Stage-Dependent Death Hazard and Conditional Survival of Patients with Papillary Thyroid Cancer

Dear Editor:

The incidence of papillary thyroid cancer (PTC) has increased substantially worldwide in the recent decades. ¹ Generally, PTC has an indolent course and a favorable prognosis, with a 10-year disease-specific survival (DSS) of more than 95%. ² However, patients with distant metastases have a substantially lower 10-year DSS rate (∼50%). ³ Moreover, since the 1980s, the incidence of differentiated thyroid cancer with distant metastasis has been steadily increasing in the United States. ³ Traditionally, survival rates are estimated from the time of diagnosis. However, the probability of survival changes over time, especially when patients have lived for a given period after the diagnosis. This study aimed to evaluate the survival of patients with PTC using the annual hazard rates of cancer death and conditional survival (CS). Death hazard illustrates the absolute hazard of death at any instant among the remaining at-risk individuals. ⁴ CS represents the probability of surviving additional years for patients who have lived for a designated period. ⁵

We utilized data from the Surveillance, Epidemiology, and End Results (SEER) database, including patients diagnosed with PTC between 2000 and 2019 as per the following International Classification of Diseases for Oncology, 3rd Edition histological codes: 8050, 8260, 8340–8344, 8350, 8450–8460. The exclusion criteria were as follows: (i) no thyroidectomy, (ii) more than one type of primary malignancy, (iii) unknown tumor stage, and (iv) 0 days of follow-up. We classified the tumor stage at diagnosis using the SEER Summary Stage as localized (confined to the thyroid), regional (extension into adjacent tissue or lymph nodes), and distant metastasis (lung, bone, liver, brain, distant nodes, and other sites).

The definition of distant metastatic disease was based on the presence or absence of distant organ or extracervical lymph node metastases. Metastatic location data were extracted from patients between 2010 and 2019, since this information was reported in the SEER database from 2010 onward. The last available follow-up date was November 2021. This study was exempt from review by the Institutional Review Board of the First Hospital of China Medical University owing to the use of de-identified data.

The primary endpoint of this study was DSS, measured from the time of diagnosis to death due to PTC. The maximum likelihood estimate from the piecewise exponential model was used to evaluate the annual hazard rate and the kernel-smoothing method was used to display the graphics. The Kaplan–Meier method and log-rank test were used for estimation and between-group comparison of actuarial DSS, respectively. CS was defined as the probability of surviving an additional y years, given that a patient has already lived for x years, and may be summarized as follows: CS(x|y) = S(x+y)/S(x), where S(x) represents overall survival at x years. The CS was computed from life table survival data. A two-tailed p-value <0.05 was considered statistically significant. All analyses were conducted using the Stata software (version 16.0; Stata Corporation Ltd., College Station, TX).

We included 135,607 patients, followed for a median of 6.92 years (interquartile range 3.25–11.50 years). Of these patients, females (78.5%), white (80.5%), and localized disease (67.4%) predominated. During the study follow-up period, 1528 (1.1%) patients died of PTC. The annual hazard curve of cancer mortality for the entire population presented a single-peaked distribution, which emerged at 2.5 years after diagnosis. Subsequently, the curve declined from slowly to quickly until the end of the follow-up period (Fig. 1a). The annual hazard curve of patients with distant metastasis presented a similar pattern; however, that of patients with localized or regional disease showed two almost parallel lines. The death hazard rate of patients with distant metastasis was higher than that of patients with localized or regional disease during the entire follow-up period (Fig. 1b).

OPEN IN VIEWER

FIG. 1.

Annual hazard rate of cancer death for the entire cohort (a) and patients with different SEER Summary Stages (b); Kaplan–Meier curves for actuarial disease-specific survival for the entire cohort (c) and patients with different SEER Summary Stages (d); the 10-year conditional disease-specific survival at different time points for the entire cohort (e) and patients with different SEER Summary Stages (f). SEER, Surveillance, Epidemiology, and End Results.

Most deaths (1327 of 1528, 86.8%) occurred within 10 years after diagnosis, with the majority (923 of 1528, 60.4%) occurring within the first 5 years. The 5- and 10-year DSS rates of the entire study population were 99.19% ± 0.03% and 98.57% ± 0.04%, respectively (Fig. 1c). When stratified according to SEER Summary Stage, the 5-year DSS rates for patients with localized, regional, and distant disease were 99.91% ± 0.01%, 98.88% ± 0.06%, and 82.79% ± 0.72%, while the 10-year DSS rates were 99.77 ± 0.02%, 97.80% ± 0.10%, and 74.34% ± 0.94%, respectively (Fig. 1d). Statistically significant differences were found in DSS rates among the different tumor stages (p < 0.001).

As the overall time a patient has already survived has increased, the conditional DSS has also increased. For example, if a patient was alive at 1, 2, 3, 4, and 5 years after diagnosis, then their probabilities of remaining alive at 10 years were 98.69%, 98.72%, 98.74%, 98.72%, and 98.79%, respectively (Fig. 1e). The 10-year conditional DSS was stable in patients with regional disease and excellent in those with local disease. The 10-year conditional DSS was the worst, but the improvement over time was the most substantial in patients with distant metastasis (Fig. 1f).

The impact of the location and number of metastatic lesions on the annual hazard curve of cancer mortality and conditional DSS was further evaluated in post hoc secondary analyses. Data on metastatic lesions at diagnosis were available for 695 patients and these patients had a median follow-up period of 2.83 years (interquartile range 1.08–5.50 years). The lungs (47.9%) were the most common site for distant metastasis, followed by the bone (16.6%), distant lymph nodes (10.9%), brain (1.2%), and liver (1.2%). Approximately 22.2% of patients developed distant metastasis to multiple sites. Finally, 151 (21.7%) patients died from PTC. As there were few patients with brain or liver metastases, they were not included in the following statistical analyses.

Patients with multiple metastatic sites had a higher death hazard rate during the first 5 years and worse 5-year conditional DSS than those with a single metastatic site (hazard rate: 0.12 [confidence interval {CI}: 0.10–0.14] vs. 0.04 [CI: 0.03–0.05], p < 0.001; conditional DSS: 68.91% [CI: 65.46–72.36%] vs. 83.97% [CI: 79.77–88.17%], p < 0.001); however, they appeared to have greater survival improvement over time (Fig. 2a, b). Among patients with single-site metastasis, patients with lung metastasis had the highest death hazard rate (0.07 [CI: 0.06–0.08]) during the first 2.5 years; thereafter, those with bone metastasis had the highest death hazard rate (0.08 [CI: 0.07–0.09]) (Fig. 2a). Patients with distant lymph node metastases had a 5-year conditional DSS of 98.99% [CI: 94.04–99.99%], which was relatively stable. The 5-year conditional DSS improved gradually in patients with lung metastasis, whereas it worsened in patients with bone metastasis farther from the diagnosis (Fig. 2b).

OPEN IN VIEWER

FIG. 2.

Annual hazard rate of cancer death (a) and 5-year conditional disease-specific survival (b) for papillary thyroid cancer patients with distant metastasis.

One of the most important findings of this study was that the annual hazard curve of cancer mortality for the entire population presented a single-peaked distribution at 2.5 years and declined until the end of the follow-up period. We previously reported a double-peaked annual hazard curve of cancer mortality with surges at 10 and 20 years after surgery in a Japanese cohort consisting of 400 patients with PTC. ² Potential factors contributing to the difference in the annual hazard curves between these two cohort studies included race/ethnicity, sample size, risk stratification, surgical procedure, adjuvant radioactive iodine therapy administered, and follow-up time. Furthermore, we found that patients with multiple metastatic sites had a higher death hazard rate than those with single-site metastasis. Among patients with single-site metastasis, those with bone metastasis had the highest death hazard rate.

Our results demonstrated that conditional DSS improved over time in a stage-dependent manner among patients with PTC, and the greatest improvement was observed in patients with distant metastases, which is consistent with the results of the study by Banerjee et al. ⁶ Patients with multiple metastatic sites had worse 5-year conditional DSS but achieved more obvious survival improvement than patients with a single metastatic site. Different time-varying patterns of conditional DSS were displayed according to different metastatic sites in patients with a single metastatic site. Previous studies have explored the CS of patients with thyroid cancer. ^{6 –9}

In 2009, Bleyer et al. explored how the CS changed as a function of age at diagnosis in cancer patients and reported that the improvement in the 5-year CS of young patients with thyroid cancer was less than that in older patients. ⁷ Janssen-Heijnen et al. also reported that younger patients with thyroid cancer had better 5-year CS, and older patients had more obvious improvement in 5-year CS over time. ⁸ Ito et al. demonstrated that the 5-year CS of patients with thyroid cancer increased over time. ⁹ Banerjee et al. described the 5-year conditional DSS of 43,392 patients with well-differentiated thyroid cancer based on sex, age, and stage using data from the SEER registry between 1998 and 2005. ⁶ They found that younger patients had better 5-year conditional DSS for both sexes. ⁶

Furthermore, the 5-year conditional DSS of patients with different tumor stages presented different profiles. ⁶ For patients with localized and regional disease, the 5-year conditional DSS was excellent and relatively static over time, whereas for those with distant metastasis, the 5-year conditional DSS was the worst but gradually improved over time. ⁶ However, prior studies did not focus exclusively on PTC or consider the sites of distant metastases. The results of this study are consistent with our recent findings in anaplastic thyroid cancer (ATC) patients. ¹⁰ The worst CS and greatest improvement were observed in patients with advanced tumor stages. Unlike PTC, CS improved over time in patients with ATC at any tumor stage. ¹⁰ This discrepancy may be due to the different types of malignancies.

The advantages of our study include use of a nationally representative population-based cohort with long-term follow-up. However, this study is subject to several limitations. First, treatment information was limited, and it is possible that treatments varied over the study period. Second, the study is subject to U.S. population, and the findings may not be generalizable to other populations. Third, no data on tumor recurrence were available, and we were unable to estimate recurrence-free survival. Fourth, data on socioeconomic status were limited. Finally, some of the secondary analyses on location and number of metastatic lesions were performed post hoc.

Using data from the SEER registry, we showed an improvement in the conditional DSS for PTC. This information is valuable for consideration in survivorship care of individuals diagnosed with PTC.