BS in Statistics
Background The 5-year survival rate for patients with breast cancer is much higher than patients with other types of cancer. Due to this longer survival period, breast cancer patients also tend to have increased rates of lost to follow-up, when compared to other cancers. When a patient becomes lost, the occurrence of distant metastasis cannot be reliably ascertained, unless the patient had a breast cancer-specific (BC) death. The impact of lost patients on recurrence rates and disease-free survival (DFS) was explored in breast cancer patients seen at the City of Hope from 1997 to 2012.
Methods Female breast cancer patients with a stage of 0, I, II, or III at diagnosis were included in these analyses (N=2,358). Of these patients 1,937 were deemed non-lost and 421 were lost. Kaplan-Meier estimates for DFS were stratified by lost status. Cox proportional hazards models were built to adjust for multiple predictors such as age group at diagnosis, race, comorbidity score, final cancer stage at diagnosis, and lymphovascular invasion (LVI) status. BC death rates were compared between non-lost and lost patients using a chi-square test. Missed recurrences were estimated and recurrence rates were calculated. Patients were separated into 20 groups based on propensity scores from a logistic regression model using categorical distance between the patient’s residence and the City of Hope, age group at diagnosis, final cancer stage at diagnosis, hormone receptor status, and her2/neu status to predict the probability of becoming lost. Lost patients were removed and replaced with simulated lost patients. Simulated lost patients were sampled with replacement from the non-lost patients within each group and then one year of information was removed from those patients. The new 5-year DFS rate was calculated. This process of simulating lost patients and recalculating the 5-year DFS was bootstrapped 1,000 times.
Results The 5-year DFS rate was 84.6% for non-lost patients and 95.1% for lost patients. Adjusting for age, race, comorbidity score, stage, and LVI status, the risk of death or recurrence is 61.0% lower for lost patients compared to non-lost patients. The BC death rate was 8.2% for non-lost patients and 2.4% for lost patients. This difference in BC death rates may be due to delays in death information for lost patients. There were 66 observed missed recurrence and 42 estimated unobserved missed recurrences. The observed recurrence rate was 7.1% and the estimated recurrence rate was 11.7%. The mean 5-year DFS rate for simulated lost patients was 86.2%.
Conclusion There are missing recurrences for both non-lost and lost patients, yielding a lower observed recurrence rate than estimated and inflated DFS rates. Lost patients lead to even more missing recurrence information, yielding larger differences in the observed rates and estimated rates. Researchers could mention the lost to follow-up rate and the possible effects on DFS to avoid misleading rates.