Clinical utility of longitudinal ctDNA monitoring by multiplex MS-ddPCR for risk stratification and follow-up in rectal cancer

Abstract

Introduction

Personalized treatment strategies in rectal cancer aim to balance escalation and de-escalation based on recurrence risk. Accurately identifying which patients will benefit from each approach is essential for optimizing outcomes and guiding follow-up. However, current clinical methods lack the precision needed to reliably predict response and long-term prognosis.

Methods

In this feasibility study, we evaluated the prognostic utility of a novel methylation-specific droplet digital PCR (MS-ddPCR) multiplex assay in 56 patients with locally advanced rectal cancer (LARC) undergoing neoadjuvant treatment (nT)  and surgery. Circulating tumor DNA (ctDNA) was analyzed at four time points (baseline, during nT, preoperatively, 6 months post-surgery). Associations between ctDNA status and dynamics with tumor regression grade (TRG), disease recurrence, and overall survival (OS) were assessed using receiver operating characteristics (ROC) analyses and survival statistics.

Results

ctDNA was detected in 59% of the patients at baseline. Preoperative ctDNA had limited discriminative value for pathologic response, AUC 0.60 (95% CI 0.45–0.76). In contrast, ctDNA positivity 6 months postoperatively was strongly associated with recurrence within 2 years following surgery, AUC 0.96 (95% CI, 0.91–1.00). CtDNA positivity 6 months post-surgery was associated with inferior 2-year DFS (38% vs 94%, p for log-rank < 0.001) and 3-year OS (63% vs 100%, p for log-rank < 0.001).

Conclusion

With this MS-ddPCR assay, preoperative ctDNA showed limited prognostic value, whereas ctDNA 6 months postoperatively was strongly associated with recurrence and overall survival. The absence of an immediate postoperative sample limited assessment of early molecular response—a time point critical for guiding treatment decisions and follow-up strategies—underscoring the need for earlier sampling in future studies to optimize ctDNA-guided management. Given the small cohort and exploratory design, these findings are hypothesis-generating and support further validation of the assay in larger, prospective trials.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00384-025-05055-w.

Introduction

The management of rectal cancer is increasingly focusing on personalized treatment strategies, balancing escalation and de-escalation according to risk of disease recurrence [1–4]. Accurately identifying which patients are most likely to benefit from each approach is essential to optimize treatment efficacy and minimize toxicity. However, current methods lack the precision needed to reliably predict therapeutic response and long-term outcomes.

In this context, molecular biomarkers—such as circulating tumor DNA (ctDNA)—have emerged as promising candidates to refine this process. Multiple studies, including a recent systematic review, have demonstrated the prognostic value of ctDNA and its ability to detect minimal residual disease (MRD), while ongoing trials continue to explore its role in precision oncology [5–7]. Nevertheless, clinical implementation remains limited by suboptimal sensitivity and specificity, assay heterogeneity, lack of standardized cut-offs, and the absence of large-scale validation [8, 9].

Methylation-specific droplet digital PCR (MS-ddPCR) represents a promising alternative to tumor-informed methods, by targeting tumor-specific DNA methylation patterns—early and consistent epigenetic events in colorectal carcinogenesis [10, 11]. The use of multiplex assays enables simultaneous detection of multiple methylations, improving diagnostic accuracy and offering a comprehensive assessment of tumor heterogeneity [12]. These assays provide high sensitivity and specificity while remaining time- and cost-efficient, making them ideal for clinical application.

We have developed a novel MS-ddPCR multiplex assay. Preliminary data suggest that it can detect ctDNA in localized rectal cancer and that ctDNA dynamics in metastatic colorectal cancer may reflect treatment response and disease progression.

The objective of this study was to evaluate whether this assay can (1) identify poor responders to neoadjuvant therapy, (2) predict disease recurrence within two years following surgery, and (3) serve as a prognostic marker for disease-free and overall survival. We also explored whether ctDNA dynamics could enhance predictive performance. Ultimately, this feasibility study aimed to assess a clinically practical assay as a minimally invasive tool to inform treatment decisions and follow-up in rectal cancer.

Methods

Study design and patients

Patients with locally advanced rectal cancer (LARC) were enrolled in a biomarker protocol, if they were scheduled for curative preoperative radiotherapy, chemoradiotherapy, or chemotherapy followed by surgery. Exclusion criteria included distant metastasis, other malignancies diagnosed within the previous 5 years, or if they received experimental treatment 30 days prior to treatment start. Although the original protocol included a sample size calculation, recruitment was slower than anticipated, and therefore, all eligible patients enrolled at the time of analysis were included.

The study protocol was approved by The Regional Committees on Health Research Ethics for Southern Denmark (S-20160097) and conducted in accordance with the Declaration of Helsinki. All participants provided written and oral informed consent prior to inclusion.

Evaluation

Magnetic resonance imaging (MRI) was performed at baseline, for radiotherapy dose planning, during treatment, and prior to surgery. Computed tomography (CT) scans were obtained at baseline and preoperatively. Postoperative follow-up included clinical examinations and CT imaging every 6 months during the first and second year, and then annually for a total follow-up period of 5 years following surgery.

Blood sampling and DNA isolation

Blood samples were collected at baseline (pre-treatment), during treatment, pre-operatively (approximately 3 months after inclusion) and at every follow-up visit thereafter. Samples were drawn into EDTA tubes and processed within 4 h. Tubes were centrifuged at 2000g for 10 min and the isolated plasma was stored at −80 °C until use. The second centrifugation step at 10,000g for 10 min was performed at the day of DNA extraction. For quality control (QC) of purification efficiency, ~ 36,000 copies of the exogenous control cysteine-rich polycomb-like protein 1 (CPP1) were added to each plasma sample before DNA isolation [13]. Cell-free DNA (cfDNA) was isolated from 4 ml plasma using the QIAsymphony DSP circulating DNA kit on the QIAsymphony SP instrument (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The DNA was eluted in 60 µl QIAsymphony Plasma Elution Buffer and 140 µl H2O.

Droplet digital PCR

Pre-analytical quality control (QC) was performed by four assays in multiplex using a single well per sample. Briefly, purification efficiency was assessed as the recovered fraction of CPP1 after DNA isolation [13]. Lymphocyte DNA contamination was estimated using an assay targeting the VDJ rearranged IGH locus specific for B cells (PBC), and a PBC/cfDNA ratio > 2% was considered above the limit[13]. Total cfDNA concentration and contamination with high molecular weight (HMW) DNA were assessed by quantifying a short (65 bp) and a long (250 bp) fragment of the EMC7 (ER membrane protein complex subunit 7) gene, with a long/short ratio of > 0.7 deeming a sample unquantifiable[14].

For the methylation analysis, cfDNA was concentrated to a final volume of 20 µl using Amicon Ultra 30 K 0.5 ml columns (Merck, Darmstadt, Germany). The cfDNA was subsequently bisulfite-converted (BSC) with the EZ DNA Methylation-Lightning kit (Zymo Research, Irvine, CA, USA), according to the manufacturer’s instructions, using a 150 µl reaction subsequently eluted in 15 µl elution buffer. Positive controls, including universal human methylated control DNA (Zymo Research, Irvine, CA, USA), and negative controls comprising genomic DNA from whole blood and water, were converted alongside the samples.

A methylation-specific ddPCR multiplex assay targeting five DNA biomarkers—three tissue-conserved (FRMD4B, GDAP1L1, CELF2) and two cancer-specific (NPY, CLIP4)—along with the reference gene ALB, was used to analyze bisulfite-converted (BSC) DNA. Droplet generation was performed using the Automated Droplet Generator (Bio-Rad), and PCR amplification was conducted in duplicate using a Veriti™ Thermal Cycler (Applied Biosystems, Waltham, MA, USA). Details of the primer sequences are provided in Supplementary Table 1. Droplets were read on the QX600 Droplet Reader and data was analyzed using the QXManager Software 2.2 (Bio-Rad). The limit of blank (LoB) for each marker was defined using cfDNA samples from a control cohort of healthy donors (N = 60). Samples were classified as ctDNA positive if the number of observed droplets was above the LoB (NPY = 4, CLIP4 = 1, CELF2 = 2, GDAP1L1 = 2, FRMD3B = 3) for more than one marker (≥ 2), except when NPY and GDAP1L1 were the only positive markers.

The analyses of ctDNA dynamics were based on quantitative changes in each marker’s ratio (in relation to the reference gene ALB) and the Poisson distribution 95% confidence interval (CI), as described by Jacobsen et al. [15]. A significant increase (I) or decrease (D) in ctDNA fraction was defined by non-overlapping 95% CI between two time points. Overlapping intervals were classified as stable values (SV). Changes in ctDNA status (e.g., both samples positive = PP, both samples negative = NN, conversion from negative to positive = NP, or positive to negative = PN) were combined with ctDNA dynamics to categorize patients into two groups: those showing evidence of increasing ctDNA levels (NPSV, NPI, PPI) and those showing stable or decreasing levels (NNSV, PNSV, PND, PPD, PPSV). Samples were excluded from the dynamics analysis if ctDNA was not available at both time points.

Statistical analysis and data visualization

Patient characteristics were presented as frequencies with percentages (%) for categorical variables and as medians with interquartile range (IQR) for continuous variables.

Circulating tumor DNA was analyzed both as a binary variable (positive vs. negative) and as a dynamic variable based on changes in ctDNA fraction, as previously described. ctDNA dynamics were assessed from baseline to the preoperative time point in association with pathological response, and from the preoperative to the six-month postoperative time point in association with disease recurrence within 2 years.

The discriminative performance of ctDNA for pathological treatment response and disease recurrence was assessed using receiver operating characteristic (ROC) analysis. To assess the robustness of the findings, bootstrap resampling was performed. Discriminative ability was quantified by the area under the ROC curve (AUC), and comparisons between models were conducted using the DeLong test. Pathologic response was categorized as either a good response (tumor regression grade (TRG) 1 or 2) or a poor response (TRG 3–5). Disease recurrence was defined as either local relapse or distant metastasis occurring within 18 months following the post-surgery blood sample (equivalent to 2 years following surgery). Patients who experienced an event before surgery or before the relevant sampling time point were excluded from the analyses.

Time-to-event outcomes were analyzed with a cut-off date of November 1, 2024. Disease-free survival (DFS) was defined as the time from the 6-month postoperative ctDNA time-point to the first event (local recurrence, distant metastasis, or death) within 2 years. Overall survival (OS) was defined as death from any cause within 3 years from the ctDNA sampling time point. Kaplan–Meier curves were used to estimate survival probabilities, and differences between ctDNA-positive and ctDNA-negative groups were assessed using the log-rank test.

Reporting and analyses were performed in accordance with the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) guidelines where relevant (Supplementary Table 2). Statistical analyses were performed using Stata version 18.0 (StataCorp, College Station, TX, USA).

Results

Patient characteristics

Between September 2016 and November 2021, 57 patients were enrolled in the study. One patient was excluded from ctDNA analysis due to insufficient sample quality, resulting in a final evaluable cohort of 56 patients. The cohort consisted of 57% males and 43% females, with a median age of 67 years (IQR 61–73). All patients presented with locally advanced disease except for one patient, diagnosed with T1N0 (Table 1).

Table 1.

Descriptive characteristics

	N = 56
Sex, n (%)
Female	24 (43%)
Male	32 (57%)
Age, median (IQR)	67 (61–73)
cT category, n (%)
1	1 (2%)
2	1 (2%)
3	40 (71%)
4	14 (25%)
cN category, n (%)
0	5 (9%)
1	19 (34%)
2	32 (57%)
Treatment
Capox	6 (11%)
Intensified RT	4 (7%)
Standard nCRT	45 (80%)
Standard RT	1 (2%)
Tumor regression grade (Mandard), n (%)
TRG 1	12 (21%)
TRG 2	5 (9%)
TRG 3	19 (34%)
TRG 4	15 (27%)
TRG 5	0 (0%)
Missing	5 (9%)

cT, clinical Tumor; cN, clinical node; nCRT, neoadjuvant chemoradiotherapy; RT, radiotherapy

Neoadjuvant treatment consisted of standard chemoradiotherapy (50.4 Gy in 28 fractions with concomitant capecitabine at 1650 mg/m²) in 45 patients (80%). Among the remaining patients, six received four cycles of neoadjuvant CAPOX (capecitabine 825 mg/m² × 2 daily and oxaliplatin 130 mg/m² every 3 weeks), and one patient received radiotherapy only, 50.4 Gy in 28 fractions. Four patients received intensified radiotherapy: two were treated off-protocol - with a watch-and-wait approach - as they did not meet inclusion criteria for the organ preserving protocol at the time, and two had locally advanced tumors with invading adjacent organs and received intensified treatment to both the primary tumor and elective lymph nodes (Table 1). Data on treatment dose reductions were not available. None of the patient received adjuvant chemotherapy.

Surgical resection was performed in 51 patients (91%). Based on tumor regression grade (TRG), 17 (33%) patients were classified as good responders (TRG 1–2), and 34 (67%) as poor responders (TRG 3–5). One patient underwent surgery following treatment for metastases identified on the preoperative CT scan. Of the five patients who did not undergo surgery, three opted for an organ-preserving strategy, one initiated systemic therapy after being diagnosed with metastases preoperatively, and one was lost to follow-up during neoadjuvant treatment.

Plasma samples for ctDNA analysis were available in 53 patients (95%) at baseline, 53 (95%) during neoadjuvant treatment, 42 (75%) preoperatively, and 43 (77%) 6 months postoperatively (Fig. 1 & Supplementary Table 2). Complete serial samples across all four time-points were available in 35 patients (66%). Among the 21 patients with incomplete sampling, 10 were missing samples from more than one time point (Fig. 2).

Fig. 1.

Illustration of time-points for ctDNA analysis, neoadjuvant treatment and surgery. The fraction of patients who are ctDNA+ and ctDNA− is showed at each of the time-points. Median time from sampling to treatment start/end, and before and after surgery are available from the Supplementary data (Supplementary Table 2). ctDNA, circulating tumor DNA. cT, clinical Tumor; cN, clinical node; nCRT, neoadjuvant chemoradiotherapy; RT, radiotherapy. Created in https://Biorender.com

Fig. 2.

Swimmers plot showing individual patients (N = 55) ctDNA status at the different blood collection time-points (baseline, during treatment, pre-operative, 6-months postoperative). Light gray color indicates additional blood sampling after the 6-month postoperative time-point in false negatives/positives. A Patients with disease recurrence at any time point from baseline (n = 16) (B) Patients without recurrence (n = 39). ctDNA, circulating tumor DNA

ctDNA for response assessment and identification of early recurrence

Preoperative ctDNA status (n = 42) showed limited ability in distinguishing between good and poor responders (TRG 1–2 vs. TRG 3–5), AUC of 0.60 (95% CI, 0.45–0.76). When assessing ctDNA dynamics between baseline and the preoperative time point (n = 38), discriminatory performance decreased further, AUC 0.52 (95% CI, 0.39–0.65). A direct comparison between preoperative ctDNA status and ctDNA dynamics (n = 38) revealed no improvement in predictive accuracy.

While preoperative ctDNA did not effectively differentiate between good and poor responders, it is noteworthy that only 1 of the 9 patients with available preoperative samples who achieved pathological complete response (TRG 1) following nCRT was ctDNA positive before surgery. This patient was also ctDNA positive 6 months postoperatively but was later confirmed as a false positive and became ctDNA negative. In addition, one other patient became ctDNA positive at 6 months and subsequently developed metastatic disease.

We next assessed whether ctDNA positivity 6 months postoperatively was associated with disease recurrence within 2 years of surgery. This analysis (n = 42) yielded an AUC of 0.96 (95% CI, 0.91–1.00), with a sensitivity of 100%, specificity of 92%, and an overall classification accuracy of 93%. When analyzing ctDNA dynamics from the preoperative to postoperative time point (n = 35), the AUC reached 1.00 (95% CI, 1.00–1.00), with 100% sensitivity, specificity, and classification accuracy. Comparison between the two models revealed no statistically significant difference. Bootstrap resampling confirmed the robustness of these findings, with no meaningful variation across iterations.

Postoperative ctDNA and its association with recurrence and DFS

After a median follow-up of 60 months (IQR 48–60 months) following surgery, 13 of 50 patients (26%) developed recurrence: three (6%) had isolated local regrowth, seven (14%) had distant metastases, and three (6%) presented with both. One of three patients managed non-operatively developed local recurrence and metastases.

Disease-free survival after 2 years of follow-up from the 6-month postoperative time-point was 38% (95% CI 0.09–0.67) in ctDNA-positive patients versus 94% (95% CI 0.78–0.99) in ctDNA-negative patients (log-rank p < 0.001; Fig. 3).

Fig. 3.

Kaplan–Meier plot showing A DFS stratified by ctDNA six months after surgery B OS stratified by ctDNA six months after surgery. Patients were censored at the time of last follow-up if no event occurred. DFS, disease-free survival; OS, overall survival; ctDNA, circulating tumor DNA

Of the 13 evaluable patients with recurrence, 10 had ctDNA data available at the 6-month postoperative time-point. Among these, ctDNA positive had a median time to recurrence of 3.9 months (IQR, 1.6–11.0 months), compared to 31.6 months (IQR, 20.3–42.4 months) in ctDNA-negative patients. The patient who developed recurrence following non-operative management had no available sample at this time-point.

Prognostic value of ctDNA

By the cut-off date (November 2024), ten patients (20%) out of 51 patients who underwent surgery had died. Of these, nine had experienced disease recurrence. Seven patients with recurrence remained alive at the last follow-up.

Three years following the 6-month post-operative sampling time-point (3.5 years after surgery), OS was 100% in ctDNA-negative patients versus 63% (95% CI 0.23–0.86) in ctDNA-positive patients (log-rank p < 0.001; Fig. 3). No significant differences in 5-year OS were observed at baseline, during neoadjuvant treatment or preoperatively (Supplementary Fig. 1A–C).

Exploratory analyses of postoperative ctDNA in false-negative and false-positive patients

Three patients were ctDNA positive 6 months postoperatively without developing disease recurrence during follow-up. Of these, two became ctDNA negative at subsequent blood draws—one at 11 months and the other at 6 months after the initial postoperative sample. The third patient remained ctDNA positive at 3 months after the initial positive result but converted to negative 12 months postoperatively (Fig. 2).

Five patients were ctDNA negative at the 6-month postoperative time-point but subsequently experienced disease recurrence. One patient developed a local recurrence and had a ctDNA-positive sample 1 month prior to detection; a previous sample, taken 16 months earlier, was negative. Another patient developed bone and lung metastases; the only sample available—which was taken 46 months before recurrence—was negative. Among the three patients who developed liver metastases, one was ctDNA positive 13 months prior to diagnosis; another was positive 1 month before recurrence but was ctDNA negative 13 months prior; and the third was ctDNA negative 8 months before disease detection.

Discussion

In this study, we evaluated the feasibility and prognostic potential of a novel MS-ddPCR multiplex assay in patients with LARC. Preoperative ctDNA demonstrated limited ability to distinguish between good and poor responders, with an AUC of 0.60 (95% CI, 0.45–0.76), consistent with findings reported by Alden et al. [16]. Although only a few studies have provided detailed measures of sensitivity and specificity, many have reported inconsistent associations between ctDNA status and treatment response. These discrepancies likely reflect heterogeneity in assay platforms, definitions of response (e.g., cCR, pCR, or TRG), and differences in sample timing. For example, Zhou et al. observed that preoperative ctDNA negativity was associated with pCR and favorable response (CAP 0–1), whereas Khakoo et al. found no consistent association between post-nCRT ctDNA and MRI-based TRG [17, 18]. Notably, Zhou et al. used a tissue-agnostic sequencing-based approach, whereas Khakoo et al. monitored ctDNA using ddPCR informed by the highest variant allele frequency in the tumor. These differences underscore the challenges of comparing results across studies and emphasize the need for standardized response criteria, and harmonized blood-sampling protocols to clarify the role of ctDNA at this important time-point.

We did not investigate the association of ctDNA with pCR, as only nine patients had available preoperative blood samples. Of these, all were ctDNA negative except for one patient, who became ctDNA negative following surgery. Although based on a small cohort, a recent abstract by Kagawa et al. suggests a potential association between pCR and ctDNA status after total neoadjuvant treatment [19]. Given that nearly one-third of patients will develop regrowth following curative-intent chemoradiation, identifying true complete responders through a blood-based biomarker—like ctDNA—could significantly improve patient selection for non-operative management.

We evaluated ctDNA dynamics from baseline to the preoperative time-point, and performance did not improve. While ctDNA dynamics have demonstrated value in metastatic cancer—identifying poor responders earlier[20]—their significance in localized rectal cancer has been uncertain. Part of this uncertainty may stem from inconsistent definitions of ctDNA dynamics across studies. Standardized frameworks—such as the proposed ctDNA-RECIST criteria and broader methodological standardization initiatives—are essential to enable reproducible and clinically meaningful use of ctDNA across institutions [21, 22].

Importantly, our study demonstrated that ctDNA detection six months after surgery was associated with disease recurrence. These findings are in line with previous research and support the role of ctDNA as a non-invasive biomarker for early identification of patients with high risk of recurrence. Although ctDNA dynamics modestly improved classification accuracy, the observations remain uncertain and will require further investigation.

Positivity of ctDNA 6 months post-surgery was strongly associated with inferior OS, consistent with prior studies [17, 23]. In contrast, ctDNA measured at earlier time points was not associated with OS in our cohort. This difference may be explained by several factors with the most important being lack of statistical power.

Despite the strong prognostic value of postoperative ctDNA, a subset of patients remained ctDNA positive without developing recurrence. This may reflect biological noise near the detection limit or persistent cfDNA release from post-treatment tissue remodeling. Indeed, radiation-induced changes can persist for months [24], potentially contributing to prolonged cfDNA release. Interestingly, we observed that false positivity was primarily driven by tissue-specific methylations.

In contrast, all patients who developed late recurrences (> 2 years following surgery) were ctDNA negative at the 6-month postoperative time-point—an expected finding given previous studies reporting a median interval of approximately 10 months between ctDNA detection and radiologic confirmation of recurrence [25]. Of the four patients who developed metastases, three eventually became ctDNA-positive before disease detection.

ctDNA detection rates may vary by metastatic site, being lower in patients with lung or peritoneal involvement compared to those with liver metastases [26]. This was unlikely a contributing factor in our cohort, as nearly all late relapses (patients who were ctDNA negative at 6 months post-surgery) involved liver metastases. Interestingly, two of the three patients who were ctDNA positive at the 6-month blood draw and developed metastases had lung involvement. These findings suggest that tumor burden and growth dynamics, rather than metastatic site alone, may have played a more significant role. Studies in colorectal cancer have shown that ctDNA levels correlate with tumor size, proliferation rate, and overall tumor burden [27, 28].

This study has several limitations. First, the small sample size limited statistical power and precluded meaningful subgroup analyses. Missing data across time points limited statistical power further and the absence of a post-surgical blood sample restricted our ability to evaluate ctDNA during this important phase. Second, the use of a tumor-agnostic assay with limited number of targets may have reduced analytical sensitivity compared to tumor-informed methods. Third, bisulfite conversion likely contributed to ctDNA loss, decreasing the amount of detectable ctDNA. Finally, the single-center design and absence of a dedicated power calculation may limit both the interpretation and generalizability of the findings. External validation in larger, multicenter cohorts will therefore be essential to confirm these proof-of-concept observations.

Despite these limitations, the study benefits from long follow-up, enabling meaningful survival analyses, and from the use of a clinically feasible MS-ddPCR multiplex assay that is cost- and time-effective and scalable for routine practice. In addition, serial ctDNA sampling at four time points provided valuable insights into ctDNA dynamics throughout treatment and follow-up.

In this prospective LARC cohort, ctDNA detected using the MS-ddPCR assay showed poor performance in separating good from poor responders following nCRT. However, ctDNA detection 6 months postoperatively was associated with both increased risk of recurrence and reduced overall survival—underscoring its potential utility for postoperative risk stratification. Although these findings require validation in large, well-designed prospective trials, they are consistent with previous research and suggest that a tumor-agnostic, methylation-based multiplex assay may support risk stratification in LARC.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Edina D. Lauridsen, Luisa Matos Do Canto, and Signe Timm. The first draft of the manuscript was written by Edina D. Lauridsen and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported entirely by in-house resources from the department, without external funding. All ddPCR consumables and related costs were covered by the departmental budget.

Data availability

The data supporting the findings of this study are not publicly available due to ethical and legal restrictions related to participant confidentiality. Data may be made available upon reasonable request to the corresponding author, subject to approval by the relevant institutional review board and compliance with applicable data-sharing agreements.

Declarations

Ethics approval

The study protocol was approved by The Regional Committees on Health Research Ethics for Southern Denmark (S-20160097) and carried out in accordance with the Declaration of Helsinki.

Consent to participate

All participants provided written and oral informed consent prior to inclusion.

Competing interests

The authors declare no competing interests.