Care delivery among patients with early-onset versus average-onset cholangiocarcinoma: a nationwide retrospective analysis Authors: US Grewal, A Nottke, BT Loeffler, SJ Concors, NJ Hornstein, TJ Brown, A Lamarca, JW Valle Published: 2026 ABSTRACT AbstractBackgroundCholangiocarcinoma (CCA) is an aggressive biliary malignancy with increasing incidence and poor prognosis. Although traditionally considered a disease of older adults, early-onset CCA (EOCCA; age <50 years) may represent a clinically distinct subgroup. Data evaluating diagnostic and treatment trajectories in EOCCA remain limited.Patients and methodsUsing the Komodo Healthcare Map claims database (2017-2024), patients with CCA were identified using International Classification of Diseases, 10th Revision, codes and categorized as having EOCCA or average-onset CCA. Outcomes included time from first diagnosed symptom to CCA diagnosis, diagnostic delay (≥30 days), time to first intervention, and type of first intervention. Multivariable logistic, log-gamma, and competing-risk regression models evaluated associations between demographic and insurance factors and care delivery outcomes.ResultsAmong 116 106 patients with CCA, 5146 (4.4%) had EOCCA. Of 43 137 patients with symptoms documented in their claims data, EOCCA was independently associated with longer time to diagnosis [expβ 1.13, 95% confidence interval (CI) 1.07-1.20, P < 0.001] and higher odds of diagnostic delay of ≥30 days [odds ratio (OR) 1.24, 95% CI 1.12-1.36, P < 0.001]. Overall time to first intervention did not differ significantly between cohorts; however, patients with EOCCA were more likely to receive cancer-directed therapy (OR 1.12, 95% CI 1.01-1.25). Within the EOCCA cohort, longer diagnostic intervals reduced the likelihood of intervention, and disparities in treatment receipt were observed among racial minorities and publicly insured patients.ConclusionsThe EOCCA cohort represents a clinically vulnerable population characterized by significant diagnostic delays and sociodemographic disparities in care delivery. Earlier recognition and system-level interventions to improve access and streamline diagnostic pathways are needed to optimize outcomes in younger patients with CCA.Key words: biliary tract cancers, cholangiocarcinoma, diagnostic delays, disparities, early-onset, treatment delays Introduction Cholangiocarcinoma (CCA) is an aggressive malignancy that originates in the biliary tract epithelium and has an increasing incidence worldwide, largely driven by the increase in intrahepatic CCA.1,2 Owing to the aggressive nature of CCA, most patients are diagnosed with unresectable or advanced disease.3 Despite therapeutic advancements in immunotherapy and targeted therapies, the prognosis among patients with CCA remains poor.4 Although generally considered a disease of older adults, emerging reports highlight that early-onset CCA (EOCCA), defined as diagnosis before the age of 50 years, may represent a clinically distinct subgroup with unique molecular characteristics and disease biology.5, 6, 7, 8 Data on other gastrointestinal cancers indicate that clinically relevant differences between younger and older adults may extend beyond differences in disease biology and response to therapies. For example, patients with early-onset colorectal cancer may experience significant delays in diagnosis and initiation of therapeutic interventions.9,10 Atypical symptom presentation, delayed recognition, and differential access to treatment are some potential causes hypothesized to be associated with these delays. However, similar data describing diagnostic and treatment delays in EOCCA remain unavailable. Timely diagnosis and intervention are critical in CCA because patients diagnosed with resectable disease may have the opportunity to undergo treatment (surgical resection and adjuvant chemotherapy) with curative intent.11,12 However, among patients with advanced disease, therapeutic options are limited to palliative intent systemic therapy, wherein most patients die of their disease ∼12-13 months after diagnosis.13,14 Limited existing data have shown that patients with CCA may experience delays in diagnosis and treatment initiation, which may also impact outcomes.15,16 However, large-scale population-level data analyzing delays in diagnosis and treatment of CCA are lacking. Such analyses are particularly important for younger patients, in whom lack of awareness, unexpectedness (of malignancy), misdiagnosis, underinsurance, and heterogeneous care access may disproportionately influence outcomes. In this study, we leveraged the Komodo Healthcare Map (Komodo Health, Inc., New York, NY), a national longitudinal claims dataset, to characterize diagnostic and treatment trajectories in EOCCA compared with average-onset CCA (AOCCA). We examined patterns of symptom presentation, time from symptom onset to diagnosis, time to first intervention (TTFI), and the influence of demographic and insurance-related factors on these outcomes. Patients and Methods Data for this study were obtained from the Komodo Health Sentinel database (Komodo Healthcare Map), a real-world dataset that integrates disparate sources of de-identified, patient-level claims data to construct longitudinal health care journeys. Komodo aggregates data from clearinghouse, payer (≥150 payers), and provider sources, enabling longitudinal follow-up across health care settings. The dataset supports robust claims-based analyses with national coverage, large sample size, and extended follow-up periods, allowing comprehensive characterization of patient care trajectories. Using Komodo Healthcare Map claims data from 2017 through 2024, patients with a diagnosis of CCA were identified using International Classification of Diseases, 10th Revision (ICD-10) codes C22.1, C24.0, C24.8, and C24.9 with either one inpatient claim or two non-inpatient claims on different days. Patients were classified as having EOCCA (age <50 years) or AOCCA (age ≥50 years) on the basis of age at diagnosis. ICD-10 codes corresponding to common CCA-related symptoms were used to determine the earliest documented symptom and to estimate time intervals from first symptom onset to diagnosis and from diagnosis to first intervention (Supplementary Table S1, available at https://doi.org/10.1016/j.esmogo.2026.100405). Interventions were identified from procedural claims and categorized as surgery, systemic therapy, or palliative care. Cancer-directed therapy was operationally defined as undergoing surgery and/or receiving systemic anticancer therapy, whereas palliative interventions referred to symptom-directed supportive procedures or care measures. Patients with missing key demographic or coverage information including gender, race/ethnicity, rurality, and insurance status were excluded. Symptom-related outcomes included the presence of a documented first symptom, the time (days) from first diagnosed symptom to CCA diagnosis, and the occurrence of a diagnostic delay, defined as ≥30 days from symptom onset to diagnosis. Treatment-related outcomes included TTFI within 6 months of diagnosis, the type of first intervention (cancer-directed versus palliative), and the continuous time from diagnosis to first intervention. TTFI was calculated from the date of diagnosis to the earliest of first intervention or 6 months, whichever occurred first. Patients were censored at their date of insurance coverage loss. Death was treated as a competing risk. For analyses of time from first symptom to diagnosis, wait times of zero were replaced with a random uniform value between 0 and 0.1 days to enable log-gamma modeling. Of note, the duration of 30 days was chosen as a metric on the basis of a prior analysis showing that patients with a similarly aggressive gastrointestinal malignancy such as pancreatic cancer take a median of <30 days to start treatment after first contact with a physician.17 Baseline characteristics were summarized using descriptive statistics and chi-square tests were used to compare categorical variables across groups. Logistic regression was used to assess the association between demographic characteristics and the odds of (i) presenting with a first symptom, (ii) experiencing a diagnostic delay of ≥30 days, and (iii) receiving a cancer-directed versus palliative first intervention. Among symptomatic patients, log-gamma regression was used to model the continuous time from symptom onset to diagnosis with results expressed as exponentiated beta coefficients and corresponding 95% confidence intervals (CIs). For treatment-related analyses, Fine–Gray competing-risk regression estimated subdistribution hazard ratios (HRs) and 95% CIs for the likelihood of receiving a first intervention within 6 months and log-gamma regression was used to evaluate predictors of time from diagnosis to intervention. Univariable analyses were conducted for all covariates and variables with P < 0.05 were included in multivariable models. All tests were two-sided with statistical significance set at P < 0.05. Analyses were carried out using R version 4.2.1 (R Core Team; Vienna, Austria) (https://www.R-project.org/). Results We included 116 106 patients with CCA, of whom 5146 (4.4%) had EOCCA and 110 960 (95.6%) had AOCCA. Baseline characteristics are presented in Supplementary Table S2 (available at https://doi.org/10.1016/j.esmogo.2026.100405). Overall, 43 137 patients presented with a documented first CCA-related symptom, of the 43 137 symptoms, EOCCA accounted for 2046 (4.7%) and AOCCA for 41 091 (95.3%). Symptomatic presentation was slightly more frequent in EOCCA than in AOCCA [2046 of 5146 (39.8%) versus 41 091 of 110 960 (37.0%)]; however, after adjustment for demographic and insurance factors, diagnosis timing (EOCCA versus AOCCA) was not independently associated with the odds of presenting with a first symptom [odds ratio (OR) 0.98, 95% CI 0.92-1.04, P = 0.48]. Overall, the median time from first CCA symptom to diagnosis was 39 days [interquartile range (IQR) 8-134 days] and 24 055 of 43 137 patients (55.8%) experienced a delay of ≥30 days. Patients with EOCCA had longer diagnostic intervals than patients with AOCCA [median 51 days (IQR 14-155 days) versus 39 days (IQR 7-133 days)] with similar variability in the distribution [mean ± standard deviation (SD) 95.1 ± 104.1 versus 85.2 ± 101.9 days] (Figure 1). After multivariable adjustment for gender, race/ethnicity, and insurance status, EOCCA remained significantly associated with longer time to diagnosis, corresponding to a 13% increase in the mean diagnostic interval relative to AOCCA (expβ 1.13, 95% CI 1.07-1.20, P < 0.001) (Table 1). When diagnostic delay was dichotomized as ≥30 versus <30 days, EOCCA was again associated with a higher likelihood of prolonged delay. In the multivariable model, EOCCA remained an independent predictor of delay of ≥30 days with 24% increased odds of a delay of ≥30 days relative to AOCCA (OR 1.24, 95% CI 1.12-1.36, P < 0.001) (Table 2). Distribution of diagnostic and treatment initiation intervals in EOCCA versus AOCCA. AOCCA, average-onset cholangiocarcinoma; CCA, cholangiocarcinoma; EOCCA, early-onset cholangiocarcinoma. Table 1Multivariable analysis of diagnostic delays in patients with cholangiocarcinoma (n = 43 137)CharacteristicPatientsexpβ95% CIDiagnosis timing43 137 AOCCA41 091Ref EOCCA20461.131.07-1.20Gender43 137 Male23 051Ref Female20 0861.041.02-1.06Race/ethnicity43 137 White28 127Ref Asian or Pacific Islander26710.940.90-0.99 Black or African American49421.071.03-1.11 Hispanic or Latino58461.030.99-1.06 Other15511.000.94-1.06Insurance at diagnosis43 137 Commercial8433Ref Medicaid34901.061.01-1.11 Medicare31 2141.061.02-1.09AOCCA, average-onset cholangiocarcinoma; CI, confidence interval; EOCCA, early-onset cholangiocarcinoma; Ref, reference. Multivariable analysis of diagnostic delays in patients with cholangiocarcinoma (n = 43 137) AOCCA, average-onset cholangiocarcinoma; CI, confidence interval; EOCCA, early-onset cholangiocarcinoma; Ref, reference. Table 2Multivariable analysis of diagnostic delays of ≥30 days among patients with cholangiocarcinoma (n = 43 137)CharacteristicPatientsOR95% CIDiagnosis timing43 137 AOCCA41 091Ref EOCCA20461.241.12-1.36Gender43 137 Male23 051Ref Female20 0861.091.05-1.13Race/ethnicity43 137 White28 127Ref Asian or Pacific Islander26710.870.80-0.94 Black or African American49421.050.99-1.12 Hispanic or Latino58461.010.96-1.07 Other15511.010.91-1.12Insurance at diagnosis43 137 Commercial8433Ref Medicaid34901.030.95-1.11 Medicare31 2140.960.91-1.01AOCCA, average-onset cholangiocarcinoma; CI, confidence interval; EOCCA, early-onset cholangiocarcinoma; OR, odds ratio; Ref, reference. Multivariable analysis of diagnostic delays of ≥30 days among patients with cholangiocarcinoma (n = 43 137) AOCCA, average-onset cholangiocarcinoma; CI, confidence interval; EOCCA, early-onset cholangiocarcinoma; OR, odds ratio; Ref, reference. Among patients with EOCCA who presented with a CCA-related symptom, upon multivariable analysis, gender (P = 0.002) and insurance type (P = 0.006) remained independently associated with time to diagnosis. Female patients with EOCCA experienced a 16% increase in diagnostic interval duration (expβ 1.16, 95% CI 1.06-1.28) compared with male patients. Medicaid and Medicare patients experienced 12% (expβ 1.12, 95% CI 1.01-1.25) and 27% (expβ 1.27, 95% CI 1.07-1.51) longer times to diagnosis, respectively, compared with commercially insured patients (Table 3). Evaluation of binary diagnostic delay (≥30 days versus <30 days) demonstrated that only gender was significantly associated with prolonged delay upon univariable logistic regression (P < 0.001). Female patients with EOCCA had 39% higher odds of experiencing a diagnostic delay of ≥30 days compared with male patients (OR 1.39, 95% CI 1.16-1.66). Race/ethnicity, rurality, and insurance were not statistically significant predictors of a ≥30-day diagnostic delay (Table 4). Table 3Multivariable analysis of diagnostic delays in patients with early-onset cholangiocarcinoma (n = 2046)CharacteristicPatientsexpβ95% CIGender2046 Male1071Ref Female9751.161.06-1.28Insurance at diagnosis2046 Commercial1196Ref Medicaid6671.121.01-1.25 Medicare1831.271.07-1.51CI, confidence interval; Ref, reference. Multivariable analysis of diagnostic delays in patients with early-onset cholangiocarcinoma (n = 2046) CI, confidence interval; Ref, reference. Table 4Univariable analysis of sociodemographic and insurance factors associated with diagnostic delays ≥30 days in early-onset cholangiocarcinoma (n = 2046)CharacteristicPatientsOR95% CIGender2046 Male1071Ref Female9751.391.16-1.66Race/ethnicity2046 White990Ref Asian or Pacific Islander1590.960.68-1.35 Black or African American3171.090.84-1.42 Hispanic or Latino4661.130.90-1.43 Other1141.440.96-2.20Rurality2046 Urban916Ref Both10821.070.89-1.28 Rural481.440.78-2.76Insurance at diagnosis2046 Commercial1196Ref Medicaid6671.180.97-1.43 Medicare1831.190.86-1.65CI, confidence interval; OR, odds ratio; Ref, reference. Univariable analysis of sociodemographic and insurance factors associated with diagnostic delays ≥30 days in early-onset cholangiocarcinoma (n = 2046) CI, confidence interval; OR, odds ratio; Ref, reference. Among 102 743 patients with CCA and adequate follow-up for intervention analyses, 4643 (4.5%) had EOCCA and 98 100 (95.5%) had AOCCA. Overall, just over half of the patients received a first intervention within 6 months of diagnosis (n = 52 444, 51.0%), whereas 12 452 (12.1%) died without receiving an intervention and 37 847 (36.8%) were censored. In univariable Cox regression, EOCCA was associated with a higher likelihood of receiving any intervention within 6 months compared with AOCCA (HR 1.28, 95% CI 1.23-1.32). However, after adjustment for gender, race/ethnicity, rurality, and insurance, cohort status was no longer independently associated with TTFI (EOCCA: HR 1.03, 95% CI 0.99-1.07, P = 0.14). Among the 52 444 patients who did receive an intervention, the median time from diagnosis to first intervention was 0.9 months (IQR 0.4-1.8 months) and did not differ meaningfully between AOCCA and EOCCA (median 0.9 versus 0.8 months; mean 1.3 months in both cohorts). Systemic therapy was the most common first intervention (n = 30 937, 59.0%), followed by palliative care (n = 13 055, 24.9%) and surgery (n = 8452, 16.1%). When categorized as cancer-directed (surgery or systemic therapy) versus palliative care, 39 389 (75.1%) received a cancer-directed first intervention. In multivariable logistic regression, EOCCA remained independently associated with greater odds of a cancer-directed first intervention compared with AOCCA (OR 1.12, 95% CI 1.01-1.25, P = 0.03). In Cox regression restricted to EOCCA, insurance status (P < 0.001) and the interval from symptom onset to diagnosis (P < 0.001) were significantly associated with TTFI within 6 months. Patients with Medicaid (HR 0.84, 95% CI 0.77-0.91, P < 0.001) and Medicare (HR 0.62, 95% CI 0.54-0.70, P < 0.001) coverage were significantly associated with decreased likelihood of receiving a first intervention compared with commercially insured patients. Each additional month from first CCA symptom to diagnosis was associated with a 3% decrease in the likelihood of receiving an intervention (HR 0.97, 95% CI 0.95-0.99, P < 0.001). In multivariable analysis of time from diagnosis to first intervention among treated patients with EOCCA, gender and insurance remained significant. Female patients experienced a 9% shorter mean time to intervention compared with male patients (expβ 0.91, 95% CI 0.85-0.98, P = 0.016). In contrast, Medicaid and Medicare coverage were associated with longer TTFI relative to commercial insurance: Medicaid with a 16% increase (expβ 1.16, 95% CI 1.07-1.26, P < 0.001) and Medicare with an 18% increase (expβ 1.18, 95% CI 1.04-1.35, P = 0.010) (Table 5). Upon multivariable logistic regression, black patients with EOCCA had 30% lower odds of receiving a cancer-directed (versus palliative) first intervention compared with white patients (OR 0.70, 95% CI 0.53-0.91, P = 0.008). Hispanic/Latino patients showed a similar trend toward lower odds (OR 0.78, 95% CI 0.61-1.00, P = 0.050). Relative to commercially insured patients with EOCCA, those with Medicaid (OR 0.50, 95% CI 0.41-0.62) and Medicare (OR 0.46, 95% CI 0.34-0.63) were also less likely to receive a cancer-directed first intervention (P < 0.001 for both) (Table 6). Table 5Multivariable analysis of factors associated with time from diagnosis to first intervention in early-onset cholangiocarcinoma (n = 2744)CharacteristicPatientsexpβ95% CIGender2744 Male1309Ref Female14350.910.85-0.98Insurance at diagnosis2744 Commercial1680Ref Medicaid8051.161.07-1.26 Medicare2591.181.04-1.35CI, confidence interval; Ref, reference. Multivariable analysis of factors associated with time from diagnosis to first intervention in early-onset cholangiocarcinoma (n = 2744) CI, confidence interval; Ref, reference. Table 6Multivariable analysis of factors associated with receipt of cancer-directed versus symptom-directed care in early-onset cholangiocarcinoma (n = 2744)CharacteristicPatientsOR95% CIP valueRace/ethnicity27440.044 White1437Ref Asian or Pacific Islander1851.100.73-1.70 Black or African American4230.700.53-0.91 Hispanic or Latino5490.780.61-1.00 Other1500.910.59-1.45Insurance at diagnosis2744<0.001 Commercial1680Ref Medicaid8050.500.41-0.62 Medicare2590.460.34-0.63CI, confidence interval; OR, odds ratio; Ref, reference. Multivariable analysis of factors associated with receipt of cancer-directed versus symptom-directed care in early-onset cholangiocarcinoma (n = 2744) CI, confidence interval; OR, odds ratio; Ref, reference. Discussion To our knowledge, this represents the first and largest analysis investigating care delivery among patients with EOCCA. In our large, population-based study, we showed that EOCCA is associated with significant differences in diagnostic and treatment trajectories compared with AOCCA. Patients with EOCCA experienced significantly longer intervals from first symptom to diagnosis and higher odds of diagnostic delays of ≥30 days. Although younger patients who reached the treatment pathway were more likely to receive cancer-directed rather than palliative first interventions, significant disparities within EOCCA were also noted. There is a growing body of evidence that highlights the increase in incidence of CCA among younger adults. Existing studies have demonstrated that patients with EOCCA are more likely to present with more advanced disease compared with AOCCA.5, 6, 7 In our study, in addition to a significantly longer median time from symptom onset to diagnosis, delays of ≥30 days were also more common in the EOCCA cohort. Considering the aggressive clinical course of CCA, these data propose the possibility that diagnostic delays may at least in part explain the relatively more advanced disease at diagnosis. Although patients in the EOCCA cohort in our study were more likely to present with symptoms before diagnosis, several factors may explain diagnostic delays. There is a common lack of access to primary care among young adults and an inherent diagnostic bias that propagates the notion that cancers including gastrointestinal cancers are primarily a disease of older individuals.18,19 Our finding that female sex was independently associated with longer diagnostic intervals in EOCCA is particularly concerning and aligns with broader literature showing sex-based disparities in diagnostic workup across cancer types, even among younger adults.20, 21, 22, 23 Diagnostic delays in women may be driven by a complex interplay of patient-, provider-, and system-level factors including symptom normalization, fear and denial, and competing caregiving or work responsibilities.20 Our findings merit further investigation into drivers of diagnostic delays among younger women with CCA. Additionally, among patients with EOCCA, both Medicaid and Medicare coverage were independently associated with significantly longer times to diagnosis. These data reflect multiple structural barriers including limited access to subspecialty care that contribute to delayed diagnosis and adversely affect outcomes among patients with CCA who are uninsured or covered by less comprehensive insurance plans, such as Medicaid.24,25 Notably, although patients in the EOCCA cohort experienced diagnostic delays, we did not identify differences in time to intervention after diagnosis when compared with patients with AOCCA. In fact, patients in the EOCCA cohort were more likely to receive a cancer-directed first intervention compared with patients in the AOCCA cohort, a finding that has consistently been noted in prior investigations analyzing treatment patterns and outcomes in EOCCA.5, 6, 7, 8 We further noted that diagnostic delays in EOCCA were significantly and negatively associated with the timing of cancer-directed therapy (longer diagnostic delays were associated with delayed initiation of interventions). These data underscore the need to amplify efforts targeting timely diagnosis of CCA among younger patients to help facilitate prompt institution of cancer-directed interventions. Similar to patterns observed with diagnostic delays, patients in the EOCCA cohort who were females or had Medicaid/Medicare coverage experienced longer delays in initiation of therapeutic interventions. However, it is important to acknowledge that reduced likelihood of intervention among patients with EOCCA with delayed diagnosis may reflect multiple overlapping factors including more advanced disease stage at presentation, impaired functional status, limited access to specialized care, socioeconomic barriers, or other unmeasured confounders not captured in claims data. Our data also uncovered some key racial disparities in receipt of cancer-directed interventions among patients with EOCCA. Black patients were significantly less likely than white patients to receive cancer-directed first-line therapy and Hispanic/Latino patients showed a similar trend. These data mirror prior findings suggesting relatively lower odds of curative resection and inferior long-term outcomes among patients with CCA representing racial and ethnic minorities.26,27 Thus, although younger age may increase the likelihood of treatment, this advantage does not appear equitably distributed across racial and ethnic subgroups. This study has limitations inherent to claims-based analyses including reliance on administrative coding for diagnoses, procedures, and symptoms, which may be subject to misclassification or coding variability. The claims-based nature of the Komodo Healthcare Map dataset limits access to detailed clinical variables including disease stage, resectability, tumor burden, laboratory values, radiographic findings, molecular profiling, and performance status. Therefore, these findings should be interpreted as associations in care delivery patterns rather than definitive evidence of causative disparities. The observed differences between EOCCA and AOCCA may reflect a combination of health care access, symptom recognition, disease biology, and unmeasured clinical confounders that cannot be fully disentangled within administrative claims data. However, the large sample size, national coverage, and longitudinal structure of the Komodo dataset enable evaluation of real-world care patterns that are not readily captured in single-institution or clinical trial cohorts, which is a unique strength of our current analysis. Notably, although we noted important inequities based on insurance coverage, it is important to highlight that Medicare coverage among patients aged <50 years likely represents a medically complex subgroup with disability-related eligibility or significant comorbidity burden. We also acknowledge that the time period of 30 days has not been prospectively validated in patients with CCA as a standard metric to study diagnostic and treatment delays. Further work with built-in sensitivity analyses is needed to validate these findings. In conclusion, our large, population-based analysis highlights that patients with EOCCA represent a clinically vulnerable subgroup of patients. Younger patients with CCA should be recognized as a high-risk population for delayed diagnosis, warranting greater clinical vigilance and a higher diagnostic index of suspicion. Delayed diagnosis in this population was particularly pronounced in females and those with Medicaid/Medicare coverage and was shown to result in delayed initiation of treatment interventions. These data underscore the need for system-level interventions to streamline diagnostic pathways for suspected biliary malignancies including standardized referral algorithms, expedited imaging protocols, and early multidisciplinary evaluation, especially among younger individuals and under-represented populations. Further prospective studies are needed to validate these findings and uncover modifiable factors driving diagnostic delays and treatment disparities among patients with EOCCA. Additionally, future studies linking diagnostic and treatment delays with clinically meaningful oncologic outcomes are also needed. Acknowledgments Data Availability Data are not publicly available. Data used for the analysis may be made available upon appropriate request. Funding None declared. Disclosure The authors have declared no conflicts of interest. Supplementary data Supplementary Material References Associated Data Supplementary Materials Supplementary Material Data Availability Statement Data are not publicly available. Data used for the analysis may be made available upon appropriate request.