The median (IQR) was 61 (58–73) mL/min for cases and 91 (79–102) mL/min for noncases, a statistically significant difference (P = .001). However, renal function alone could not discriminate cases from noncases.

Figure 2.

Comparison of renal function in cases and noncases. Renal function was based on estimated creatinine clearance (CrCl).22 Patients with CrCl >110 mL/min were overweight. Median CrCl was significantly different between noncases (91 mL/min, interquartile range 79–102) compared with cases (61 mL/min, interquartile range 58–73), P < .005

Comparison of patients with similar THY/DHT values, suggested that patients who had elevated ratios (>4.0) did not experience clinically significant toxicity when renal function was good. For example, patient 4019 had a THY/DHT ratio of 4.77, CrCl of 93 mL/min and experienced G0 diarrhoea and G1 hand–foot syndrome. Patient 4021 had THY ratio of 4.35 a CrCl of 95 mL/min and had G1 mucositis. In contrast, patient 4018 had a THY ratio of 4.95, CrCl was 69 mL/min and this patient experienced G3 diarrhoea.

We then considered 2 ways of taking kidney function into account in the test. Firstly, we divided the THY/DHT by CrCl to obtain adjusted scores, THY/DHT_norm (Figure 3a). Secondly, a logistic regression model with THY/DHT and CrCl as explanatory variables (Table 3) was used determine the predicted probabilities of being a case and the resulting diagnostic test scores are shown (Figure 3b). The normalised values THY/DHT_norm could detect a significant difference (P = .0087) between cases and noncases. The ROC curves for all 3 putative tests are shown in Figure 4. Whilst the logistic regression approach seems to provide the best discrimination between cases and noncases there was no statistically significant difference between the ROC curve for all 3 approaches (P = .09). With a cut‐off based on a specificity of 100%, the adjusted THY/DHT_norm value test correctly predicted 6 of 9 cases.

Figure 3.

(A) Thymine challenge test result adjusted to estimated renal function (THY/DHT_norm). Median normalised values were significantly different (P < .01) between cases (0.084, interquartile range 0.03–0.13) and noncases (0.029, 0.011–0.054). Open circles are the patients identified as DPYD variant carriers (B) The predicted probability score from logistic regression model using THY test result and CrCl as explanatory variables. Median values are shown

Table 3.

Logistic regression model for thymine challenge test (THY/DHT ratio) and estimated renal function

Ccn	Coefficient	Standard error	z	P > [z]	[95% Confidence interval]
THY_DHT	0.257531	0.2502487	1.03	0.303	–0.2329434	0.7480135
CrCln	–0.0834603	0.0382244	–2.18	0.029	–0.1583786	–0.0085419
_cons	4.614254	3.154017	1.46	0.143	–1.567506	10.79601

Figure 4.

Receiver operator characteristic (ROC) curves using thymine test result (THY/DHT); data normalised to estimated renal function (THY/DHT_norm) and predicted probability score (logistic regression model). ROC area (95% confidence interval) were 0.7101 (0.4949–0.92538); 0.7971 (0.57448–1.00) and 0.8792 (0.72778–1.00), respectively

4. DISCUSSION

Numerous studies have assessed endogenous uracil levels as a surrogate measure of DPD enzyme activity (reviewed in18). However, our study is the first to assess a THY challenge test in a cohort of patients to determine the ability of this test to discriminate between patients who can vs those who cannot tolerate standard fluoropyrimidine treatment. Although the data reported in this manuscript are from a small study, where the majority of the patients were treated with capecitabine monotherapy, the cohort appeared to reflect the disposition of fluoropyrimidines observed in numerous larger studies. In particular, the prospective incidence of G3 and G4 GI toxicity was similar to that reported in larger trials.1, 2, 3, 4, 9 In addition, the prevalence of deleterious DPYD variants, calculated from the prospective patients only, was similar to the expected prevalence in European populations, which was the predominant ancestry reported in the cohort. Our data were also consistent with the extensive literature on the low sensitivity of DPYD variant testing with detection of risk of G3–4 GI adverse events.

Our earlier work assessing the THY challenge test focussed on the plasma pharmacokinetics of THY and its catabolites in healthy participants19 and in a group of 6 patients who had previously experienced G3–4 fluoropyrimidine toxicity.20 Intensive serial blood sampling was required to detect individuals with aberrant THY pharmacokinetic disposition (either the ETA phenotype or decreased metabolic clearance). This intensive approach will probably not be logistically feasible in many oncology clinics. Our preliminary data also demonstrated that the THY/DHT urinary ratio was elevated in the 6 fluoropyrimidine toxicity cases assessed.20 Hence, for this current study we chose to focus on assessment of the THY challenge test combined with collection of 0–4‐h cumulative urine rather than intensive serial pharmacokinetic sampling.

The values for the 0–4 h‐urinary THY/DHT ratios in the current study were comparable to the range reported previously in healthy volunteers and also in the previous case series.19, 20 We observed a non‐normal distribution of urinary THY/DHT ratios, with a remarkably high number of patients in the cohort (48%) with elevated ratios (THY/DHT > 4). This finding reflects earlier work, which identified 48% of patients with elevated uracil/dihydrouracil ratios.24 This suggests that aberrant pyrimidine disposition may not be uncommon in cancer patients.

In this small preliminary study, the THY challenge test result had relatively good performance (ROC area 0.71) for detection of patients at risk of severe toxicity. Previous oral loading studies using uracil as a phenotypic probe have also shown some success in discriminating DPD activity status. However, in some of these studies the uracil metabolic ratio was evaluated only in a subset of patients with previously known DPYD variant alleles and/or inherited low DPD activity in peripheral blood mononuclear cells. Recently, a uracil challenge dose was assessed in a toxicity case series (n = 47). Using a limited sampling strategy (T = 120 min) the plasma uracil/dihydrouracil ratio could correctly identify 19 of the toxicity cases with known inherited DPYD deficiency.25 However, the majority (60%) of these severe‐toxicity cases had apparently normal DPD activity (as measured in peripheral blood mononuclear cells) and could not be identified by this limited sampling strategy. This reiterates that an inherited deficiency in DPD activity only accounts for a small proportion of toxicity cases and is in agreement with the data presented in this manuscript.

The uracil challenge dose was also administered to a group of colorectal cancer patients 48 h after capecitabine treatment.26 This study compared the uracil pharmacokinetics in 12 patients with metastatic disease with 12 patients receiving adjuvant treatment. No information about capecitabine‐related adverse events was provided for these patients. Notably, the ability of the uracil challenge dose (using a limited sampling strategy) to discriminate toxicity cases, including those who are not DPYD variant, from patients who tolerate treatment does not appear to have been directly assessed. In contrast, assessment of the ¹³C‐uracil breath test in a small case–control study of patients (n = 20, Grade 0–1 toxicity vs n = 13, G3–4 toxicity) could discriminate toxicity cases with a sensitivity of 61.5% and specificity of 85%.27 This is remarkably similar to the performance of the THY challenge test reported in this manuscript.

Although THY challenge could detect known DPYD variant carriers, we did not find a simple association with aberrant THY disposition (i.e. elevated THY/DHT urinary ratio) and G3–4 GI toxicity. Renal clearance is also known to be important for 5‐FU disposition and our data suggest that although some patients may be vulnerable (i.e. THY/DHT >4), if renal function was good, they appeared to tolerate standard fluoropyrimidine treatment. With both aberrant THY disposition (THY/DHT ratio) and estimated renal function (CrCl) as explanatory variables, logistic regression provided a predicted probability score and this improved the performance of the challenge test (ROC area 0.88). The additional role of moderately decreased renal function on risk of toxicity is not unexpected since previous studies have reported an inverse association between creatinine clearance and risk of fluoropyrimidine toxicity.28, 29 In addition, renal function has been reported to contribute to risk of toxicity, with OR of 0.85 (95% confidence interval 0.78–0.94) per 10 mL/min/1.73m² of CrCl.30

However, this approach could not account for the severe toxicity observed in 3 patients. Other factors that may influence the risk of severe toxicity include low levels of thymidylate synthase (TYMS), the therapeutic target of fluoropyrimidine drugs, as well as variability in transmembrane uptake.31 Associations between TYMS genotype and toxicity risk have been reported previously, although the importance of this is unclear.32, 33 We are currently evaluating this cohort of patients for variants of TYMS.

Our goal is to develop a predictive marker for tolerance of standard doses of fluoropyrimidine drugs, regardless of the underlying cause of aberrant pyrimidine disposition. We are currently assessing the performance of the THY challenge test in a larger prospective study of patients receiving standard fluoropyrimidine combination regimens for treatment of GI or advanced breast cancer.

COMPETING INTERESTS

There are no competing interests to declare.

CONTRIBUTORS

N.H. and J.D. conceived the study. N.H., J.D., K.S., D.P. and M.F. contributed to the research design. K.B., C.B., S.H.J., E.B., P.B., D.P. and M.F. undertook data collection. N.H., J.D., K.B., K.S., D.P. and M.F. prepared the manuscript.

DATA SHARING

Author selects not to share data.

Supporting information

Table S1. Primer sequences used

Helsby NA, Duley J, Burns KE, et al. A case–control study to assess the ability of the thymine challenge test to predict patients with severe to life threatening fluoropyrimidine‐induced gastrointestinal toxicity. Br J Clin Pharmacol. 2020;86:155–164. 10.1111/bcp.14153