The burdens of lung cancer involved multiple primary cancers and its occurring patterns–SEER Analysis between 1973 and 2006

The prognosis of malignancies has improved in recent years, subsequent primary cancers (SPCs) have become more frequent. This study investigates the patterns of lung cancer involved multiple primary cancers. We enrolled 206,619 primary lung cancer patients and 2,071,922 patients with other primary malignancies from Surveillance, Epidemiology and End Results (SEER) database. Observed annual risk (OAR) and absolute numbers were used to describe the risk of SPC and observed cases of SPC per 10,000 person-years at risk. Overall, OAR of SPCs following lung cancer was 176.28. At follow-up, 41.26% of SPCs occurred within 12–59 months while the highest OAR appeared after 120 months. The overall OAR of subsequent lung cancer after other malignancies was 27.90. Overall, the highest OAR and the highest absolute numbers of subsequent lung cancers were noticed 60–119 months and over 120 months post-diagnosis, respectively. Ten related cancers were listed. Our findings encourage surveillance for 10 common SPCs in lung cancer survivors during follow-up as well as screening for lung cancer after 10 common malignancies.

SPCs following lung and bronchus cancer. As shown in Table 2, the overall OAR of SPCs was 176.28 per 10,000 PYR. The OAR was 212.27 per 10,000 PYR in males (n = 7,659, OAR = 212.27, AER = 30.05), and 134.88 per 10,000 PYR in females (n = 4,232, AER = 16.97). Table 2 and Fig. 1 describe the OARs and EARs for the 10 most common SPCs after lung and bronchus cancer. The observed incidence, as measured by OAR, ranged from low OAR for larynx (n = 392, OAR = 5.81, AER = 3.96) to high OAR for prostate cancer (n = 2,213, OAR = 61.33, AER = −7.87). Stratification by gender revealed that OARs of all selected SPCs were higher in males than in females, with the exception of breast cancer. In terms of the absolute number, the most common SPC was breast cancer (n = 1,202, OAR = 38.31, AER = −0.18) in female patients, while prostate cancer (n = 2,213, OAR = 61.33, AER = −7.87) was the most frequent SPC in males. Furthermore, incidences varied across different latency periods but with similarly fluctuant patterns. Presented in Fig. 2, Appendix Tables A1 and A2, 40.64% SPCs occurred in the early period (12-59 months) in terms of the absolute number, and this phenomenon persisted in all selected SPCs for both males or females. Moreover, OARs increased with follow-up and peaked at around 10 years after initial diagnosis (n = 2,786, OAR = 193.82, AER = 25.20), meaning that 1 out of every 51 lung and bronchus cancer patients would develop a SPC each year of this period. Furthermore, late latency (more than 120 months) was the most dangerous period in 4 selected SPCs among males and 4 among females.
According to Fig. 3, Appendix Tables A3 and A4, most subsequent lung and bronchus cancers occur 10 years after initial diagnosis (n = 19,999, OAR = 28.21, AER = 2.26) in the field of absolute numbers. This was also observed in 8 sites in female patients and 4 sites in males. Regarding the OAR, however, the highest risk of subsequent lung cancer was found 60-119 months after initial diagnosis (n = 16,671, OAR = 28.88, AER = 4.04). This persisted in males (n = 10,522, OAR = 39.05, AER = 3.9) but shifted to a later period in female patients (after 120 months, n = 9,647, OAR = 22.31, AER = 2.01). The risk patterns during follow-up between the two genders were similar in that the highest OAR was observed in the late period (after 120 months) in 6 sites in males and 6 sites in females.

Discussion
In the current population-based study, we reported the risks of SPCs after lung and bronchus cancer as well as the risks of lung and bronchus cancer after other malignancies. Recently, various SPCs have been reported after lung cancer in several centers [13][14][15] . In the present study, we examined absolute numbers, OARs, and occurrence patterns of the 10 most common SPCs, emphasizing the need for directed surveillance of these conditions. Our results showed overall OAR of SPCs was 176.28 per 10,000 PYR, with AER of 23.97. In other words, 176 of 10,000 patients will develop a SPCs as a result of the history of lung cancer in one year, of which, 24 SPCs were additional occurrences beyond the number of primary cancers expected in the reference population. The highest OARs of SPC were observed in male prostate cancer patients and female breast cancer patients, consistent with the data of the general population, being the most frequently diagnosed malignancies of the male and female population 4 . Moreover, the risk of some SPCs varied by gender. This suggests the use of different surveillance strategies for male and female patients with primary lung and bronchus cancer as their initial diagnosis.
During the follow-up period, absolute numbers and OARs varied substantially across periods but with similar patterns. Consistent with previous studies 8,9 , most SPCs occurred during the early period (12-59 months) in terms of the absolute number. In addition, we utilized the OAR to investigate the annual risks of SPCs in each period. The highest OAR was found in the late period (more than 120 months) either in overall SPCs or in selected malignancies, followed by the mid-term period (60-119 months). For instance, overall, 1 out of every 368 patients with lung cancer would develop a colorectal cancer each year during the follow-up period. Importantly, the risk rose to 1 in every 313 patients during 60-119 months and 1 in every 278 patients after 120 months. Showing an arbitrary mismatch with current follow-up strategies, which only emphasize early-phase screening, our results suggested that the frequency of surveillance of SPCs should be maintained throughout the entire follow-up period for each lung cancer survivor, and more intensely in the later period. Special attention needs to be paid to SPCs we described. Specific plans should be carefully framed by oncologists based on our findings and the NCCN guidelines of each malignancy [16][17][18] . In this study, we also ranked our results by primary cancer sites with the 10 most frequent subsequent lung cancers and performed the analyses accordingly ( Fig. 1 and Table 3). Primary cancers with the most subsequent lung cancer cases included prostate cancer among males and breast cancer among females, probably resulting from the already high incidences in the general population. The OARs of lung cancer following selected cancers yielded a 2-to 34-fold increase compared to the general incidence 4 . Furthermore, OARs were higher in males than in females among all selected non-genital malignancies. These data indicated that males with such primary cancers should be screened for lung cancer more frequently than females.
Although the highest absolute number and OAR of overall subsequent lung and bronchus cancer were observed for mid-term (60-119 months) and late (more than 120 months) periods, patterns varied by primary malignancy sites and gender. Such discrepancies might be attributed to different characteristics and therapies of primary cancers 14,15 . In our present work, we summarized the risk patterns of subsequent lung and bronchus cancer by following 10 common cancers to provide evidence supporting the revision of the screening schedules. Consequently, surveillance of subsequent lung and bronchus cancer should be emphasized and given equal attention during each follow-up period, and a specific schedule should be made by oncologists according to the primary malignancies and NCCN guidelines of lung cancer 19 .
Previous studies have typically reported the risk ratios of the total observed number to the total expected number in different follow-up periods. However, the results were not comparable due to the different durations of each period (6 months vs. 48 months vs. 60 months vs. unknown), so the ratios of the total number were inappropriate to describe the risk. The current research has analyzed the annual incidence (OAR) of each subsequent primary cancer, resulting in more objective understanding of actual risk. Taking the SPCs that followed lung and bronchus cancer, for instance, although the AER of larynx cancer was increased by 3.96 per 10,000 PYR, their OAR was only 5.81. In other words, 1 out of 1721 lung cancer survivors will develop a larynx cancer each year, raised from 1/5405 in the reference population. However, the incidence of lung and bronchus cancer patients getting prostate cancer was 1/163 every year, which is still 10-fold that of larynx cancer, even though it is decreased from 1/145. In addition, we believed even the AER showed a decrease of 0.18 after lung cancer diagnosis, the breast cancer is still the top malignancy demanding surveillance on females.
The major highlight of this study is its population-based nature, utilizing a collection of a large number of patients with minimized selection biases and long-term follow-up. SEER*stat software also allowed us to analyze the risks of diseases in different follow-up latencies and thus helped us discover the patterns of SPCs. More importantly, the current study introduced absolute numbers and OARs to report the results of SPCs, which can provide a more accurate portrayal of the cancer burden in malignancy survivors.
Indeed, our study has a number of limitations mostly due to the sole application of the SEER database. First, distributions of histological types vary across different malignancies and with a good amount of unspecified records; therefore, we decided not to perform stratification by histological types in the analyses. Furthermore, the SEER database's inaccessibility to known cancer risk factors, such as tobacco use, BMI, genetic data, chemotherapy, and the like, is also a major shortcoming that we have to address. These limitations prevented us from incorporating these factors into further analysis.
In conclusion, our findings refine the current surveillance strategy of SPCs for lung and bronchus cancer survivors during different follow-up periods as well as screening for subsequent lung and bronchus cancer in patients with other common malignancies.