Residential location shapes initial prostate cancer care in Korea (2010–2020): A nationwide stage- and risk-stratified study

Abstract

Purpose

To characterize nationwide patterns of initial prostate cancer (PCa) management in Korea from 2010 to 2020, stratified by clinical stage, risk, and residential context.

Materials and Methods

We performed a secondary analysis of a chart-abstracted cohort from 51 general hospitals. Initial treatment was classified into ten mutually exclusive categories (active surveillance; surgery; surgery after androgen deprivation therapy [ADT]; radiotherapy; radiotherapy after ADT; docetaxel after ADT; androgen receptor–targeted agent [ARTA] after ADT; ADT-only ≥1 year; other active treatment; unknown). Localized disease (T1–T4N0M0) was stratified by risk (low, intermediate, and high). Residential context was coded as capital versus non-capital and urban versus rural. National Comprehensive Cancer Network risk stratification (version 2.2024) was applied uniformly across survey years.

Results

Among 27,075 men, initial management differed significantly by stage, risk, and residence. In localized disease, surgery predominated, active surveillance was most frequent in low-risk disease, and radiotherapy and ADT-only increased with higher risk (p<0.001). Over time, ADT-only decreased in metastatic disease, with increases in radiotherapy and systemic intensification (p=0.004). Capital and urban residence were associated with greater use of surgery in localized disease and systemic intensification in metastatic disease, whereas non-capital and rural residence relied more on ADT-only (all p<0.001). In multivariable analyses, residential context remained independently associated with treatment selection after adjustment for age, prostate-specific antigen, grade, and stage.

Conclusions

Initial PCa management in Korea during the 2010s reflected contemporary adoption of surgical and systemic therapies but demonstrated persistent geographic variation, indicating residential context influences real-world access to guideline-concordant care.

Graphical Abstract

INTRODUCTION

Prostate cancer (PCa) is a major male malignancy globally, and its incidence has risen steadily in South Korea over the past two decades with demographic aging and wider prostate-specific antigen (PSA) testing [1]. Especially in the last decade, treatment practice shifted with the diffusion of robotic-assisted radical prostatectomy (RARP), advances in radiotherapy, and adoption of androgen receptor–targeted agents (ARTAs) [2].

Nationwide Korean reports using insurance or administrative data have described utilization trends and regional differences [2,3,4,5]; however, most analyses were derived from such datasets and therefore did not permit detailed stratification by clinical stage or biological risk—the key determinants of initial management [3,4]—and did not reliably capture contemporary modalities such as RARP and ARTAs [3,4,5]. Other Korean studies have examined selected modalities or subgroups [6,7,8,9,10], yet none has provided a nationwide view resolved simultaneously by clinical stage, biological risk, and residential context. Consequently, these studies cannot fully assess guideline-concordant, contemporary initial management.

To address these gaps, we analyzed a nationwide hospital-based cohort assembled from 51 general hospitals between 2010 and 2020, whose parent analysis described national distributions of clinicopathologic variables among 27,075 men with biopsy-confirmed disease [11]. Using this chart-abstracted dataset, which enables detailed PCa risk stratification, we examined initial management across stage- and risk-defined strata and evaluated variation by residential context (urban vs. rural; capital vs. non-capital). Leveraging real-world clinical data spanning both traditional and contemporary treatment modalities, this study offers an integrated nationwide description of how Korean men entered the PCa treatment pathway during the 2010s.

MATERIALS AND METHODS

1. Study design, setting, and oversight

This study is a secondary analysis of a retrospectively assembled, nationwide hospital-based dataset that covers six survey years (2010, 2012, 2014, 2016, 2018, 2020) [11]. The protocol originally evaluated the stage- and risk-stratification patterns at the time of the diagnosis of PCa based on the biopsy information. Fifty-one general hospitals (≥200 beds) with urology departments participated across all 17 administrative districts. Each site contributed cases diagnosed during the survey years and entered structured information into a standardized electronic case-report form (eCRF). Among the 15 items included in the eCRF, item 11 captured the initial treatment received after diagnostic biopsy, classified into the 10 predefined categories described below. Item 12 ascertained whether any local therapy was performed within one year after biopsy, and item 13 specified the type of local therapy administered (Supplementary Figs. 1, 2, 3). Trained abstractors extracted demographic, clinical, imaging, pathology, and treatment-initiation data from medical records, with built-in range and logic checks (e.g., serum PSA level, date order) applied to ensure data consistency. Because treatment data were abstracted directly from medical records rather than insurance claims, the eCRF captured actual treatments administered regardless of reimbursement status, including off-label or self-paid use of agents such as docetaxel and ARTAs. Ethical approval for this study was obtained from the Institutional Review Board of Yeungnam University Medical Center (approval number: YUMC 2022-11-017), with a waiver of informed consent for the use of de-identified data.

2. Participants and key variables

Eligible patients were men with histologically confirmed PCa diagnosed in a survey year at a participating hospital. Exclusion criteria were: (1) prior definitive local therapy before the index biopsy, (2) incomplete staging information required for the relevant analysis, and (3) indeterminate residence at diagnosis for residential-context analyses.

Clinical stage at diagnosis followed routine institutional pathways and was categorized as T1–T4N0M0 (localized), N1M0 (node-positive), or M1 (metastatic). Localized disease was stratified by National Comprehensive Cancer Network (NCCN) risk-stratification category (low, intermediate, and high) when components were available; cases lacking sufficient elements were labeled “unknown” and excluded from risk-stratified denominators. Risk stratification was applied retrospectively and uniformly to all cases across all survey years using a single contemporary definition based on the NCCN Guidelines (version 2.2024), ensuring internal consistency throughout the study period. Contemporary NCCN guideline categorized high-risk as any one of the following: PSA levels ≥20 ng/mL, clinical stage ≥cT3, and International Society of Urological Pathology (ISUP) grade group ≥4. Conversely, low-risk was defined as PSA levels <10 ng/mL, clinical stage ≤cT2a, and ISUP grade group 1. Residential context was recorded as one of 17 administrative districts and dichotomized as urban versus rural (dong vs. eup/myeon). The urban–rural classification was based on the patients’ residential address at the time of diagnosis, as recorded in the medical chart, using the Korean administrative classification of “dong” (urban) versus “eup/myeon” (rural) areas. A second dichotomy contrasted capital (Seoul, Incheon, Gyeonggi) versus non-capital regions.

Initial management was assigned to ten mutually exclusive categories applied identically across survey years: (1) active surveillance; (2) surgery; (3) surgery after androgen-deprivation therapy (ADT); (4) radiotherapy; (5) radiotherapy after ADT; (6) docetaxel after ADT; (7) ARTA after ADT; (8) ADT-only (≥1 year, no other treatment); (9) other active treatment; and (10) unknown. When modalities occurred in close succession, classification reflected the initial clinical intent (curative local therapy first; otherwise systemic-therapy schema).

3. Statistical analysis

The primary endpoint was the distribution of initial management (ten categories) within each analytic stratum defined by stage, risk, and residential context in each survey year. Results are presented as counts and within-group percentages. A two-sided α=0.05 was used, and very small p-values were reported as p<0.001. Group differences were assessed using Pearson χ² tests, excluding categories with zero total counts across comparison groups.

Exploratory univariable analyses assessed differences in patient characteristics and treatment distributions using χ² tests for categorical variables and Wilcoxon rank-sum tests for continuous variables. Multivariable logistic regression models estimated adjusted odds ratios for: (1) surgery versus non-surgical management in localized disease and (2) ADT-only versus any active treatment across stage groups. Covariates included age, PSA, Gleason grade, clinical stage, survey year, and residential context (urban vs. rural; capital vs. non-capital); NCCN risk category was additionally included for models restricted to localized disease. These models complemented the descriptive primary analyses.

This study is reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.

RESULTS

1. Overall patterns of initial management

Baseline characteristics by stage and NCCN risk are summarized in Table 1. Curative local therapies predominated in localized disease, whereas ADT-only and systemic-intensified therapy were more frequent in node-positive and metastatic PCa. Within localized PCa (T1–T4N0M0), the distribution of the four dominant treatment strategies—active surveillance, surgery±ADT, RT±ADT, and ADT-only—varied significantly across NCCN risk strata (Fig .1). Active surveillance was most common in low-risk disease, surgery±ADT in intermediate-risk disease, and both RT±ADT and ADT-only in high-risk disease (χ²=1,256.7, df=6, p<0.001).

Table 1. Urban–rural comparison across ten initial treatment categories by stage and risk.

Stage	Risk group	Region	Active surveillance	Surgery	Surgery after ADT	RT	RT after ADT	Doce after ADT	ARTA after ADT	ADT-only (≥1 y)	Other active Tx.	Unknown	p-value
T1–T4N0M0	High	Urban	62 (1.3)	2,007 (43.4)	104 (2.3)	161 (3.5)	486 (10.5)	11 (0.2)	1 (0.0)	961 (20.8)	22 (0.5)	806 (17.4)	0.066
		Rural	32 (1.7)	778 (41.8)	48 (2.6)	63 (3.4)	156 (8.4)	4 (0.2)	1 (0.1)	447 (24.0)	11 (0.6)	321 (17.2)
T1–T4N0M0	Intermediate	Urban	301 (5.0)	3,407 (56.6)	83 (1.4)	244 (4.1)	248 (4.1)	1 (0.0)	2 (0.0)	602 (10.0)	31 (0.5)	1,100 (18.3)	<0.001
		Rural	108 (4.8)	1,171 (51.5)	48 (2.1)	84 (3.7)	105 (4.6)	2 (0.1)	0 (0.0)	313 (13.8)	11 (0.5)	430 (18.9)
T1–T4N0M0	Low	Urban	171 (16.9)	553 (54.6)	5 (0.5)	29 (2.9)	18 (1.8)	1 (0.1)	0 (0.0)	47 (4.6)	15 (1.5)	174 (17.2)	0.22
		Rural	58 (22.4)	127 (49.0)	1 (0.4)	3 (1.2)	2 (0.8)	0 (0.0)	0 (0.0)	14 (5.4)	2 (0.8)	52 (20.1)
AnyT N1M0	-	Urban	6 (0.9)	187 (26.9)	22 (3.2)	19 (2.7)	109 (15.7)	7 (1.0)	3 (0.4)	257 (37.0)	2 (0.3)	82 (11.8)	0.024
		Rural	2 (0.7)	60 (20.8)	5 (1.7)	1 (0.3)	52 (18.1)	2 (0.7)	1 (0.3)	137 (47.6)	2 (0.7)	26 (9.0)
AnyT AnyN M1	-	Urban	12 (0.8)	74 (4.9)	24 (1.6)	28 (1.8)	141 (9.3)	99 (6.5)	43 (2.8)	930 (61.4)	12 (0.8)	152 (10.0)	0.726
		Rural	9 (1.1)	46 (5.5)	12 (1.4)	16 (1.9)	67 (8.0)	43 (5.2)	17 (2.0)	539 (64.6)	6 (0.7)	79 (9.5)

Values are presented as number (row %); omnibus Pearson χ² across the ten categories within each stratum.

The “Unknown” category (17%–20% of cases) primarily includes patients whose initial treatment information was not recorded in the electronic case-report form at the time of data abstraction. This may reflect patients referred to other institutions for treatment after diagnostic biopsy, those lost to follow-up before treatment initiation at the diagnosing hospital, or cases in which treatment decisions were pending at the time of data collection.

ADT, androgen deprivation therapy; RT, radiotherapy; ARTA, androgen receptor–targeted agent; Tx., treatment.

Fig. 1. Temporal trends in initial treatment distribution by disease stage across survey years (2010–2020). (A) Localized disease (T1–T4N0M0). (B) Node-positive disease (N1M0). (C) Metastatic disease (M1). Bars represent the proportion of patients receiving each initial treatment category in the corresponding year. ADT, androgen deprivation therapy; RT, radiotherapy.

2. Temporal trends in initial management (2010–2020)

Temporal patterns demonstrated stage-specific differences (Fig. 2). In localized disease, surgery±ADT remained the leading modality with modest increases in RT±ADT and a decline in ADT-only over time (p<0.001). Treatment distributions for N1M0 disease were stable (p=0.686). In metastatic PCa (M1), ADT-only decreased steadily across survey years, accompanied by rising use of RT±ADT and systemic intensification (p=0.004).

Fig. 2. Risk-stratified distribution of initial treatment in localized prostate cancer (T1–T4N0M0) according to NCCN (National Comprehensive Cancer Network) risk category (low, intermediate, and high). ADT, androgen deprivation therapy; RT, radiotherapy.

3. Residential context and treatment distribution

Initial treatment distribution differed significantly by residential context across all stages (capital vs. non-capital and urban vs. rural; all Pearson χ² tests, p<0.001) (Fig. 3). In localized disease (T1–T4N0M0), curative local therapy—particularly surgery—was more frequently used in capital and urban regions, whereas ADT-only was more common in non-capital and rural areas (e.g., non-capital vs. capital: 24.0% vs. 13.1%). Similar patterns were observed in N1M0 disease, where ADT-only predominated in non-capital regions (53.1% vs. 32.4%), while capital regions showed greater use of surgery or radiotherapy. In metastatic disease (M1), ADT-only remained the most common strategy overall but was again more frequent in non-capital areas (74.3% vs. 61.9%), whereas systemic intensification (docetaxel or ARTA) was used more often in capital regions (12.5% vs. 8.5%). Among patients receiving surgery as local therapy, RARP accounted for a significantly larger proportion in capital versus non-capital regions (59.9% vs. 41.9%) and in urban versus rural settings (53.4% vs. 41.4%), while radiotherapy constituted a greater share in non-capital and rural areas (both comparisons p<0.001) (Fig. 4).

Fig. 3. Geographic differences in initial treatment distribution by disease stage. (A) Localized disease (T1–T4N0M0). (B) Node-positive disease (N1M0). (C) Metastatic disease (M1). Comparisons are shown for capital versus non-capital regions and for urban versus rural residence. p-values were derived from Pearson χ² tests and were <0.001 for all stage-specific comparisons. ADT, androgen deprivation therapy; RT, radiotherapy.

Fig. 4. Geographic differences in the choice of local treatment modality among patients receiving local therapy. Comparisons are shown for capital versus non-capital regions and for urban versus rural residence. p-values were derived from Pearson χ² tests and were <0.001 for both regional comparisons. RP, radical prostatectomy.

4. Multivariable analyses

In multivariable logistic regression analyses, age, PSA level, Gleason grade, and clinical stage were independently associated with initial treatment selection. After adjustment for these clinical factors, residential context remained significantly associated with treatment choice, with capital versus non-capital residence showing a consistent association with receipt of surgery in non-metastatic disease (Table 2) and with systemic intensification in metastatic counterparts (Table 3), whereas urban versus rural residence demonstrated a more limited effect.

Table 2. Independent determinants of surgical selection as primary treatment in non-metastatic prostate cancer.

Variable		Reference	p-value	Exp(B)	95% CI for Exp(B)
					Lower	Upper
Urban vs. rural residence
	Age <75 y	Age ≥75 y	<0.001	4.800	4.432	5.199
	PSA <20 ng/mL	PSA ≥20 ng/mL	<0.001	1.875	1.705	2.063
	ISUP grade group <4	ISUP grade group ≥4	<0.001	1.272	1.176	1.375
	T stage <T3	cT stage ≥cT3	<0.001	0.775	0.697	0.862
	Residential area: rural	Residential area: urban	0.023	0.917	0.851	0.988
Capital vs. non-capital residence
	Age <75 y	Age ≥75 y	<0.001	4.853	4.478	5.260
	PSA <20 ng/mL	PSA ≥20 ng/mL	<0.001	1.893	1.719	2.084
	ISUP grade group <4	ISUP grade group ≥4	<0.001	1.247	1.153	1.349
	T stage <T3	cT stage ≥cT3	<0.001	0.779	0.699	0.867
	Residential area: capital	Residential area: non-capital	<0.001	1.778	1.659	1.905

Models were adjusted for age, PSA level, ISUP grade group, clinical T stage, NCCN (National Comprehensive Cancer Network) risk category, survey year, and residential context.

PSA, prostate-specific antigen; ISUP, International Society of Urological Pathology; CI, confidence interval.

Table 3. Independent determinants of systemic intensification beyond ADT as primary treatment in metastatic prostate cancer.

Variable		Reference	p-value	Exp(B)	95% CI for Exp(B)
					Lower	Upper
Urban vs. rural residence
	Age ≥75 y	Age <75 y	0.003	0.633	0.470	0.853
	PSA ≥20 ng/mL	PSA <20 ng/mL	0.051	1.718	0.998	2.956
	ISUP grade group ≥4	ISUP grade group <4	0.312	1.289	0.788	2.110
	cT stage ≥cT3	T stage <T3	0.018	1.763	1.104	2.817
	Residential area: urban	Residential area: rural	0.114	1.292	0.940	1.774
Capital vs. non-capital residence
	Age ≥75 y	Age <75 y	0.002	0.626	0.465	0.843
	PSA ≥20 ng/mL	PSA <20 ng/mL	0.048	1.734	1.005	2.993
	ISUP grade group ≥4	ISUP grade group <4	0.321	1.284	0.783	2.107
	cT stage ≥cT3	T stage <T3	0.018	1.768	1.104	2.831
	Residential area: non-capital	Residential area: capital	0.011	0.678	0.501	0.916

Models were adjusted for age, PSA level, ISUP grade group, clinical T stage, survey year, and residential context.

ADT, androgen deprivation therapy; PSA, prostate-specific antigen; ISUP, International Society of Urological Pathology; CI, confidence interval.

DISCUSSION

This nationwide hospital-based study demonstrates that residential context is associated with initial PCa management in Korea, even when treatment decisions are examined within clearly defined stage and risk strata. While variation by disease biology is expected, the present analysis shows that capital versus non-capital and urban versus rural residence remain aligned with distinct first-line treatment patterns in routine practice. Because staging and risk stratification were chart-verified rather than inferred from claims, these geographic differences are interpreted within a clinically coherent framework. Geographic variation in PCa care has been consistently reported across health systems, with residential location associated with differences in presentation and treatment patterns, underscoring that the observed associations are not unique to the Korean context [12,13].

Compared with Western countries, the treatment landscape for localized PCa in Korea during the study period was characterized by a markedly higher predominance of surgery relative to radiotherapy. In a US population-based SEER (Surveillance, Epidemiology, and End Results) analysis of 398,074 men with localized PCa (2004–2013), surgery and radiotherapy were utilized in roughly equal proportions (each approximately 38%), with 24% managed by expectant management [14]. By contrast, in our Korean cohort, surgery (±ADT) accounted for approximately 45%–55% of localized disease management while radiotherapy (±ADT) comprised only 5%–10%—representing a surgery-to-radiotherapy ratio of approximately 5–10:1 compared with the roughly 1:1 ratio observed in the United States. This disparity likely reflects a combination of factors including the relatively limited availability of radiotherapy infrastructure outside tertiary centers in Korea, reimbursement frameworks favoring surgical approaches, and differences in patient–physician shared decision-making practices between Asian and Western settings. For metastatic disease, systemic intensification with docetaxel and ARTAs was adopted earlier in Western settings following landmark trials, whereas our data show that ADT-only remained the predominant first-line strategy in Korea, particularly outside the capital region. Recent Korean consensus efforts have addressed these gaps by establishing evidence-based treatment recommendations for metastatic hormone-sensitive PCa [15].

Across the 2010s, treatment distributions observed in this cohort were broadly consistent with known secular trends in Korean PCa care, including sustained predominance of surgery in localized disease and increasing use of RT and systemic intensification in advanced disease. These patterns parallel the diffusion of RARP and the adoption of contemporary systemic agents and are concordant with prior Korean studies based on administrative data and modality-specific registries [2,3,4,6,7,10,11]. In PCa specifically, prior studies have shown that greater travel distance and treatment accessibility are associated with lower utilization of radiotherapy, supporting travel burden as a practical determinant of initial treatment selection [16]. The higher proportion of neoadjuvant ADT followed by surgery, rather than radiotherapy, observed in non-capital and rural regions may partly reflect the travel burden associated with fractionated radiotherapy courses, which typically require repeated visits over several weeks and disproportionately affect patients with limited access to radiation facilities [16,17,18]. The principal contribution of the present study is not the identification of these trends per se, but their resolution by stage, risk, and residential context at the time of initial treatment decision.

Geographic gradients were evident across disease stages. In localized disease, capital and urban residence were associated with higher use of curative local therapy, particularly surgery, whereas ADT-only accounted for a larger proportion of initial management in non-capital and rural settings. In node-positive and metastatic disease, ADT-only remained the dominant initial strategy overall but was most frequent in non-capital and rural regions, while RT and systemic intensification were more commonly used in capital regions. These associations persisted after adjustment for age, PSA, grade, and stage, indicating that residential context contributed independently to treatment selection.

Among patients receiving local therapy, further regional differences were observed in the choice of modality. RARP accounted for a larger share of surgical treatment in capital and urban settings, whereas RT constituted a greater proportion of local therapy in non-capital and rural regions. This pattern is consistent with prior evidence indicating increasing centralization of radical prostatectomy at high-volume centers alongside diffusion of robotic platforms, which can increase patient travel and shape access to surgical modalities [19]. This finding suggests that geographic variation extends beyond whether local therapy is delivered to how it is delivered, likely reflecting differential access to resource-intensive surgical platforms and facility-based treatments. Recent Korean data further demonstrate that active surveillance eligibility criteria and protocol elements also vary across institutions and regions, suggesting that geographic variation extends even to the management of low-risk disease [20].

The observed patterns are consistent with existing Korean literature describing centralization of PCa care. Studies of medical travel have shown that non-Seoul residents frequently seek definitive PCa treatment in the capital, particularly for surgery, reflecting concentration of high-volume centers and advanced technology [21,22,23]. National analyses of radiotherapy utilization have similarly documented increasing use over time alongside persistent regional variation, indicating that capacity expansion alone has not eliminated geographic differences in access [24]. Together, these reports provide context for our finding that residence correlates with the treatment actually delivered once disease severity is specified.

International evidence supports similar mechanisms. Systematic reviews and narrative syntheses across cancer sites have consistently reported lower radiotherapy utilization among patients facing greater travel distance or time to treatment facilities [17,18]. Population-level studies have highlighted persistent rural–urban disparities in cancer care, implicating structural and socioeconomic factors that influence access to timely and guideline-concordant treatment [25]. Geospatial analyses further demonstrate how facility distribution shapes real-world care pathways [26], and PCa–specific data from Australia show later presentation and longer travel intervals outside capital hubs [27]. The present findings align with this broader literature, suggesting that travel burden and facility distribution function as practical determinants of modality choice across health systems.

It should be noted that in Korea’s National Health Insurance system, reimbursement for systemic intensification agents (e.g., docetaxel and ARTAs) in node-positive (N1M0) disease was limited during the study period, as these agents were primarily approved for metastatic castration-resistant indications [15]. The low proportion of systemic intensification observed in N1M0 cases (~1%–2%) reflects these reimbursement constraints, and some cases may represent off-label use, clinical trial participation, or self-paid treatment.

Several strengths of this study warrant emphasis. The cohort includes 51 hospitals across all 17 administrative districts, capturing a broad cross-section of Korean practice. Stage and risk were chart-verified, and initial management was classified using harmonized, mutually exclusive definitions applied consistently across survey years. The repeated cross-sectional design allows temporal patterns to be interpreted without reliance on single-year snapshots.

Important limitations should be acknowledged. First, this is a hospital-based cohort rather than a population census, and participating centers were not exhaustive; however, the cohort captured approximately 37% of incident PCa cases during survey years, supporting external relevance. Second, outcome endpoints such as survival, toxicity, or quality of life were not available, limiting inference to patterns of care rather than effectiveness. Third, some stratified analyses involved small denominators, particularly when stage, risk, and residence were examined simultaneously, and these estimates should be interpreted cautiously. Fourth, detailed socioeconomic indicators, comorbidity burden, and facility-level characteristics were not uniformly captured, precluding adjustment for these factors and leaving open alternative explanations related to patient preference or local practice culture.

CONCLUSIONS

In this nationwide, stage- and risk-resolved analysis, initial management of PCa in Korea during the 2010s reflected contemporary adoption of surgical, radiotherapeutic, and systemic treatment strategies but varied substantially according to residential context. Even after accounting for disease stage and risk, capital and urban residence was associated with greater use of resource-intensive and guideline-concordant modalities, whereas non-capital and rural residence were characterized by greater reliance on ADT-only therapy, highlighting persistent geographic disparities in real-world PCa care.

SUPPLEMENTARY MATERIALS

Supplementary materials can be found via https://doi.org/10.4111/icu.20260160.

Supplementary Fig. 1

Distribution of initial treatment in localized prostate cancer (T1–T4N0M0) across 17 administrative regions. ADT, androgen deprivation therapy; RT, radiotherapy.

Supplementary Fig. 2

Distribution of initial treatment in node-positive prostate cancer (N1M0) across 17 administrative regions. ADT, androgen deprivation therapy; RT, radiotherapy.

Supplementary Fig. 3

Distribution of initial treatment in metastatic prostate cancer (M1) across 17 administrative regions. ADT, androgen deprivation therapy; RT, radiotherapy.