Evaluating Serum Thymidine Kinase 1 in Patients with Hormone Receptor–Positive Metastatic Breast Cancer Receiving First-line Endocrine Therapy in the SWOG S0226 Trial

Endocrine therapy (ET), including the aromatase inhibitor anastrozole (A) and the selective estrogen receptor (ER) degrader fulvestrant (F), is effective for the treatment of hormone receptor–positive metastatic breast cancer (MBC), but resistance is a major clinical problem (1). We have previously reported that in the SWOG prospective randomized clinical trial S0226, the combination of A + F improves progression-free survival (PFS) and overall survival (OS) in selected patients with MBC, specifically, in patients without prior adjuvant tamoxifen (2, 3). Moreover, targeted agents such as CDK4/6 inhibitors (palbociclib, ribociclib, and abemaciclib), mTOR inhibitors (everolimus), and a PIK3CA inhibitor (alpelisib) improve PFS when combined with standard ET as first- and second-line therapy for hormone receptor–positive MBC (4–7).

Combination ET or ET plus other targeted agents is associated with increased toxicities and costs compared with single-agent ET. Thus, identification of patients who may not need combination therapy would serve to spare them these adverse events. Currently, there is no tool available to help clinicians tailor treatment for each patient, underlining the compelling need for identification of biomarkers of resistance and response to ET agents (8, 9).

Thymidine kinase (TK) is a fundamental enzyme in DNA synthesis and cellular proliferation (10). High versus low levels of cellular TK are associated with poorer prognosis in many cancer types, including breast (11, 12). TK has both cytosolic (TK1) and mitochondrial (TK2) forms (13). TK1 activity can be measured in serum (sTK1) and is frequently elevated in patients with cancer (11, 14). Preliminary studies have suggested that high sTK1 is associated with poor prognosis in MBC treated with chemotherapy or ET (15–17). However, data from the latter are limited.

We conducted this prospective retrospective study of sTK1 in specimens collected from participants in S0226 to test the hypothesis that baseline and serial levels would be prognostic in patients with hormone receptor–positive MBC treated with first-line ET.

This was a prospectively designed retrospective translational medicine study of S0226 (ClinicalTrials.gov number NCT00075764), for which the conduct and final survival outcomes have been reported previously (2, 3). All patients who were eligible and evaluable for the primary analysis of S0226 (N = 694) were considered for inclusion in this study. The study was approved by an Institutional Review Board at each participating site, and all participants (or participant's guardian) provided written informed consent before enrollment. The study was conducted in accordance with the International Conference on Harmonization Good Clinical Practice guidelines and provisions of the Declaration of Helsinki. Submission of blood for future research was voluntary and all patients who contributed blood gave written informed consent for use of the serum samples.

Briefly, S0226 was a phase III, randomized clinical trial in which postmenopausal women with previously untreated hormone receptor–positive [ER-positive, progesterone receptor (PgR)-positive, or both] MBC were randomly assigned to receive either 1 mg A orally daily or to the combination of A + F (F: 500 mg intramuscular as loading dose, followed by 250 mg on days 14, 28 and 250 mg maintenance monthly thereafter). Randomization was stratified by prior adjuvant tamoxifen use (3). A prospectively written study plan for this correlative study was approved by the NCTN Core Correlative Science Committee.

Details regarding patient eligibility, accrual, and overall conduct of the trial have been reported previously (2). ER, PgR, and HER2 status were determined locally by routine pathology at the treating institutions.

Serum samples from the designated collection timepoints of baseline and cycles 2, 3, 4, and 7 were included, if available. Cycle length was 28 days. Serum samples were stored at the SWOG Specimen Repository. After removing identifying information, aliquots of approximately 1 mL for each timepoint were shipped to Biovica Inc. for assessing the enzymatic activity of the sTK by the ELISA-based DiviTum assay (Biovica, Inc.), according to the manufacturer's instructions and as reported previously (15). Biovica was blinded to treatment assignment and outcome. The DiviTum assay has a lower limit of quantitation of 20 DiviTum units per liter (Du/L). Recommended working range is from 20 up to 4,000 Du/L or to the plate reader's asserted upper limit of detection if this limit corresponds to a TK activity lower than 4,000 Du/L. At 100 Du/L, the assay has a coefficient of variation (CV) lower than 20%. The TK activity unit was originally determined with a reference sample of recombinant TK. 1,000 Du/L corresponds to the activity that was obtained with 1,000 ng TK/L. Clinical samples have been observed to have activities ranging from less than 10 Du/L to greater than 100,000 Du/L. All specimens were assayed in duplicate and the final result is the mean of both results. The overall CV of the specimens assayed in this study was ≤ 10%. If an individual sample had a larger CV between the duplicates, the sample was reanalyzed.

For this study, we chose a predetermined cut-off point of 200 Du/L which was an approximate median from prior studies for the hormone receptor–positive MBC population included in S0226 (18). As secondary analysis, we also explored the median sTK1 level and divided the sTK1 values into quartiles (Supplementary Fig. S1).

Assay results were returned to SWOG Statistical Center for analysis to determine the association between sTK levels and clinical outcomes overall and then specifically for patients who had not received adjuvant tamoxifen. The assigned treatment arms of A and A + F were compared using intent-to-treat analyses. The primary outcomes were PFS (defined as time from randomization to progression or death due to any cause) and OS (defined as time from randomization to death from any cause), as prescribed in the parent clinical trial protocol. Patients who did not meet the outcome definition were censored at last follow-up. Survival curves were estimated using the Kaplan–Meier method; 95% confidence intervals (CI) for median (m)PFS and mOS were calculated using the method of Brookmeyer and Crowley. Testing was by log-rank test and Cox regression with stratification on prior adjuvant tamoxifen therapy when both strata are used. The analysis of prognosis from cycle 2 values used a landmarked analysis starting at 56 days. A time-dependent Cox model was also used to include baseline sTK1 and current value of sTK1 with stratification by prior adjuvant tamoxifen use. In this analysis, the most recent measurement of sTK1 was used as a time-dependent covariate with adjustment for baseline sTK1.

All results are reported according to REMARK (Reporting recommendations for tumour MARKer prognostic studies) criteria (19).

A total of 707 patients were enrolled in S0226, but 13 were ineligible, and therefore 694 were included in the original analysis (Fig. 1; ref. 3). sTK1 was evaluated in 1,726 specimens which wereevaluable from baseline or later timepoints [Baseline n = 432 (62% of the 694 patients eligible for S0226); and at cycles 2, 3, 4, 7: n = 365, 342, 324, 263, respectively]. Not all patients had consecutive specimen collection (e.g., specimens might have been available at baseline and cycles 2 and 4, but not cycles 3 and 7).

The patients enrolled in this correlative sTK1 study subcohort had similar characteristics to the original S0226 cohort (Table 1). Survival outcomes did not differ between those patients that had baseline sTK1 values (n = 432) and those that did not (n = 262; P = 0.57 for PFS and P = 0.86 for OS in stratified log-rank tests). Treatment effects comparing single agent A versus combination A + F were also similar in this subset compared with the full cohort. Prior analyses (2, 3) had shown that the benefit of the combination was only seen in the subset without prior adjuvant tamoxifen treatment (N = 282/432; 65.3%), so all prediction analyses were restricted to that subset.

Furthermore, most characteristics were not significantly different between the cohorts with high versus low sTK1, with a few exceptions. Compared with low sTK1, patients with high sTK1 were more likely to have measurable disease (64% vs. 53%, P = 0.019) and more likely to be diagnosed as de novo MBC (48% vs. 36%, P = 0.012), but less likely to have bone-only disease (13% vs. 25%, P = 0.005).

All the samples available were assessed using DiviTum assay with an 100% evaluation rate. Serum TK1 was successfully analyzed in all 1,726 specimens and was elevated in 40% (171/432) of patients at baseline. Median pretreatment level of sTK1 was 135 Du/L (range: 19–35,340; Supplementary Table S1). In serial specimens, sTK1 was elevated in 28% (102/365), 21% (73/342), 17% (55/324), and 15% (40/263) of patients at cycles 2, 3, 4, and 7, respectively. Median sTK1 for the follow-up timepoints was 73 (19–28,470), 65 (19–20,140), 57 (19–55,250), and 50 Du/L (19–27,380) at cycles 2, 3, 4, and 7, respectively.

Overall, patients with high baseline sTK1 had significantly worse PFS than those with low sTK1 [HR = 1.76; 95% CI (1.43–2.16); P < 0.0001; median PFS 11.2 vs. 17.3 months] (Fig. 2A; Table 2). This strong prognostic effect was not altered by inclusion of treatment, age, and disease characteristics in a multivariable analysis (Supplementary Table S2). In patients with no prior adjuvant tamoxifen and high sTK1, PFS was significantly better for those treated with A + F versus A alone [HR = 0.64; 95% CI (0.43–0.95); P = 0.027; median PFS 13.6 vs. 8.7 months] (Fig. 2B; Table 2). Conversely, in patients with low sTK1, we observed no difference in PFS between the A + F versus A arms [HR = 0.85; 95% CI (0.61–1.19); P = 0.34] (Fig. 2C) with no prior adjuvant tamoxifen. The interaction of treatment arm and sTK1 was not statistically significant, though power was low to test the interaction (P = 0.26). Patients with previous tamoxifen did not show a treatment difference overall (2) or separately by sTK1 (Supplementary Fig. S2A and S2B).

Differences in OS for high versus low sTKI at baseline were more pronounced than for PFS. Patients with high sTK1 had significantly worse OS than those with low sTK1 [HR = 2.38; 95% CI (1.91–2.98); P < 0.0001; median OS 30 vs. 58 months] (Fig. 3A; Table 2). Even after adjustment for treatment arm, age, and disease characteristics, sTK1 remained highly prognostic (any other variables prognostic in multivariable analysis). In patients with high sTK1 and no prior adjuvant tamoxifen, OS was significantly better for those treated with combinational therapy (A + F) versus A alone (Fig. 3B; Table 2) [HR = 0.58; 95% CI (0.38–0.87); P = 0.0087; 46 vs. 21 months], whereas for low sTK1, we observed no difference in OS between the A + F versus A arms (Fig. 3C; Table 2) [HR = 0.85; 95% CI (0.56–1.29); P = 0.44]. Despite this apparent difference, there was not a statistically significant interaction (P = 0.16) between randomized treatment and sTK1 for women with no prior tamoxifen.

In addition to the pre-set cut-off points, we performed an exploratory analysis of PFS and OS by quartiles (Q: Q1 <58 Du/L; Q2: 58–134; Q3: 135–491; Q4: >491) of sTK1, and observed a statistically significant association for trend of increasing marker levels (Q4 to Q1 gives a HR = 2.09; 95% CI, 1.57–2.77; P < 0.0001 for PFS and HR = 2.97; 95% CI, 2.16–4.07; P < 0.0001for OS; Supplementary Fig. S1A and S1B).

sTK1 levels were also split into four different quartiles (Q) based on pretreatment level (135 Du/L): Q1 <58 Du/L; Q2: 58–134; Q3: 135–491; Q4: >491. PFS decreased over the four quartiles (P_trend < 0.0001) with medians 17.9, 15.0, 13.8, and 11.0 months, respectively. Comparison of Q4 with Q1 gives a HR = 2.09; 95% CI, 1.57–2.77; P < 0.0001 (Supplementary Fig. S1A). Likewise, OS also decreased over the four quartiles (P_trend < 0.0001) with medians 62.1, 52.6, 41.6, and 29.3 months, respectively. Comparison of Q4 with Q1 gives a HR = 2.97; 95% CI, 2.16–4.07; P < 0.0001 (Supplementary Fig. S1B).

The prognostic role of sTK1 was also tested in subsequent timepoints in landmarked analyses. At cycle 2, patients with high sTK1 had significantly worse subsequent PFS than those with low values (HR = 1.70; P < 0.0001; Fig. 4A). Likewise, serial monitoring of sTK1 levels during therapy at cycles 3, 4, and 7 was also significantly associated with subsequent PFS (Fig. 4B–D), respectively. Patients that had sTK1 measurements at both baseline and cycle 2 were classified by baseline and cycle 2: (i) pos-pos (n = 72); (ii) pos-neg (n = 56); (iii) neg-pos (n = 24); (iv); neg-neg (n = 187). In a joint model, baseline positive sTK1 remained significant [HR = 1.63; 95% CI (1.24–2.14); P = 0.0004] while cycle 2 sTK1 was only marginal [HR = 1.33; 95% CI (1.00–1.77); P = 0.053]. In a time-dependent Cox model, both baseline sTK1 and most current value of sTK1 were included in the same model and both were statistically significant: Baseline [HR = 1.57; 95% CI (1.25–1.97); P = 0.0001; most current [HR = 1.40; 95% CI (1.08–1.81); P = 0.010].

Similar results were observed for OS. At cycle 2, high values had significantly worse subsequent PFS than those with low values (OS HR = 2.51, P < 0.0001; Fig. 4E). Likewise, serial monitoring of sTK1 levels during therapy at cycles 3, 4, and 7 was significantly associated with subsequent OS (Fig. 4F–H), respectively. Patients that had sTK1 measurements at both baseline and cycle 2 were modeled using both values. Baseline positive sTK1 was significant [HR = 2.02; 95% CI (1.51–2.70); P < 0.0001] as was the cycle 2 sTK1 [HR = 1.75; 95% CI (1.29–2.37); P = 0.0004]. In a time-dependent Cox model with both baseline sTK1 and most current value of sTK1 both were statistically significant: Baseline [HR = 1.92; 95% CI (1.51–2.43); P < 0.0001; Most current [HR = 1.96; 95% CI (1.47–2.61); P < 0.0001].

In this prospective–retrospective translational medicine study, we observed that high sTK1 using the DiviTum assay at BL is associated with worse prognosis in patients with hormone receptor–positive MBC starting first-line ET. In addition, high sTK1 at subsequent timepoints was also associated with worse PFS and OS. Furthermore, in patients with high BL sTK1 and no prior adjuvant tamoxifen, PFS and OS were significantly better for those treated with A + F versus A alone (PFS: P = 0.027; OS: P < 0.0001), whereas PFS and OS did not differ between A + F versus A alone for those with low BL sTK1 levels (PFS: P = 0.34; OS: P = 0.44).

The results of this study are consistent with previous reports in which BL sTK1 was demonstrated to be prognostic in patients with MBC on chemotherapy or ET alone or in combination with other agents (15–18, 20, 21). Our results are also concordant with those reported in the TREnd trial, where it was noted that increases in plasma TK1 levels at 4 weeks were associated with worse PFS compared with those with stable or decreasing plasma TK1 levels in patients with hormone receptor–positive MBC treated with CDK 4/6 inhibitors with or without ET (21). Similarly, in S0226, patients with high sTK1 at cycle 2 had worse PFS and OS compared with those patients with low sTK1. However, our study was exclusively conducted in patients with MBC starting first-line ET in patients treated either with A alone or A + F and TK1 was investigated in serum instead of plasma. Our results are partially dissimilar from those of other investigators, who have found that changes from 3 months to 6 months during therapy are significantly correlated to PFS and OS, whereas early changes were not correlated (18).

Taken together with results from previous studies, our results suggest that there are several potential practical applications for sTK1 in the care of patients with MBC. For example, because patients treated in S0226 with low BL sTK1 had nearly identical PFS and OS regardless of whether they received monotherapy or combination ET, baseline sTK1 could potentially be used to select patients best suited for endocrine monotherapy (mono-ET) as upfront treatment, whereas those with elevated sTK1 appeared to benefit from combination A + F.

Furthermore, these results suggest that perhaps sTK1 might be used to identify patients who might not need the addition of CDK4/6 inhibitors to first-line ET. It is possible that patients with low baseline sTK1 have indolent disease and may have excellent clinical outcomes with first-line mono-ET. At present, the current standard of care for patients with hormone receptor–positive MBC is to treat all patients with this combination upfront, exposing them to the additional toxicity and cost of the CDK4/6 inhibitors. Our data suggest that patients with low baseline sTK1 have indolent disease and may do well with single-agent first-line ET. We speculate that identification of this fraction of patients with ER-positive MBC who are very highly likely to benefit from ET monotherapy would help these patients to avoid, or at least delay, these adverse complications. In contrast, our data, consistent with those by McCartney and colleagues (17), also suggest that high sTK1 levels following initiation of a systemic therapy may identify patients who are resistant to the chosen treatment and might benefit from an alternative strategy. Each of these hypotheses would need to be validated in prospective clinical trials.

Strengths of this study include that samples were obtained from a prospective, randomized clinical trial (S0226) with robust clinical outcomes data. In addition, serum samples at BL and at subsequent timepoints were available from a large number of trial participants, with very similar characteristics between the cohort analyzed in this study and the overall patient population of the S0226 study. The patients included in this cohort were also well balanced across treatment arms. Weaknesses of this study include that patients enrolled in S0226 are not necessarily representative of our current population of patients with hormone receptor–positive MBC initiating first-line ET. For example, many patients in S0226 had de novo metastatic disease, and were therefore ET naïve at study entry. In current practice, most patients who present with hormone receptor–positive MBC have previously received adjuvant tamoxifen or aromatase inhibition in the early-stage setting. Furthermore, in the S0226 study, the dose of fulvestrant (250 mg administered intramuscularly) administered was lower than that currently used (500 mg intramuscularly) and there was no single-agent F arm. It is unknown if the superior clinical outcomes comparing the combination of A + F versus A alone would have been even greater if the higher dose of F had been administered to all patients, or if single-agent higher dose F alone would give similar results.

In conclusion, the results of this translational medicine study suggest that BL sTK1 identified patients with hormone receptor–positive MBC with a very favorable prognosis. We speculate that low BL sTK1 might identify patients who have indolent disease and may do well for a long time with first-line ET alone and could safely be treated with mono-ET as upfront treatment for their metastatic disease. Delaying CDK4/6 or mTOR inhibitors might be warranted, decreasing toxicity and cost of therapy. We further speculate that serial monitoring of sTK1 in MBC may allow early identification of patients who are refractory to the chosen treatment and who might benefit from switching to an alternative therapeutic strategy prior to detecting evidence of radiographic disease progression. Further evaluation of the predictive potential of sTK1 in hormone receptor–positive MBC is warranted in prospective clinical trials.

C. Paoletti reports grants from AstraZeneca, Menarini Silicon Biosystem, and Pfizer outside the submitted work; C. Paoletti is currently an employee of Eisai, Inc, but this work is unrelated to her current employment. W.E. Barlow reports grants from NCI and non-financial support from Biovica during the conduct of the study as well as other support from AstraZeneca and Merck outside the submitted work. E.F. Cobain reports personal fees from AstraZeneca, Athenex Oncology, and Ayala Pharmaceuticals outside the submitted work. M. Bergqvist reports personal fees and other support from Biovica outside the submitted work. J.R. Gralow reports personal fees from Roche/Genentech, AstraZeneca, Sandoz/Hexal AG, Puma, SeaGen, and Genomic Health/Exact Sciences, and other support from Novartis and Radius outside the submitted work. K.S. Albain reports other support from Seattle Genetics, AstraZeneca, Daiichi-Sankyo, and Seattle Genetics/Axio, and personal fees from Genentech/Roche and Genomic Health/Exact Sciences outside the submitted work. L. Pusztai reports grants and personal fees from from AstraZeneca, Merck, Novartis, Bristol-Myers Squibb, Genentech, Eisai, Pieris, Immunomedics, Seattle Genetics, Clovis, Syndax, H3Bio, and Daiichi outside the submitted work. P. Sharma reports other support from Cosmo Biosciences, Celgene, and Bristol-Myers Squibb; personal fees and other support from Novartis and Merck; and personal fees from Seattle Genetics, Exact Sciences, AstraZeneca, and Immunomedics outside the submitted work. A.K. Godwin reports grants from NIH/NIGMS/NCI and SWOG during the conduct of the study as well as personal fees from Sinochips Diagnostics, grants from VITAC, and other support from Clara Biotech outside the submitted work. D.F. Hayes reports personal fees and non-financial support from Bioveca during the conduct of the study as well as personal fees from Cepheid, Freenome, Inc., Artiman Ventures (Cellworks), Lexent Bio, EPIC Sciences, Inc, Salutogenic Innovations, LLC, L-Nutra, OncoCyte, Turnstone Biologics, Tempus, and Predictus, and other support from Inbiomotion, AstraZeneca, Pfizer, and Menarini Silicon Biosystems outside the submitted work; in addition, D.F. Hayes has a patent for Diagnosis and Treatment of Breast Cancer (Patent No.: US 8,790,878 B2. Date of Patent: Jul. 29, 2014. Applicant Proprietor: University of Michigan) and D.F. Hayes is designated as inventor/co-inventor (issued, licensed, and with royalties paid from Menarini Silicon Biosystems). J.M. Rae reports grants from NIH/NCI and other support from AstraZeneca and Biovica during the conduct of the study. No disclosures were reported by the other authors.

C. Paoletti: Conceptualization, formal analysis, methodology, writing–original draft, writing–review and editing. W.E. Barlow: Conceptualization, data curation, formal analysis, methodology, writing–original draft, writing–review and editing. E.F. Cobain: Writing–original draft, writing–review and editing. M. Bergqvist: Conceptualization, resources, data curation, formal analysis, methodology, writing–original draft, writing–review and editing. R.S. Mehta: Conceptualization, writing–original draft, writing–review and editing. J.R. Gralow: Conceptualization, writing–original draft, writing–review and editing. G.N. Hortobagyi: Conceptualization, writing–original draft, writing–review and editing. K.S. Albain: Conceptualization, writing–original draft, writing–review and editing. L. Pusztai: Conceptualization, writing–original draft, writing–review and editing. P. Sharma: Conceptualization, writing–original draft, writing–review and editing. A.K. Godwin: Conceptualization, writing–original draft, writing–review and editing. A.M. Thompson: Conceptualization, writing–original draft, writing–review and editing. D.F. Hayes: Conceptualization, resources, formal analysis, methodology, writing–original draft, writing–review and editing. J.M. Rae: Conceptualization, resources, data curation, methodology, writing–original draft, writing–review and editing.

Research reported in this article was supported by the NIH/NCI grants U10CA180888, U10CA180819, U24CA196175; and in part by AstraZeneca. Biovica, Inc. has funded the TK1 testing for this correlative study. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.