Meme Kanserinde Sağkalım Durumunu Etkileyen Faktörlerin İncelenmesi: Eğilim Skoru Analizi

Abstract

Amaç: Eğilim skoru (ES), incelenen değişkenlerdeki sistematik hatanın düzeltilmesi ya da ortadan kaldırılabilmesi amacıyla özellikle gözlemsel çalışmalarda kullanılan bir yöntemdir. Rosenbaum ve Rubin (1983) tarafından geliştirilen bu yöntem, bireyin ortak değişkenlere göre ilgili gruba atanmasının koşullu olasılığı olarak tanımlanır. Bu çalışmada, ES ile eşleştirme yapılarak meme kanseri nedeniyle ölümü etkileyen faktörlerin incelenmesi amaçlandı. Yöntem: Çalışmada, 423 meme kanseri hastasına ilişkin veri seti kullanıldı. Sağkalım durumu üzerine yaş, tümör büyüklüğü, radyoterapi, hormon terapisi, aksiller lenf nodu tutulumu değişkenlerinin etkisi incelendi. Veri setinde yaş değişkeni bakımından gruplar arası heterojenlik olduğu için en yakın komşu yöntemi kullanılarak eşleştirme yapıldı. Bulgular: Eşleştirme sonrası yaş değişkeninin etkisi ortadan kaldırıldı. 1:1 eşleştirme sonrası tümör büyüklüğü (p=0,009) ve aksiller pozitif lenf nodu tutulumu ≥4 (p=0,026) değişkenlerinin sağkalım durumu üzerinde anlamlı etkisinin olduğu belirlendi. 2:1 eşleştirme sonrası tümör büyüklüğü (p=0,004), radyoterapi (p=0,017) ve aksiller pozitif lenf nodu tutulumu ≥4 (p=0,001) değişkenlerinin sağkalım durumu üzerinde anlamlı etkisinin olduğu belirlendi. Sonuç: Heterojen yapıdaki verilerin doğrudan analiz edilmesi verideki gerçek etkilerin göz ardı edilmesine neden olabileceği için mutlaka ES yöntemi kullanılarak eşleştirme yapılmalıdır. Bu yöntemin en önemli dezavantajı ise eşleştirme nedeniyle veri kaybı olmasıdır. Bu nedenle çok sayıda birimle çalışılması verideki bilgi kaybının önüne geçecektir.

Keywords

• Eğilim skoru
• eşleştirme
• lojistik regresyon
• meme kanseri
• en yakın komşu yöntemi

Examination of Factors Affecting Survival Status in Breast Cancer: Propensity Score Analysis

Abstract

Objective: The propensity score (PS) is a method used especially in observational studies to correct or eliminate the systematic error in the variables studied. The method developed by Rosenbaum and Rubin (1983) is defined as the conditional probability of assigning the individual to the relevant group according to common variables. The aim of this study was to investigate the factors affecting the death of breast cancer by matching with PS. Methods: In this study, 423 breast cancer patients were used. The effects of age, tumor size, radiotherapy, hormone therapy, axillary lymph node involvement on survival were investigated. Since there was heterogeneity among the groups in terms of age variables, 1:1 and 2:1 matching was done by using the nearest neighbor method. Results: The effect of age variable after matching was eliminated. After 1:1 matching, tumor size (p=0.009) and axillary positive lymph node involvement ≥4 (p=0.026) were found to have a significant effect on survival. After 2:1 matching, tumor size (p=0.004), radiotherapy (p=0.017) and axillary positive lymph node involvement ≥4 (p=0.001) were found to have a significant effect on survival. Conclusion: Since the direct analysis of the heterogeneous structure data may lead to neglecting the actual effects in the data, it must be matched using the PS method. The most important disadvantage of this method is data loss due to matching. Therefore, working with a large number of units will prevent the loss of information in the data.

Keywords

• Propensity score
• matching
• logistic regression
• breast cancer
• nearest neighbor method

References

1. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41-55. doi: 10.1093/biomet/70.1.41
2. Kim YJ, Jung SY, Kim K. Survival benefit of radiotherapy after surgery in de novo stage IV breast cancer: a population-based propensity-score matched analysis. Scientific reports. 2019;9. doi: 10.1038/s41598-019-45016-2
3. Matsuo K, Mandelbaum RS, Machida H, et al. Decreasing secondary primary uterine cancer after breast cancer: A population-based analysis. Gynecologic oncology. 2019;154(1):169-176. doi: 10.1016/j.ygyno.2019.05.014
4. Wang J, Tang H, Li X, et al. Is surgical axillary staging necessary in women with T1 breast cancer who are treated with breast-conserving therapy? Cancer Communications. 2019;39(1):25. doi: 10.1186/s40880-019-0371-y
5. Jeon HJ, Oh J, Shin DH. Urate-lowering agents for asymptomatic hyperuricemia in stage 3–4 chronic kidney disease: Controversial role of kidney function. PloS one. 2019;14(6). doi: 10.1371/journal.pone.0218510
6. Miura S, Yamashita T, Hanyu M, Kumamaru H, Shirai S, Ando K. Propensity score-matched analysis of patients with severe aortic stenosis undergoing surgical aortic valve replacement. Open heart. 2019;6(1). doi: 10.1136/openhrt-2018-000992
7. Li X, Zhang C, Sun Z, et al. Propensity-matched analysis of adjuvant chemotherapy for completely resected Stage IB non-small-cell lung cancer patients. Lung Cancer. 2019;133:75-82. doi: 10.1016/j.lungcan.2019.04.024
8. Paek SH, Lee HA, Kwon H, Kang KH, Park SJ. Comparison of robot-assisted modified radical neck dissection using a bilateral axillary breast approach with a conventional open procedure after propensity score matching. Surgical Endoscopy. 2019:1-6. doi: 10.1007/s00464-019-06808-9

9. Leite W. Practical propensity score methods using R. Sage Publications; 2016.
10. Demir O, Dolgun A, Etikan İ, Kuyucu YE, Saraçbaşı O. Propensity skor ağırlıklandırma yönteminde denge metriklerinin performansı üzerine benzetim çalışması. Journal of Contemporary Medicine. 2017; 7(3):265-277. doi : 10.16899/gopctd.349948
11. Rosenbaum PR, Rubin DB. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. The American Statistician. 1985;39(1):33-38. doi: 10.1080/00031305.1985.10479383
12. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research. 2011;46(3):399-424. doi: 10.1080/00273171.2011.568786
13. d'Agostino RB. Propensity score methods for bias reduction in the comparison of a treatment to a non‐randomized control group. Statistics in Medicine. 1998;17(19):2265-2281. doi: 10.1002/(SICI)1097-0258(19981015)17:19<2265::AID-SIM918>3.0.CO;2-B
14. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Stürmer T. Variable selection for propensity score models. American Journal of Epidemiology. 2006;163(12):1149-1156. doi: 10.1093/aje/kwj149
15. Arun T, Imai K, Sinha F. Does the Microfinance Reduce Poverty in India? Propensity Score Matching based on a National-level Household Data. Economics Discussion Paper, The University of Manchester, September 2006, 9-22.
16. Gu XS, Rosenbaum PR. Comparison of multivariate matching methods: Structures, distances, and algorithms. Journal of Computational and Graphical Statistics. 1993;2(4):405-420. doi: 10.1080/10618600.1993.10474623
17. Stürmer T, Joshi M, Glynn RJ, Avorn J, Rothman KJ, Schneeweiss S. A review of the application of propensity score methods yielded increasing use, advantages in specific settings, but not substantially different estimates compared with conventional multivariable methods. Journal of Clinical Epidemiology. 2006;59(5):437. e431-437. e424. doi: 10.1016/j.jclinepi.2005.07.004
18. Hosmer Jr DW, Lemeshow S, Sturdivant RX. Applied logistic regression. Vol 398: John Wiley & Sons; 2013.
19. Olmos A, Govindasamy P. Propensity scores: a practical introduction using R. Journal of MultiDisciplinary Evaluation. 2015;11(25):68-88.
20. Caliendo M, Kopeinig S. Some practical guidance for the implementation of propensity score matching: Discussion paper series 1588. The Institute for the Study of Labour (IZA), Bon, Germany. 2005. doi: 10.1111/j.1467-6419.2007.00527.x
21. Amusa LB. Reducing bias in Observational Studies: An Empirical Comparison of Propensity Score Matching Methods. Turkiye Klinikleri Journal of Biostatistics. 2018, 10(1),14-26. doi: 10.5336/biostatic.2017-58633
22. Stuart EA. Matching methods for causal inference: A review and a look forward. Statistical science: a review Journal of the Institute of Mathematical Statistics. 2010;25(1):1. doi: 10.1214/09-STS313
23. Omurlu IK, Ozdamar K, Ture M. Comparison of Bayesian survival analysis and Cox regression analysis in simulated and breast cancer data sets. Expert Systems With Applications. 2009;36(8):11341-11346. doi: 10.1016/j.eswa.2009.03.058
24. Zhang Z. Propensity score method: a non-parametric technique to reduce model dependence. Annals of Translational Medicine. 2017;5(1). doi: 10.21037/atm.2016.08.57