Beyond a Simple Survival Curve: How Better Statistics Can Reveal More About Lung Cancer

MAKASSAR, January 14, 2026/STOCHASTIC When researchers study how long patients survive after a cancer diagnosis, the data are rarely complete. Some patients are still alive when the study ends, some leave the study, and others can no longer be followed. Their exact survival time is unknown, but their information is still valuable. Statistics calls this situation “right censoring.”

A study titled “Poisson-Exponential Distribution Approach to Survival Analysis of Right-Censored Data of Lung Cancer Patients” offers a way to handle this challenge. The article was published in the Jurnal Matematika, Statistika, dan Komputasi by Departemen Matematika in January 2026. It shows how a mathematical model can turn incomplete observations into a clearer picture of survival patterns.

The researchers analyzed secondary data from 48 lung cancer patients treated at Dr. Wahidin Sudirohusodo Hospital in Makassar in 2015. Of these patients, 36 had right-censored records, meaning that death was not observed during their follow-up period. The remaining 12 records contained observed deaths. Ignoring the censored cases would waste most of the available information and could distort the results.

To solve this problem, the researchers used the Poisson-Exponential distribution, a flexible model designed to describe survival when risk may change over time. They estimated the model using two methods: Maximum Likelihood Estimation (MLE), a widely used statistical approach, and Maximum Product Spacing (MPS), an alternative that can work well with small or censored datasets.

Both versions produced survival curves close to the Kaplan-Meier curve, a standard benchmark based directly on observed data. However, MPS followed the patient data more closely, especially during the middle and later months of observation. It also achieved better scores across several statistical measures used to compare models. In simple terms, MPS provided the better balance between accuracy and model efficiency.

Using the fitted models, the estimated probability of surviving beyond 12 months was approximately 76 percent, declining to around 63 percent beyond 20 months. These figures describe this particular historical dataset; they should not be interpreted as a personal prognosis or a current survival rate for all lung cancer patients. Individual outcomes can differ because of disease stage, age, tumor type, treatment, general health, and many other factors.

This work contributes directly to Sustainable Development Goal 3 (SDG 3): Good Health and Well-Being. Reliable survival analysis can help researchers evaluate treatment outcomes, identify changing risks, plan health services, and use limited clinical data more responsibly. Better statistics cannot treat cancer, but they can improve the evidence behind healthcare decisions.

The study is an important starting point, not the final answer. Its sample was small, came from one hospital, and did not include major clinical predictors such as cancer stage or treatment type. Future research should test the model on larger and more diverse patient groups. Even with these limitations, the message is clear: incomplete medical data do not have to mean incomplete understanding.

Reference
DOI:
https://doi.org/10.20956/j.v22i2.46746

Contact:
Prof. Dr. Nurtiti Sunusi, S.Si., M.Si.
+62 813-9540-8907
nurtitisunusi@unhas.ac.id /