The characteristics and survival of second primary lung cancer after Hodgkin’s lymphoma: A comparison with first primary lung cancer using the SEER database

Objective The study aimed to compare the characteristics and prognosis between patients with second primary lung cancer following Hodgkin’s lymphoma and those with primary lung cancer. Materials and methods Using the SEER 18 database, the characteristics and prognosis were compared between the second primary non-small cell lung cancer following Hodgkin’s lymphoma (HL-NSCLC) (n = 466) and the first primary non-small cell lung cancer (n = 469,851)(NSCLC-1), as well as between the second primary small cell lung cancer following Hodgkin’s lymphoma (n = 93) (HL-SCLC) and the first primary small cell lung cancer (n = 94,168) (SCLC-1). Comparisons of categorical variables were performed using Chi-square or Fisher’s test. Continuous variables were compared using the Mann-Whitney U test. Overall survival (OS) was estimated using the Kaplan-Meier method, and the difference between groups was analyzed by log-rank test. Results HL-NSCLC group had more males than NSCLC-1 group, and the median age of HL-NSCLC group was younger than that of NSCLC-1 group. Patients with HL-NSCLC showed inferior OS than those with NSCLC-1 (median: 10 months vs. 11 months, P = 0.006). Both HL-SCLC and SCLC-1 groups had poor prognosis, with median OS of 7 months (P = 0.4). The 3-year cumulative risks of death from any cause for patients with the latencies from HL to NSCLC of 0 to 5 years, >5 to 10 years, >10 to 15 years, >15 to 20 years, and>20 years were 71.8%, 82.6%, 86.8%, 85.7% and 78.5%, respectively(P = 0.020). Conclusion HL-NSCLC patients had worse prognosis than NSCLC-1 patients, while HL-SCLC patients shared similar characteristics and survival with SCLC-1 patients.


Introduction
Hodgkin's lymphoma (HL) can be diagnosed at any age and it is most common in young adults [1]. Over the past decades, the remarkable progress in the treatment of HL has greatly improved survival outcomes, which makes HL one of the most curable cancers [2,3]. With the combination of multi-agent chemotherapy and radiotherapy, the overall survival (OS) rates at 5 and 10 years for HL are 86% and 80%, respectively [1]. However, the prolonged survival following effective HL treatment has been accompanied by increased risk of subsequent malignancies, cardiovascular disease or other late effects [4][5][6][7]. According to the cancer databases from North America and Europe, approximately 6.6% to 12% of HL survivors developed second primary cancers during further follow-up, and lung cancer (LC) accounted for the most frequent type of solid tumors [6,[8][9][10]. A study from Sweden reported 2.39-fold increased risk of developing a second cancer among HL survivors, with a 3.3-fold increased risk of developing second lung cancer [9]. Lung cancer accounted for a large proportion of late deaths among HL survivors [11].
In previous studies, risk factors of developing second primary lung cancer following HL (HL-LC) have been explored, including older age at the time of HL diagnosis, male, time from HL treatment, alkylating chemotherapy, thoracic radiation, cigarette smoking and a family history of cancer [9,[11][12][13][14]. In contrast, data regarding the prognosis of HL-LC patients is very limited, and few studies have clarified the prognostic factors affecting survival outcomes. Several studies reported a median OS of 9 to 12.6 months for HL-LC patients, but their sample sizes were small and heterogeneity was high [15][16][17]. Additionally, the comparison of clinical characteristics and survival outcomes between HL-LC and the first primary lung cancer (LC-1) was extremely difficult, due to the relatively small number of HL-LC patients. Only one population-based study showed that patients with second primary non-small cell lung cancer (NSCLC) following HL (HL-NSCLC) had inferior OS compared to those with the first primary NSCLC(NSCLC-1) [18], probably due to molecular changes and limited treatment options influenced by prior therapy for HL [19]. Despite this, to date, no study has focused on the characteristics and prognosis of second primary small cell lung cancer (SCLC) following HL (HL-SCLC).
Therefore, in this study, the Surveillance, Epidemiology, and End Results (SEER) database (http://seer.cancer.gov/) was utilized to select all the patients with HL-NSCLC or HL-SCLC, as well as those with NSCLC-1 or the first primary SCLC (SCLC-1). We aimed to investigate the clinicopathological characteristics and survival outcomes of HL-LC patients, and compare them with LC-1 patients. We also explored the prognostic factors affecting OS among patients with HL-NSCLC or HL-SCLC. 2016 varying)) (https://seer.cancer.gov/data-software/documentation/seerstat/nov2018) was used for patient collection in this study. This database contained the data of patients diagnosed between 1975 and 2016. Tumor histology was classified with the third edition of the International Classification of Diseases for Oncology (ICD-O-3). This study was approved by the Ethics Committee of National Cancer Institute (SEER Program), with an approval number of SAR0040750. This observational study used de-identified and publicly available data from SEER and thus did not require formal consent from patients. Fig 1 showed the case selection diagram. Firstly, the ICD-O-3 histology codes of HL were used to select HL patients, including nodular sclerosis (9663, 9664, 9665, 9667), classical Hodgkin lymphoma (9650), lymphocyte-rich (9651), mixed cellularity (9652), lymphocyte-depleted (9653, 9655) and nodular lymphocyte predominant (9659). A total of 58208 patients with HL were identified after the first step. Secondly, the variable of "Sequence number" was collected for these HL patients, and only the patients with "Sequence number" of "1st of 2 or more primaries" were included. A total of 5881 patients who developed subsequent cancer following the first primary HL were identified after the second step. Thirdly, the ICD-O-3 topographical codes (lung and bronchus (C34.0-C34.9)) and ICD-O-3 histology codes of LC (including squamous cell carcinoma (8070-8073, 8083), adenocarcinoma (8410, 8250-8253, 8255, 8260, 8323, 8480, 8481, 8550, 8560, 8570, 8574), large cell carcinoma (8012), NSCLC (8046) and SCLC (8041-8045)) were used to select LC patients. A total of 1000107 LC patients were identified after this step. Then, the variable of "Sequence number" was collected for these LC patients, and only the patients with "Sequence number" of "2nd of 2 or more primaries" were included. A total of 65536 patients who developed lung cancer as the second primary cancer were identified after the step.
The overlapping patients in the above two groups (5881 HL patients and 65536 LC patients) were selected. And a total of 627 patients who developed HL-LC were identified after the step. Forty-four patients with survival time of 0 days, with incomplete survival data, without active follow-up, or without histological confirmation were further excluded. A minimum latency of 2 months was required between initial HL and second primary lung cancer, in order to exclude synchronous primary cancers [20]. Thus twenty-four patients with interval between the two cancers being less than two months were excluded. Finally, a total of 559 patients were included for analysis, including 466 patients with HL-NSCLC and 93 patients with HL-SCLC.
Regarding the LC-1 group, a total of 1000107 primary lung cancer patients were collected from the same database. Of them, lung cancer was not the first primary malignancy in 213927 patients, and 141180 patients were diagnosed without histological confirmation. Therefore, they were excluded from the study. 83981 patients with a survival time of 0 days, with incomplete survival data, or without active follow-up were further excluded. Finally, a total of 564019 eligible patients were included for further analysis, consisting of 469851 NSCLC-1 and 94168 SCLC-1 patients.

Retrieved variables and study endpoint
The demographics and survival data of patients were collected, including age at diagnosis, sex, year of diagnosis, race, grade, Ann Arbor stage, seer historic summary stage, laterality, survival time, survival status and causes of death. Treatment information, such as surgery, radiotherapy and chemotherapy, were also retrieved. Lung cancer patients were grouped into localized, regional or distant stage according to SEER historic summary stage. Details of this staging were described in historic SEER coding manuals (http://seer.cancer.gov/tools/codingmanuals/ historical.html/). Generally, the neoplasm confined to the ipsilateral lung or bronchus was considered to be localized. The regional disease was defined by the involvement of hilar or mediastinal nodes and/or extension to regional structures. The neoplasm with metastatic spread beyond regional nodes or structures was classified as distant stage. The study endpoint of interest was OS. OS was measured from the date of lung cancer diagnosis to the date of death from any cause or last follow-up.

Statistical analysis
Comparisons of categorical variables were performed using Chi-square or Fisher's test. Continuous variables were compared using the Mann-Whitney U test. OS was estimated using the Kaplan-Meier method, and the difference between groups was examined by the log-rank test. The covariate-adjusted OS comparison between HL-LC and LC-1 groups was conducted based on a Cox proportional hazards model. Cox proportional hazard model was used to perform univariate and multivariate analysis for OS. The survival analysis was conducted among patients with complete lung cancer stage information. Variables with P values of less than 0.1 in the univariate analyses were included into the multivariate analysis. All analyses were two-sided, and P<0.05 was defined as statistically significant. All the analyses were done using SPSS software (version 26.0, SPSS, IBM) and R version 3.6.2 (http://www.r-project.org/).

Patient characteristics at time of HL diagnosis
A total of 466 HL survivors developed NSCLC following HL, with a median latency of 10.3 years (range, 0.2-39.7 years), and 93 survivors developed SCLC following HL, with a median latency of 9.8 years (range, 0.34-35.1 years). Table 1 shows patient characteristics at time of HL diagnosis according to subsequent lung cancer stage. Patients with shorter latency from HL to lung cancer had earlier lung cancer stage, and the median latencies were 7, 10.1 and 15 years for localized, regional and distant stage groups, respectively(P<0.001). The age of patients at HL diagnosis was significantly different between the three stage groups, with median age of 55, 52 and 43 years for localized, regional and distant stage groups, respectively (P<0.001). Additionally, the HL patients diagnosed in 2000-2016 had earlier stage of lung cancer than those diagnosed in 1975-1999 (P = 0.002). In contrast, among the three stage groups for HL-SCLC, the distributions of latency, age, year at HL diagnosis and HL chemotherapy were similar. Table 2 summarized patient characteristics at the time of NSCLC diagnosis. The lung cancer stage was available in 414 patients, including 109 cases (26.3%) with localized stage, 113 (27.3%) with regional stage and 192 (46.3%) with distant stage. As for NSCLC-1, 387791 patients had complete stage information, in whom 86604 (22.3%), 113331 (29.2%) and 187856 (48.4%) patients had localized, regional and distant disease, respectively. Patients with HL-NSCLC were significantly younger than those with NSCLC-1, with median age of 61 (range, 23-87) and 67 (range, 5-105), respectively (P<0.001). Similar findings were observed among patients across three different stage groups. Males accounted for 65.5% in HL-NSCLC patients and 56.8% in NSCLC-1 patients(P<0.001). In the regional and distant stage groups, a larger proportion of patients received radiotherapy for NSCLC-1 patients, compared to the HL-NSCLC patients (regional, P = 0.014; distant, P<0.001).

Patient characteristics at time of lung cancer diagnosis
Regarding SCLC, the median age at LC diagnosis of HL-SCLC and SCLC-1 groups were 65 and 66 years, respectively (P = 0.135). There were 11.3% (n = 9), 16.3% (n = 13) and 72.5% (n = 58) of patients having localized, regional, and distant disease in the HL-SCLC group, and 5.8%, 23.3%, and 70.8% having localized, regional, and distant disease in the SCLC-1 group (P = 0.055). Fewer patients in the HL-SCLC group underwent radiotherapy compared to the SCLC-1 group (P<0.001). The detailed clinicopathologic features of patients with HL-SCLC and SCLC-1 at the time of SCLC diagnosis are displayed in Table 3. Table 4 summarizes the OS of HL-NSCLC versus NSCLC-1, and HL-SCLC versus SCLC-1. For all patients with NSCLC, the OS of HL-NSCLC group was significantly inferior than that of NSCLC-1 group, with the median OS of 10 and 11 months, respectively (P = 0.006) (Fig 2a). Similar findings were found in both localized and regional disease groups (P = 0.003 and 0.002, respectively), whereas no difference was observed between the HL-NSCLC and NSCLC-1 groups in patients with distant disease (P = 0.200) (Fig 2b). As for SCLC, the median OS were both 7 months for HL-SCLC and SCLC-1 patients (Fig 2c and 2d).

Overall survival of HL-LC versus LC-1
Covariate-adjusted OS comparison between HL-NSCLC and NSCLC-1 groups was further conducted, based on a Cox proportional hazards model that included history of HL, age at LC diagnosis, sex, race, year of LC diagnosis, LC subtype, laterality, LC grade, LC stage, surgery for LC, radiotherapy and chemotherapy for LC (Table 4). After adjusting these variables,  After multivariate analysis, sex (P = 0.029), LC stage (regional vs. localized, P = 0.009; distant vs. localized, P<0.001) and surgery for LC (P<0.032) were independent factors predicting OS.

Prognostic factors among HL-SCLC
The survival analysis was conducted in 80 HL-SCLC patients with lung cancer stage information available. As shown in

The cumulative risk of death affected by the latency between diagnosis of HL and LC
The mortality risks affected by the latency between diagnosis of HL and LC were analyzed. Notably, the 3-year cumulative risks of death from any cause for HL-NSCLC patients with latencies of 0-5 years, >5 to 10 years, >10 to 15 years, >15 to 20 years, and>20 years were 71.8%, 82.6%, 86.8%, 85.7% and 78.5%, respectively(P = 0.020) (Fig 4a). Likewise, the cumulative risk of death from NSCLC was also affected by the latency, and the 3-year cumulative risks were 55.6%, 71.2%, 77.1%, 81.0% and 69.4% for the above five latency groups, respectively (P = 0.002) (Fig 4b). As for patients with HL-SCLC, the latency between HL and SCLC showed no effect on the cumulative risk of death from either any cause (P = 0.561) or SCLC (P = 0.110) (Fig 4c and 4d).

Discussion
Due to the rarity of HL-LC, it is difficult to thoroughly investigate this disease entity in large series. Until now, few studies have reported the characteristics and outcomes of HL-LC. In this large population-based study, we performed a direct comparison of the characteristics and survival between the HL-LC group and the LC-1 group. The survival outcomes of HL-SCLC were described for the first time. We demonstrated that HL-NSCLC patients had an inferior OS compared with NSCLC-1 patients (P = 0.006), whereas no significant difference in OS was observed between HL-SCLC patients and SCLC-1 patients (P = 0.390). Besides, this study explored the prognostic factors affecting OS for HL-NSCLC and HL-SCLC, which might provide guidance for monitoring and treating these patients. According to a systematic review conducted by Lorigan et al., HL survivors had an increased risk of 9.7% for developing second lung cancer, and the risk increased with time from initial treatment of HL, for as long as 20-25 years [11]. Our study revealed a median latency of 10 years between diagnosis of HL and lung cancer. Besides, 21.9% and 11.8% of patients developed HL-NSCLC and HL-SCLC with a long latency of more than 20 years, respectively. These data indicated the importance of long-term regular medical examinations for HL survivors. The chest computed tomography (CT) should be performed to ensure the early detection of second primary lung cancer [21][22][23]. Besides, our study also indicated that patients with shorter latency, older age and HL diagnosed in a more recent year tended to have earlier stage of secondary NSCLC. There were several possible explanations for these findings. Firstly, more frequent examinations might be performed in the first five years after treatment for HL patients, which led to a large proportion of cases incidentally diagnosed. Secondly, patients with older age were more likely to undergo medical examinations for pertinent symptoms compared to those with younger age. Furthermore, the increased health awareness of people, as well as the improved diagnostic techniques, might promote early detection and diagnosis of cancer. The second primary NSCLC might have different biological characteristics compared with the first primary NSCLC, which could be examined by next-generation sequencing [24]. Prior radiotherapy or chemotherapy against HL could lead to the occurrence of genetic alterations in patients, which had an effect on the biologic behavior of the second primary lung cancer. One previous study compared the tumor tissues between patients with the first primary NSCLC and those with the second primary NSCLC after radiation for HL, and the results indicated that HL-NSCLC patients had increased number of microsatellite alterations [19]. Our study observed an inferior OS for HL-NSCLC compared to NSCLC-1, which might be partly due to the biological changes caused by prior treatment against HL.
For patients with HL-LC, the prior treatment against HL might also cause injury of normal organs (especially for the heart, lung and bone marrow), thus reducing the tolerance to treatment against lung cancer. The OS was affected not only by the deaths from specific cancer but also other causes such as complications and treatment side effects. Therefore, compared to cancer specific survival (CSS), OS was a more sensitive endpoint to examine the difference in prognosis between HL-LC and LC-1 patients. Furthermore, the quality of OS was more reliable than that of CSS, theoretically. Based on the above reasons, the OS was chosen as the endpoint of interest in the study. It should be noted that, the difference in OS between HL-NSCLC and NSCLC in this study could be also due to the higher risk of comorbidities and treatment side effects for HL-NSCLC patients.
SCLC accounts for 15-20% of lung cancer cases, which showed more aggressive behavior than NSCLC. Schoenfeld et al. identified 55 patients with secondary lung cancer from 1976 primary HL patients, and only 11 patients belonged to HL-SCLC [17]. Another study reported 27 patients with secondary lung cancer following HL, and only 2 patients were diagnosed with HL-SCLC [15]. No study has reported the survival outcomes of HL-SCLC, due to the very limited sample size. To the best of our knowledge, we conducted the first population-based study   few patients with HL-LC died of Hodgkin's lymphoma, which indicated the importance of management of primary HL disease. Besides, this study showed that a higher proportion of patients in the HL-LC group died from non-Hodgkin's lymphoma (NHL) compared with those in LC-1 groups, which was consistent with previous study. Petrakova et al. reported that the NHL was one of the most common secondary cancers among primary HL survivors, with the standardized incidence ratio of 13.1 [25]. Few studies have explored the prognostic factors affecting the survival outcome of HL-LC. In this study, based on multivariate analysis, the variables of sex, LC stage and LC surgery remained the independent factors affecting OS. Females had superior OS than males, which was consistent with previous studies involving primary NSCLC [26,27]. It was not surprising that patients with localized stage and those receiving surgery obtained superior OS than those not, since they had limited tumor burden and good performance status. Among patients with HL-SCLC, older age and not receiving chemotherapy for LC were correlated with a worse OS. It might be due to the poor tolerance to treatment for these patients.
There were several limitations in this study. First, several important variables are not available in SEER database, such as the performance status, smoking history, medical comorbidities, radiation dose-fractionation schedules, the systemic therapy information and treatment response, etc. The radiation and chemotherapy information was displayed as "Yes" or "No/ unknown". "No" and "unknown" were grouped together and further detailed information was not available. Therefore, interpretation of data is limited. Second, the SEER historic stage, instead of TNM stage was used for analysis in the study, because it is the only consistent staging system available in the database throughout the years. Thirdly, since the OS was chosen as the endpoint of interest in this study, it could not be determined whether the difference in prognosis between HL-LC and LC-1 was due to cancer specific death or other reasons, and it required further investigations. Despite these limitations, this study included a large cohort of HL-LC patients with long follow-up time, which allowed for analyses of survival outcomes and death causes.

Conclusion
In conclusion, patients who developed second primary lung cancer after Hodgkin's lymphoma had worse prognosis than those with primary lung cancer. The percentage of deaths from lung cancer in the HL-NSCLC group was lower than that in the NSCLC-1 group. However, patients with HL-SCLC patients shared similar characteristics and survival outcomes with SCLC-1 patients. Further studies are warranted to elucidate the molecular biology of HL-LC patients.