INTRODUCTION

Cancer registries serve as indispensable resources for evaluating cancer burden, formulating control policies, and assessing the impact of healthcare advancements and policy interventions. Their utility hinges on the reliability and completeness of the data. In Korea, comprehensive nationwide cancer statistics have been reported since 1999 [1], achieving high levels of completeness in documenting cancer severity information, such as the Surveillance, Epidemiology, and End Results (SEER) Summary Stage.

The tumor-node-metastasis (TNM) staging system, the most extensively utilized clinical tool for cancer staging [2], provides detailed information on tumor size, lymph node involvement, and metastasis and is a critical prognostic indicator. Integrating TNM staging into cancer registries could substantially improve the assessment of cancer screening effectiveness and treatment outcomes, supplying essential data for policy evaluation in screening, treatment, and care. In contrast, the SEER Summary Stage facilitates comparisons of disease severity across different cancer types based on anatomical features [3], but lacks the detailed resolution offered by TNM staging. Although periodic revisions of the TNM system enhance its clinical utility by incorporating diagnostic and treatment advances, these updates also present challenges. Meanwhile, the SEER Summary Stage offers a more consistent framework yet does not provide the depth of insight achieved with TNM staging.

Multiple methods have been employed to enhance the completeness of cancer severity data. For global comparisons [4-6], TNM variables are often used to generate SEER Summary Stage data. However, the inverse process of imputing TNM variables based on SEER Summary Stage information has been underexplored, despite some investigations into imputation strategies using other clinical variables and data sources [7-9].

To bridge this gap, we evaluated the impact of augmenting TNM variables with SEER Summary Stage data from the Korean Cancer Registry. We focused on 5 major cancer types—stomach, colorectal, liver, lung, and breast—and analyzed how these supplementation strategies influenced 5‐year relative survival rates.

METHODS

Data Source

Our study used data from the Korea National Cancer Incidence Database (KNCID) for 5 major cancer types (liver, colon, stomach, breast, and lung) diagnosed in Korea between 2012 and 2017 [1]. The KNCID, a population‐based cancer registry, collects data from the Korea Central Cancer Registry (KCCR) and 11 regional cancer registries. By incorporating cases not captured in hospital‐based registries, the KNCID achieved an impressive completeness rate of 98.2% in 2021 [1].

Initially, the database contained 173 084 stomach, 159 245 colorectal, 89 693 liver, 137 169 lung, and 110 292 female breast cancer cases. After excluding patients with uncertain diagnosis dates or in situ cancer, the final study population included 173 061 stomach, 159 199 colorectal, 89 639 liver, 137 103 lung, and 110 286 female breast cancer patients. Collection of SEER Summary Stage data is mandatory, whereas TNM information is gathered on an optional basis.

Tumor-node-metastasis Stage Classification

Using the seventh edition of the TNM manual [2], we converted the T, N, and M classifications into TNM stages (I, II, III, and IV) following site‐specific guidelines. For cases lacking TNM information, we directly assigned a predetermined stage using reverse logic that began with the most advanced tumor progression. Cases without classifiable TNM data were designated as “missing.”

Surveillance, Epidemiology, and End Results Summary Stage

The SEER Summary Stage categorizes cancer severity into several stages: in situ, localized, regional by direct extension only, regional with lymph node involvement only, regional with both direct extension and lymph node involvement, regional not otherwise specified (NOS), distant, and unknown.

Supplementation of Missing Tumor-node-metastasis Stage Information

We applied assumptions for missing T, N, and M values and devised two methods to determine the TNM stage. The first method defined the stage solely using available TNM data. The second method supplemented missing TNM components with SEER Summary Stage data, following approaches adopted by the International Cancer Benchmarking Partnership to improve stage data comparability and completeness across international registries [4-6]. Detailed assumptions for each cancer type and the corresponding supplementation strategies are provided in Supplemental Material 1.

Statistical Analysis

We calculated the case counts and proportions with available TNM stage data for the approaches. For each cancer site, we estimated 5‐year relative survival rates for all cases diagnosed during the study period up to 2022, including those with missing stages. Relative survival rates and their 95% confidence intervals (CIs) were computed using the Ederer II method via the ‘relsurv’ package in R [10]. The survival calculations incorporated age (in single-year increments), sex, and year of cancer registration as covariates. To evaluate the effect of supplementation on survival estimates, we compared 5‐year relative survival rates before and after supplementation for each TNM stage and cancer type. Statistical analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, NC, USA) and R version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria).

Ethics Statement

This study was reviewed and approved by the Institutional Review Board (IRB) of the National Cancer Center (IRB No. NCC 2024-0142). The data were deidentified, and the consent requirements were waived by the IRB.

RESULTS

Table 1 presents the distribution of TNM staging variables and SEER Summary Stage data across the 5 cancer types. Missing data varied considerably among cancer types, with liver cancer exhibiting the highest rates of missing TNM data (90.7% for T stage, 93.8% for N stage, and 93.5% for M stage). In contrast, breast cancer had the lowest missing rates (55.3% for T stage, 57.0% for N stage, and 65.5% for M stage). SEER Summary Stage data were substantially more complete, with unknown stages ranging from 3.0% in breast cancer to 14.1% in liver cancer—significantly lower than the missing rates for TNM variables.

Table 2 shows the effects of supplementing missing TNM staging variables using SEER Summary Stage data. The improvement in TNM completeness varied considerably by cancer type. Stomach cancer exhibited the greatest reduction in missing TNM data, decreasing from 73.8% to 23.2% after supplementation, with most previously missing cases reassigned to stage I (a 47.0 percentage point [%p] increase). Liver cancer also showed notable improvement, with missing proportions declining from 93.0% to 71.5%, primarily due to reassignment to stage IV (an 18.3%p increase). Breast cancer experienced a moderate reduction, with missing proportions decreasing from 65.1% to 51.5%. In contrast, lung and colorectal cancers showed relatively modest improvements, with reductions of 5.1%p and 4.5%p, respectively.

Table 3 compares the 5‐year relative survival rates across cancer types before and after supplementation of missing TNM stage variables using SEER Summary Stage data. Despite the substantial reassignment of missing TNM cases in stomach cancer, the changes in survival rates were modest, with the largest being a 3.6%p decrease for stage I (from 106.1 to 102.5%). Liver cancer demonstrated significant changes, particularly for stage IV, where survival rates dropped markedly from 17.7% to 7.9%, and for stage III, which increased by approximately 4%p. Additionally, survival rates for cases with missing TNM stages in liver cancer increased by 6.6%p, from 38.5% to 45.1%. Breast cancer exhibited a notable decrease in stage IV survival rates (from 47.8 to 43.5%) and a slight increase for missing stages (from 94.2 to 96.9%). Lung and colorectal cancers, which underwent minimal TNM reassignment, exhibited minor survival rate changes of less than 1%p across all stages.

DISCUSSION

This study represents the first nationwide analysis of TNM stage-specific survival rates using a Korean Cancer Registry. Our findings underscore that supplementing TNM data with SEER Summary Stage information improves data completeness while introducing only minimal changes to survival estimates.

These results align with previous studies suggesting that supplemented stage data can approximate the survival trends observed in directly staged cohorts [7-9]. However, it is important to acknowledge that those studies focused predominantly on prostate cancer, thereby limiting their generalizability. This narrow focus underscores the need for cancer-specific supplementation strategies, as the impact of supplementation varies significantly among cancer types. Expanding the range of variables collected by cancer registries to improve supplementation accuracy poses a considerable challenge; the additional burden on registry resources and personnel must be weighed against the benefits of obtaining more comprehensive data. Therefore, it is crucial to develop targeted approaches for acquiring missing information [11-13], that prioritize the most critical variables for each cancer type while minimizing the burden on registry systems.

Our evaluation of TNM supplementation using SEER Summary Stage data is situated within broader efforts to enhance cancer registry data quality. A recent study by Engholm et al. [14] examined how various assumptions for individual TNM components affected overall TNM staging using Nordic cancer registry data. Although our focus was on supplementing missing TNM information to improve data completeness, Engholm et al. [14] investigated the impact of different hierarchical rules and assumptions on TNM stage distribution. Their findings, which generally showed minor differences in stage distribution, are consistent with our observations that supplementation resulted in relatively modest changes in stage-specific survival estimates for several cancer types. Our approach—leveraging SEER Summary Stage data to supplement missing TNM information—offers a pragmatic solution when direct TNM data are unavailable, a common challenge in many registries.

Despite the improvements from supplementation, some survival estimates remain unstable or potentially biased. For instance, relative survival rates for stage I stomach, colorectal, and breast cancers exceeding 100% may reflect limitations of the Ederer II method used for estimation. A previous study in Sweden indicated that achieving stable relative survival estimates requires follow-up periods of at least 7 years [15]. Additionally, differences in patient characteristics and socioeconomic status, which can influence early cancer detection, might also affect outcomes [16-18]. Further research is necessary to determine the significance of these factors.

Liver cancer emerged as a notable exception, with supplementation-induced changes in 5‐year survival rates being considerably larger—particularly for stage IV, where there was a 9.8%p difference. This discrepancy is accentuated by the fact that only 7% of liver cancer cases had sufficient TNM data for staging before supplementation, compared with over 25% for other cancers. The extreme scarcity of TNM data in liver cancer suggests that relative survival rates derived solely from available data may be inaccurate. The low collection rate of TNM variables for liver cancer appears to be linked to its treatment landscape. Other registries have similarly reported high rates of missing data for liver cancer [19,20]. Unlike many cancers where surgical resection is common, liver cancer management often involves non-surgical treatments such as radiofrequency ablation or transarterial chemoembolization [21]. This reduced reliance on surgical procedures leads to fewer pathological reports, which are critical for staging in cancer registries. Moreover, the complex relationship between liver cancer and underlying conditions such as cirrhosis further complicates staging and treatment decisions [21-23].

Our study had several limitations that merit attention. First, our supplementation process did not distinguish between clinical and pathological TNM staging. Although this limitation may not have significantly influenced our results—a sensitivity analysis by Engholm et al. [14] demonstrated minimal differences when comparing the prioritization of pTNM and the inclusion of cTNM—the high completeness in reporting histologic types by the KCCR [1], suggests that pathology report information substantially contributed to determining both TNM and SEER Summary Stages in our study. Second, our analysis did not account for variations across histological types. This limitation arises from our primary focus on assessing the feasibility of supplementing TNM information, rather than exploring differences among histological variants. As such, our goal was to demonstrate the general potential of this approach, not to develop cancer-specific strategies. Future research should aim to develop and validate supplementation strategies tailored to specific cancer types. We anticipate that subsequent studies employing comprehensive staging surveys [24-26], which will include TNM information, SEER Summary Stage, and treatment data, will facilitate a more detailed evaluation of supplementation strategies for individual cancers. Third, we were unable to statistically evaluate the equivalence of relative survival rates before and after supplementation due to the lack of established statistical methods and benchmarks in the literature. Instead, consistent with previous studies [4-6,14], we assessed only the magnitude of the observed changes in survival rates. Fourth, our study has an additional limitation regarding the reliability and consistency of directly collected TNM stages compared to those derived via supplementation. Specifically, TNM stage data lacking detailed T, N, or M components were used only when supplementation through SEER Summary Stage or other sources was not feasible. However, we prioritized SEER Summary Stage data, as it is mandatorily collected by KCCR, assuming it to be more reliable than directly collected TNM stages. Consequently, further validation studies are needed to systematically evaluate the reliability of TNM stages generated through reverse logic based on TNM variables relative to those collected directly.

In conclusion, enhancing TNM stage completeness using the SEER Summary Stage appears feasible. Improved completeness of TNM staging and the resulting clinically relevant statistics may contribute to more effective treatment planning and informed policy formulation. Future research should consider incorporating additional data sources to evaluate broader impacts on cancer outcomes.

Supplemental Materials

Notes

Conflict of Interest

The authors have no conflicts of interest associated with the material presented in this paper.

Funding

This study was supported by the Intramural Research Program of the National Cancer Center Korea (NCC-2332570), and other research grants from the National Cancer Center Korea (NCC-2211820 and NCC-2410700).

Acknowledgements

None.

Author Contributions

Conceptualization: Choi CK, Yun EH. Data curation: Jung KW, Yun EH. Formal analysis: Choi CK, Yun EH. Funding acquisition: Choi CK, Yun EH. Writing – original draft: Choi CK, Yun EH. Writing – review & editing: Choi CK, Suh M, Jung KW, Yun EH.

REFERENCES

• 1. Park EH, Jung KW, Park NJ, Kang MJ, Yun EH, Kim HJ, et al. Cancer statistics in Korea: incidence, mortality, survival, and prevalence in 2021. Cancer Res Treat 2024;56(2):357-371. https://doi.org/10.4143/crt.2024.253ArticlePubMedPMC
• 2. Edge SB, Compton CC. The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM. Ann Surg Oncol 2010;17(6):1471-1474. https://doi.org/10.1245/s10434-010-0985-4ArticlePubMed
• 3. National Cancer Institute. SEER summary staging manual 2000 (code - C62610); 2020 [cited 2024 Jul 27]. Available from: https://evsexplore.semantics.cancer.gov/evsexplore/concept/ncit/C62610
• 4. Arnold M, Morgan E, Bardot A, Rutherford MJ, Ferlay J, Little A, et al. International variation in oesophageal and gastric cancer survival 2012-2014: differences by histological subtype and stage at diagnosis (an ICBP SURVMARK-2 population-based study). Gut 2022;71(8):1532-1543. https://doi.org/10.1136/gutjnl-2021-325266ArticlePubMed
• 5. Araghi M, Arnold M, Rutherford MJ, Guren MG, Cabasag CJ, Bardot A, et al. Colon and rectal cancer survival in seven high-income countries 2010-2014: variation by age and stage at diagnosis (the ICBP SURVMARK-2 project). Gut 2021;70(1):114-126. https://doi.org/10.1136/gutjnl-2020-320625ArticlePubMed
• 6. Walters S, Maringe C, Butler J, Brierley JD, Rachet B, Coleman MP. Comparability of stage data in cancer registries in six countries: lessons from the International Cancer Benchmarking Partnership. Int J Cancer 2013;132(3):676-685. https://doi.org/10.1002/ijc.27651ArticlePubMed
• 7. Westerberg M, Beckmann K, Gedeborg R, Irenaeus S, Holmberg L, Garmo H, et al. Choice of imputation method for missing metastatic status affected estimates of metastatic prostate cancer incidence. J Clin Epidemiol 2023;155: 22-30. https://doi.org/10.1016/j.jclinepi.2022.12.008ArticlePubMed
• 8. Luo Q, Egger S, Yu XQ, Smith DP, O’Connell DL. Validity of using multiple imputation for “unknown” stage at diagnosis in population-based cancer registry data. PLoS One 2017;12(6):e0180033. https://doi.org/10.1371/journal.pone.0180033ArticlePubMedPMC
• 9. Saito E, Hori M, Matsuda T, Yoneoka D, Ito Y, Katanoda K. Long-term trends in prostate cancer incidence by stage at diagnosis in Japan using the multiple imputation approach, 1993-2014. Cancer Epidemiol Biomarkers Prev 2020;29(6):1222-1228. https://doi.org/10.1158/1055-9965.EPI-19-1228ArticlePubMed
• 10. Perme MP, Pavlic K. Nonparametric relative survival analysis with the R package relsurv. J Stat Softw 2018;87(8):1-27. https://doi.org/10.18637/jss.v087.i08Article
• 11. Merrill RM, Sloan A, Anderson AE, Ryker K. Unstaged cancer in the United States: a population-based study. BMC Cancer 2011;11: 402. https://doi.org/10.1186/1471-2407-11-402ArticlePubMedPMC
• 12. Gurney J, Sarfati D, Stanley J, Dennett E, Johnson C, Koea J, et al. Unstaged cancer in a population-based registry: prevalence, predictors and patient prognosis. Cancer Epidemiol 2013;37(4):498-504. https://doi.org/10.1016/j.canep.2013.03.005ArticlePubMed
• 13. Howlader N, Noone AM, Yu M, Cronin KA. Use of imputed population-based cancer registry data as a method of accounting for missing information: application to estrogen receptor status for breast cancer. Am J Epidemiol 2012;176(4):347-356. https://doi.org/10.1093/aje/kwr512ArticlePubMedPMC
• 14. Engholm G, Lundberg FE, Kønig SM, Ólafsdóttir E, Johannesen TB, Pettersson D, et al. Influence of various assumptions for the individual TNM components on the TNM stage using Nordic cancer registry data. Acta Oncol 2023;62(3):215-222. https://doi.org/10.1080/0284186X.2023.2189528ArticlePubMed
• 15. Talbäck M, Rosén M, Stenbeck M, Dickman PW. Cancer patient survival in Sweden at the beginning of the third millennium —predictions using period analysis. Cancer Causes Control 2004;15(9):967-976. https://doi.org/10.1007/s10522-004-2475-1ArticlePubMed
• 16. Jun JK, Choi KS, Lee HY, Suh M, Park B, Song SH, et al. Effectiveness of the Korean National Cancer Screening Program in reducing gastric cancer mortality. Gastroenterology 2017;152(6):1319-1328. https://doi.org/10.1053/j.gastro.2017.01.029ArticlePubMed
• 17. Luu XQ, Lee K, Jun JK, Suh M, Jung KW, Choi IJ, et al. Risk of upper gastrointestinal cancer and death in persons with negative screening results: results from the National Cancer Screening Program in South Korea. Gastric Cancer 2023;26(4):580-589. https://doi.org/10.1007/s10120-023-01387-0ArticlePubMed
• 18. Lee K, Suh M, Jun JK, Choi KS. Impact of the COVID-19 pandemic on gastric cancer screening in South Korea: results from the Korean National Cancer Screening Survey (2017-2021). J Gastric Cancer 2022;22(4):264-272. https://doi.org/10.5230/jgc.2022.22.e36ArticlePubMedPMC
• 19. Engholm G, Lundberg FE, Kønig SM, Ólafsdóttir E, Johannesen TB, Pettersson D, et al. TNM stage in the Nordic Cancer Registries 2004-2016: registration and availability. Acta Oncol 2024;63: 303-312. https://doi.org/10.2340/1651-226X.2024.35232ArticlePubMedPMC
• 20. Barchuk A, Tursun-Zade R, Belayev A, Moore M, Komarov Y, Moshina N, et al. Comparability and validity of cancer registry data in the northwest of Russia. Acta Oncol 2021;60(10):1264-1271. https://doi.org/10.1080/0284186X.2021.1967443ArticlePubMed
• 21. Korean Liver Cancer Association (KLCA) and National Cancer Center (NCC) Korea. 2022 KLCA-NCC Korea practice guidelines for the management of hepatocellular carcinoma. Clin Mol Hepatol 2022;28(4):583-705. https://doi.org/10.3350/cmh.2022.0294ArticlePubMedPMC
• 22. Pons F, Varela M, Llovet JM. Staging systems in hepatocellular carcinoma. HPB (Oxford) 2005;7(1):35-41. https://doi.org/10.1080/13651820410024058ArticlePubMedPMC
• 23. Marsh JW, Dvorchik I, Bonham CA, Iwatsuki S. Is the pathologic TNM staging system for patients with hepatoma predictive of outcome? Cancer 2000;88(3):538-543. https://doi.org/10.1002/(sici)1097-0142(20000201)88:3<538::aid-cncr7>3.0.co;2-hArticlePubMed
• 24. Kim J, Hong S, Lee JJ, Won YJ, Lee ES, Kang HS, et al. Analysis of the tumor characteristics in young age breast cancer patients using collaborative stage data of the Korea Central Cancer Registry. Breast Cancer Res Treat 2021;187(3):785-792. https://doi.org/10.1007/s10549-021-06107-9ArticlePubMed
• 25. Oh CM, Kong HJ, Kim E, Kim H, Jung KW, Park S, et al. National Epidemiologic Survey of Thyroid cancer (NEST) in Korea. Epidemiol Health 2018;40: e2018052. https://doi.org/10.4178/epih.e2018052ArticlePubMedPMC
• 26. Shin W, Ham TY, Park YR, Lim MC, Won YJ. Comparing survival outcomes for cervical cancer based on the 2014 and 2018 International Federation of Gynecology and Obstetrics staging systems. Sci Rep 2021;11(1):6988. https://doi.org/10.1038/s41598-021-86283-2ArticlePubMedPMC

Figure & Data

References

Citations

Citations to this article as recorded by