Original Article
Increased Risk of Additional Cancers Among Patients With Gastrointestinal Stromal Tumors: A Population-Based Study James D. Murphy, MD, MS1; Grace L. Ma2,3; Joel M. Baumgartner, MD2; Lisa Madlensky, PhD, CGC4; Adam M. Burgoyne, MD, PhD5; Chih-Min Tang, PhD2; Maria Elena Martinez, PhD4; and Jason K. Sicklick, MD2
BACKGROUND: Most gastrointestinal stromal tumors (GISTs) are considered nonhereditary or sporadic. However, single-institution studies suggest that GIST patients develop additional malignancies at increased frequencies. It was hypothesized that greater insight could be gained into possible associations between GISTs and other malignancies with a national cancer database inquiry. METHODS: Patients diagnosed with GISTs (2001-2011) in the Surveillance, Epidemiology, and End Results database were included. Standardized prevalence ratios (SPRs) and standardized incidence ratios (SIRs) were used to quantify cancer risks incurred by GIST patients before and after GIST diagnoses, respectively, in comparison with the general US population. RESULTS: There were 6112 GIST patients, and 1047 (17.1%) had additional cancers. There were significant increases in overall cancer rates: 44% (SPR, 1.44) before the GIST diagnosis and 66% (SIR, 1.66) after the GIST diagnosis. Malignancies with significantly increased occurrence both before and after diagnoses included other sarcomas (SPR, 5.24; SIR, 4.02), neuroendocrine-carcinoid tumors (SPR, 3.56; SIR, 4.79), non-Hodgkin lymphoma (SPR, 1.69; SIR, 1.76), and colorectal adenocarcinoma (SPR, 1.51; SIR, 2.16). Esophageal adenocarcinoma (SPR, 12.0), bladder adenocarcinoma (SPR, 7.51), melanoma (SPR, 1.46), and prostate adenocarcinoma (SPR, 1.20) were significantly more common only before the GIST diagnosis. Ovarian carcinoma (SIR, 8.72), small intestine adenocarcinoma (SIR, 5.89), papillary thyroid cancer (SIR, 5.16), renal cell carcinoma (SIR, 4.46), hepatobiliary adenocarcinoma (SIR, 3.10), gastric adenocarcinoma (SIR, 2.70), pancreatic adenocarcinoma (SIR, 2.03), uterine adenocarcinoma (SIR, 1.96), non–small cell lung cancer (SIR, 1.74), and transitional cell carcinoma of the bladder (SIR, 1.65) were significantly more common only after the GIST diagnosis. CONCLUSIONS: This is the first population-based study to characterize the associations and temporal relations between GISTs and other cancers by both site and histological type. These associations may carry important clinical implications for future cancer screening and treatment strategies. Cancer 2015;121:2960-7. C 2015 American Cancer Society. V KEYWORDS: gastrointestinal stromal tumor (GIST), multiple primary neoplasm, second primary neoplasms, Surveillance, Epidemiology, and End Results (SEER), synchronous neoplasms.
INTRODUCTION Approximately 5% of all gastrointestinal stromal tumors (GISTs), the most common mesenchymal tumors of the gastrointestinal (GI) tract, have a hereditary etiology.1 Known heritable GIST syndromes are caused by germline mutations in KIT (c-KIT), platelet-derived growth factor receptor a (PDGFRa), neurofibromin 1, and succinate dehydrogenase subunits.2 Patients and families with these genomic alterations frequently develop multiple benign and malignant tumors. However, these syndromes account for only approximately 5% of GISTs2-6; the remaining 95% are considered sporadic. Results of several descriptive, single-institution case series suggest that patients with sporadic GISTs develop synchronous or metachronous malignancies with frequencies far exceeding the often cited 1-in-9 lifetime chance of developing 2 primary cancers.5,7-11 A review of published case series of sporadic GISTs, which includes an additional set of cases from a single institution, is the largest study published to date and includes 4813 patients.8 In this review, the frequency of additional malignancies varied from 4.5% to 33% among the various series. A subsequent study by Pandurengan et al7
*Corresponding author: Jason K. Sicklick, MD, Division of Surgical Oncology, Department of Surgery, Moores Cancer Center, University of California San Diego, 3855 Health Sciences Drive, MC 0987, La Jolla, CA 92093-0987; Fax: (858) 228-5153; [email protected] 1 Department of Radiation and Applied Sciences, Moores Cancer Center, University of California San Diego, La Jolla, California; 2Division of Surgical Oncology, Department of Surgery, Moores Cancer Center, University of California San Diego, La Jolla, California; 3Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pennsylvania; 4Department of Family and Preventive Medicine, Moores Cancer Center, University of California San Diego, La Jolla, California; 5 Division of Medical Oncology, Department of Internal Medicine, Moores Cancer Center, University of California San Diego, La Jolla, California.
See editorial on pages 2871-3, this issue. The first 2 authors contributed equally to this article. Additional Supporting Information may be found in the online version of this article. DOI: 10.1002/cncr.29434, Received: November 25, 2014; Revised: February 22, 2015; Accepted: February 25, 2015, Published online April 30, 2015 in Wiley Online Library (wileyonlinelibrary.com)
2960
Cancer
September 1, 2015
Increased Cancer Risk in GIST Patients/Murphy et al
confirmed these findings, with 159 of 783 GIST patients (20.3%) developing 1 or more additional malignancies over the study period. Other studies have also reported associations between GISTs and desmoids,12 acute myeloid leukemia,9 and other GI malignancies discovered incidentally during resections for GISTs or GISTs discovered incidentally during resections for other GI malignancies.13-15 When it is considered in its totality, the literature supports a possible association between GISTs and other malignancies not typically associated with known hereditary disorders.16 However, except for the link between GISTs and leukemia,9 disease associations have been reported only in a qualitative fashion, and none of these published reports have been based on populationlevel data. To address these important gaps in the literature, we used national cancer registry data to investigate the possibility of nonrandom associations between GISTs and other malignancies. Specifically, we assessed the risk of cancer by site and histology both before and after GIST diagnoses. As such, our report is the first populationbased study to quantify the frequency and temporal relations of GISTs and other cancer histologies not linked with known hereditary disorders.
MATERIALS AND METHODS Patients
We used data from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) database, which consists of 18 regional cancer registries that collect data on incident cancer diagnoses across the United States. These registries account for approximately 28% of the US population and include patients with GISTs diagnosed between 2001 and 2011. GISTs before 2001 were excluded because of the high prevalence of misdiagnoses and miscoding in the 1990s.17 We identified cases with histologically confirmed GISTs with GI tumor site codes (C150-C189, C199, C209-C212, C218, C220-C221, C239-C260, C268-C269, C480-C482, and C488) and the GIST histology code (International Classification of Diseases for Oncology, 3rd edition [ICD-O-3], code 8936). To exclude patients with a high likelihood of hereditary syndromes, we eliminated those diagnosed under the age of 20 years. Notably, there were 16 patients who developed another GIST after their initial GIST diagnoses. However, we could not ascertain whether these were cases of metachronous but sporadic GISTs, miscoded cases of metastatic GISTs, or GISTs associated Cancer
September 1, 2015
with multiple tumor syndromes. Therefore, we elected to include only the first GIST diagnosis in our analysis. Statistical Analysis
Cancer occurrence before GISTs was estimated with standardized prevalence ratios (SPRs), which were defined as the number of observed cases of additional cancers divided by the number of expected cases.18 The number of expected cases was calculated by the multiplication of the total at-risk person-years by the cancer prevalence specific to the patient’s age (in 5-year intervals), sex, race, and SEER registry grouping. The registries were divided into groups according to the dates on which they began collecting data (ie, 1973, 1992, and 2000). The at-risk time period extended from the date on which the registry started collecting data (or date of birth, if later) through the month before a patient’s GIST diagnosis. The cancer risk after GISTs was estimated with standardized incidence ratios (SIRs), which were defined as the number of observed cases divided by the number of expected cases of additional cancers.19,20 The number of expected cases was estimated by the multiplication of the total at-risk person-years by the cancer incidence specific to the patient’s age (in 5-year intervals), sex, race, year of GIST diagnosis (in 5-year intervals), and SEER registry grouping (as defined previously). The at-risk time period extended from the month of the GIST diagnosis through the date of death or last follow-up. The 2000 US standard population21 was used as the reference population when we were determining the expected prevalence (with the SPR) and incidence (with the SIR). We calculated SPRs and SIRs for any additional cancers as well as site-specific cancers. SPRs and SIRs greater than 1.0 indicated an excess in prevalence or incidence with respect to the general population. We assumed that the total number of observed events followed a Poisson distribution, and this allowed us to estimate the 95% confidence intervals (CIs) for SPRs and SIRs.22-25 We conducted subgroup analyses by stratifying patients by demographic and diseasespecific characteristics. The likelihood ratio test was used to assess heterogeneity in SIRs or SPRs between different subgroups of patient characteristics.20,26 In addition, we assessed the occurrence of cancer at varying times before and after the GIST diagnosis to elucidate the period of increased excess over time. All statistical tests were 2sided, and P values less than .05 were considered significant. Data were acquired directly from SEER and imported into SAS software, and all analyses were conducted with SAS software (version 9.4; SAS, Cary, NC). 2961
Original Article
RESULTS We identified 6112 patients diagnosed with GISTs between 2001 and 2011. The total observation period for additional cancers was 123,453 person-years, which included 101,551 years before the GIST diagnosis and 21,902 years after the GIST diagnosis. There were 1047 patients (17.1% or 1 in 5.8 patients) with a total of 1208 additional cancers. Specifically, 651 patients (62.2%) had other cancers diagnosed before GIST, 467 (44.6%) had other cancers diagnosed during the same month or after GIST, and 71 (6.8%) had other cancers diagnosed both before and after the GIST diagnosis. Among those with additional malignancies, 143 (13.7%) had more than 1 additional cancer, and 16 (1.5%) had more than 2 additional cancers. Table 1 presents demographic and clinical characteristics for the GIST cohort. Supporting Figure 1 (see online supporting information) shows the distribution of additional cancers by anatomic site and/or organ system. Less common malignancies, such as sarcomas, neuroendocrine-carcinoid tumors, and mesotheliomas, were placed in distinct categories because of their unique histologies and clinical significance. Overall, the most common neoplasms before and after GISTs were those of the genitourinary tract (35.8%; specifically, cancers of the prostate [57.4%], bladder [15.3%], and kidneys [14.1%]) and the GI tract (17.2%; specifically, colorectal adenocarcinomas [73.1%]; Supporting Fig. 1A). Breast cancer (11.9%) and respiratory cancer (8.2%) were also common, as were hematologic neoplasms (6.6%). Malignancies were then divided into those found before (Supporting Fig. 1B) and after GISTs (Supporting Fig. 1C). We then compared the occurrence of additional cancers among GIST patients and the US population. GIST patients in the SEER database had a 44% increased prevalence of cancers occurring before the GIST diagnosis (SPR, 1.44; 95% CI, 1.33-1.55) and a 66% increased relative risk of developing cancers after the GIST diagnosis (SIR, 1.66; 95% CI, 1.52-1.81; Fig. 1). Cancers with significantly increased occurrence both before and after the GIST diagnosis included sarcomas, neuroendocrinecarcinoid tumors, colorectal adenocarcinoma, and nonHodgkin lymphoma. Malignancies with a significantly elevated prevalence only before GISTs included esophageal adenocarcinoma, bladder adenocarcinoma, melanoma, and prostate adenocarcinoma. Malignancies with a significantly elevated incidence only after GISTs included small bowel adenocarcinoma, papillary thyroid cancer, renal cell carcinoma, gastric adenocarcinoma, hepatobiliary adenocarcinoma, pancreatic adenocarcinoma, non–small 2962
TABLE 1. Demographic and Clinical Characteristics of the GIST Cohort (n 5 6112) Characteristic Age at GIST diagnosis 20-39 y 40-49 y 50-59 y 60-69 y 70-79 y 80 y Sex Female Male Race White/unknown Black Other Ethnicity Hispanic Non-Hispanic/unknown Year of diagnosis 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 GIST location Esophagus Stomach Small intestine Colorectal Hepatobiliary Pancreas Retroperitoneum Peritoneum, omentum, and mesentery Other digestive organs GIST size 2 cm >2, 5 cm >5, 10 cm >10 cm Unknown
No.
%
325 739 1289 1553 1398 808
5.3 12.1 21.1 25.4 22.9 13.2
2860 3252
46.8 53.2
4320 1079 713
70.7 17.7 11.7
560 5552
9.2 90.8
408 522 505 521 531 494 517 577 604 730 703
6.7 8.5 8.3 8.5 8.7 8.1 8.5 9.4 9.9 11.9 11.5
33 3368 1762 343 5 23 57 126 384
0.5 55.1 28.8 5.6 0.1 0.4 0.9 2.1 6.3
430 1351 1847 1478 1006
7.0 22.1 30.2 24.2 16.5
Abbreviation: GIST, gastrointestinal stromal tumor.
cell lung cancer (NSCLC), and transitional cell carcinoma of the bladder. Among women, there was also an increased post-GIST incidence of ovarian carcinoma, uterine adenocarcinoma, and other genitourinary cancers. There were also several rare cancers that appeared significantly more often in the GIST cohort versus the general US population. However, because some had only 1 reported case in the SEER database, we elected to exclude them from Figure 1. These included prostatic transitional cell carcinoma before GISTs (SPR, 51.5) and squamous cell carcinoma of the breast after GISTs (SIR, Cancer
September 1, 2015
Increased Cancer Risk in GIST Patients/Murphy et al
Figure 1. Cancer development before and after the diagnosis of GIST. The occurrence of each cancer before the diagnosis of GIST is reported with the SPR (solid squares), and the occurrence after the diagnosis of GIST is reported as the SIR (white circles). Horizontal lines illustrate associated 95% CIs. Only cancers with statistically significantly elevated SPRs and SIRs (P 10 cm; in particular, patients with tumors 2 cm had the highest likelihood of having addiCancer
September 1, 2015
tional neoplasms before and after GISTs. Overall, these findings suggest that ethnicity and tumor size affect the risk of developing additional cancers. When we analyzed the period of increased occurrence of additional cancers, we found that the maximum increase occurred within 1 year before and after the GIST diagnosis (Fig. 2). The median latency period from the diagnosis of the first cancer to the diagnosis of GIST was 3.6 years for all patients, although it should be noted that SEER registries with longer follow-up (ie, earlier data collection dates) also had longer median latencies: the median latency was 6.1 years in the 1973 registries, 3.5 years in the 1992 registries, and 1.5 years in the 2000 registries. The median time from the diagnosis of GIST to the diagnosis of a subsequent cancer was 10 months for the entire cohort. 2963
Original Article TABLE 2. Occurrence of Additional Cancers Before and After the Diagnosis of GIST According to Select Patient Characteristics Cancers Before GIST
Age at GIST diagnosis 20-39 y 40-49 y 50-59 y 60-69 y 70-79 y 80 y Sex Female Male Race White/unknown Black Other Ethnicity Hispanic Non-Hispanic/unknown GIST size 2 cm >2, 5 cm >5, 10 cm >10 cm Unknown GIST site Stomach Small intestine Other Registry group 1973-2011 1992-2011 2000-2011
Cancers After GIST
SPR (95% CI)
Pa
Observed
Expected
SIR (95% CI)
Pa
1.8 11.8 48.1 125.5 179 118.5
2.82 1.87 2.16 1.53 1.33 1.20
(0.92-6.58) (1.17-2.83) (1.77-2.62) (1.32-1.76) (1.17-1.51) (1.01-1.41)
.11
6 24 76 144 176 79
2.1 13.7 48.4 97.5 101.3 38.6
2.86 1.75 1.57 1.48 1.74 2.05
(1.05-6.22) (1.12-2.61) (1.24-1.97) (1.25-1.74) (1.49-2.01) (1.62-2.55)
.71
280 423
199.4 285.2
1.40 (1.24-1.58) 1.48 (1.34-1.63)
.57
201 304
117.8 183.8
1.71 (1.48-1.96) 1.65 (1.47-1.85)
.89
526 126 51
369.3 74.8 40.5
1.42 (1.31-1.55) 1.68 (1.40-2.01) 1.26 (0.94-1.66)
.41
357 97 51
228.3 50.4 22.9
1.56 (1.41-1.73) 1.92 (1.56-2.35) 2.23 (1.66-2.93)
.26
33 670
37.6 447
0.88 (0.60-1.23) 1.50 (1.39-1.62)
.02
31 474
22.2 279.4
1.40 (0.95-1.98) 1.70 (1.55-1.86)
.47
91 187 182 126 117
37 109.8 144.7 113.4 79.8
2.46 1.70 1.26 1.11 1.47
.0009
66 119 154 94 72
19.1 71.3 96.8 69.9 44.6
3.46 1.67 1.59 1.34 1.61
(2.67-4.40) (1.38-2.00) (1.35-1.86) (1.09-1.65) (1.26-2.03)
.02
406 192 105
279 126.2 79.4
1.46 (1.32-1.60) 1.52 (1.31-1.75) 1.32 (1.08-1.60)
.73
303 140 62
171.6 86.9 43.1
1.77 (1.57-1.98) 1.61 (1.36-1.90) 1.44 (1.10-1.84)
.57
339 108 256
243.4 76 165.1
1.39 (1.25-1.55) 1.42 (1.17-1.71) 1.55 (1.37-1.75)
.81
183 71 251
106 42.3 153.3
1.73 (1.49-2.00) 1.68 (1.31-2.12) 1.64 (1.44-1.85)
.94
Observed
Expected
5 22 104 192 238 142
(1.98-3.02) (1.47-1.97) (1.08-1.45) (0.93-1.32) (1.21-1.76)
Abbreviations: CI, confidence interval; GIST, gastrointestinal stromal tumor; SIR, standardized incidence ratio; SPR, standardized prevalence ratio. a The P values represent a likelihood ratio test for assessing heterogeneity in SIRs or SPRs between different subgroups of patient characteristics.
DISCUSSION To the best of our knowledge, this is the first populationbased study with present-day SEER ICD-O-3 coding to quantify the occurrence of specific malignancies before and after GISTs. Although some of our results confirmed previous studies, we have also refuted many earlier findings, identified several novel associations, and provided a comprehensive description and statistical examination to support the existence of nonrandom associations between GISTs and other malignancies. Our results showed that the increased occurrence of other cancers was specific to the anatomic site and the histological type, and it had a temporal relation with the GIST diagnosis. In particular, the data pointed to the possibility of many new, clinically relevant associations between GISTs and cancers distinct from known disorders such Carney’s triad/tetrad (eg, gastric GISTs, pulmonary chondromas, extra-adrenal paragangliomas, and 2964
adrenal adenomas),27-29 familial GIST syndromes (eg, activating germline mutations in KIT and PDGFRa with multifocal GISTs),3,30-33 neurofibromatosis type 1 (eg, GISTs, sarcomas, periampullary/pancreatic neuroendocrine tumors, and pheochromocytomas),34-39 and Carney-Stratakis syndrome (eg, GISTs and paragangliomas).40-42 Because sarcomas and neuroendocrine tumors, which are found in neurofibromatosis, constituted only 4.0% and 3.1% of the 1208 additional cancers, respectively, this suggests that other etiologies might be responsible for the observed trends. In corroboration with single-institution reports, our study shows that 17.1% of the GIST patients developed additional cancers. Moreover, anatomic sites of additional malignancies were fairly similar to those reported in previous studies, including 1 multiseries analysis, which found second cancers in 13% of GIST patients and elevated risks of melanoma and GI, lung, and prostate cancers.8 A Cancer
September 1, 2015
Increased Cancer Risk in GIST Patients/Murphy et al
Figure 2. Timeline for the occurrence of additional cancers with respect to the diagnosis of GIST. The occurrence of additional cancers with respect to the time of the GIST diagnosis is represented. The occurrence of cancer before the diagnosis of GIST is quantified with the SPR, and the occurrence of cancer after the diagnosis of GIST is quantified with the SIR. Error bars represent the 95% confidence intervals, and asterisks represent P
Increased Risk of Additional Cancers Among Patients With Gastrointestinal Stromal Tumors: A Population-Based Study James D. Murphy, MD, MS1; Grace L. Ma2,3; Joel M. Baumgartner, MD2; Lisa Madlensky, PhD, CGC4; Adam M. Burgoyne, MD, PhD5; Chih-Min Tang, PhD2; Maria Elena Martinez, PhD4; and Jason K. Sicklick, MD2
BACKGROUND: Most gastrointestinal stromal tumors (GISTs) are considered nonhereditary or sporadic. However, single-institution studies suggest that GIST patients develop additional malignancies at increased frequencies. It was hypothesized that greater insight could be gained into possible associations between GISTs and other malignancies with a national cancer database inquiry. METHODS: Patients diagnosed with GISTs (2001-2011) in the Surveillance, Epidemiology, and End Results database were included. Standardized prevalence ratios (SPRs) and standardized incidence ratios (SIRs) were used to quantify cancer risks incurred by GIST patients before and after GIST diagnoses, respectively, in comparison with the general US population. RESULTS: There were 6112 GIST patients, and 1047 (17.1%) had additional cancers. There were significant increases in overall cancer rates: 44% (SPR, 1.44) before the GIST diagnosis and 66% (SIR, 1.66) after the GIST diagnosis. Malignancies with significantly increased occurrence both before and after diagnoses included other sarcomas (SPR, 5.24; SIR, 4.02), neuroendocrine-carcinoid tumors (SPR, 3.56; SIR, 4.79), non-Hodgkin lymphoma (SPR, 1.69; SIR, 1.76), and colorectal adenocarcinoma (SPR, 1.51; SIR, 2.16). Esophageal adenocarcinoma (SPR, 12.0), bladder adenocarcinoma (SPR, 7.51), melanoma (SPR, 1.46), and prostate adenocarcinoma (SPR, 1.20) were significantly more common only before the GIST diagnosis. Ovarian carcinoma (SIR, 8.72), small intestine adenocarcinoma (SIR, 5.89), papillary thyroid cancer (SIR, 5.16), renal cell carcinoma (SIR, 4.46), hepatobiliary adenocarcinoma (SIR, 3.10), gastric adenocarcinoma (SIR, 2.70), pancreatic adenocarcinoma (SIR, 2.03), uterine adenocarcinoma (SIR, 1.96), non–small cell lung cancer (SIR, 1.74), and transitional cell carcinoma of the bladder (SIR, 1.65) were significantly more common only after the GIST diagnosis. CONCLUSIONS: This is the first population-based study to characterize the associations and temporal relations between GISTs and other cancers by both site and histological type. These associations may carry important clinical implications for future cancer screening and treatment strategies. Cancer 2015;121:2960-7. C 2015 American Cancer Society. V KEYWORDS: gastrointestinal stromal tumor (GIST), multiple primary neoplasm, second primary neoplasms, Surveillance, Epidemiology, and End Results (SEER), synchronous neoplasms.
INTRODUCTION Approximately 5% of all gastrointestinal stromal tumors (GISTs), the most common mesenchymal tumors of the gastrointestinal (GI) tract, have a hereditary etiology.1 Known heritable GIST syndromes are caused by germline mutations in KIT (c-KIT), platelet-derived growth factor receptor a (PDGFRa), neurofibromin 1, and succinate dehydrogenase subunits.2 Patients and families with these genomic alterations frequently develop multiple benign and malignant tumors. However, these syndromes account for only approximately 5% of GISTs2-6; the remaining 95% are considered sporadic. Results of several descriptive, single-institution case series suggest that patients with sporadic GISTs develop synchronous or metachronous malignancies with frequencies far exceeding the often cited 1-in-9 lifetime chance of developing 2 primary cancers.5,7-11 A review of published case series of sporadic GISTs, which includes an additional set of cases from a single institution, is the largest study published to date and includes 4813 patients.8 In this review, the frequency of additional malignancies varied from 4.5% to 33% among the various series. A subsequent study by Pandurengan et al7
*Corresponding author: Jason K. Sicklick, MD, Division of Surgical Oncology, Department of Surgery, Moores Cancer Center, University of California San Diego, 3855 Health Sciences Drive, MC 0987, La Jolla, CA 92093-0987; Fax: (858) 228-5153; [email protected] 1 Department of Radiation and Applied Sciences, Moores Cancer Center, University of California San Diego, La Jolla, California; 2Division of Surgical Oncology, Department of Surgery, Moores Cancer Center, University of California San Diego, La Jolla, California; 3Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pennsylvania; 4Department of Family and Preventive Medicine, Moores Cancer Center, University of California San Diego, La Jolla, California; 5 Division of Medical Oncology, Department of Internal Medicine, Moores Cancer Center, University of California San Diego, La Jolla, California.
See editorial on pages 2871-3, this issue. The first 2 authors contributed equally to this article. Additional Supporting Information may be found in the online version of this article. DOI: 10.1002/cncr.29434, Received: November 25, 2014; Revised: February 22, 2015; Accepted: February 25, 2015, Published online April 30, 2015 in Wiley Online Library (wileyonlinelibrary.com)
2960
Cancer
September 1, 2015
Increased Cancer Risk in GIST Patients/Murphy et al
confirmed these findings, with 159 of 783 GIST patients (20.3%) developing 1 or more additional malignancies over the study period. Other studies have also reported associations between GISTs and desmoids,12 acute myeloid leukemia,9 and other GI malignancies discovered incidentally during resections for GISTs or GISTs discovered incidentally during resections for other GI malignancies.13-15 When it is considered in its totality, the literature supports a possible association between GISTs and other malignancies not typically associated with known hereditary disorders.16 However, except for the link between GISTs and leukemia,9 disease associations have been reported only in a qualitative fashion, and none of these published reports have been based on populationlevel data. To address these important gaps in the literature, we used national cancer registry data to investigate the possibility of nonrandom associations between GISTs and other malignancies. Specifically, we assessed the risk of cancer by site and histology both before and after GIST diagnoses. As such, our report is the first populationbased study to quantify the frequency and temporal relations of GISTs and other cancer histologies not linked with known hereditary disorders.
MATERIALS AND METHODS Patients
We used data from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) database, which consists of 18 regional cancer registries that collect data on incident cancer diagnoses across the United States. These registries account for approximately 28% of the US population and include patients with GISTs diagnosed between 2001 and 2011. GISTs before 2001 were excluded because of the high prevalence of misdiagnoses and miscoding in the 1990s.17 We identified cases with histologically confirmed GISTs with GI tumor site codes (C150-C189, C199, C209-C212, C218, C220-C221, C239-C260, C268-C269, C480-C482, and C488) and the GIST histology code (International Classification of Diseases for Oncology, 3rd edition [ICD-O-3], code 8936). To exclude patients with a high likelihood of hereditary syndromes, we eliminated those diagnosed under the age of 20 years. Notably, there were 16 patients who developed another GIST after their initial GIST diagnoses. However, we could not ascertain whether these were cases of metachronous but sporadic GISTs, miscoded cases of metastatic GISTs, or GISTs associated Cancer
September 1, 2015
with multiple tumor syndromes. Therefore, we elected to include only the first GIST diagnosis in our analysis. Statistical Analysis
Cancer occurrence before GISTs was estimated with standardized prevalence ratios (SPRs), which were defined as the number of observed cases of additional cancers divided by the number of expected cases.18 The number of expected cases was calculated by the multiplication of the total at-risk person-years by the cancer prevalence specific to the patient’s age (in 5-year intervals), sex, race, and SEER registry grouping. The registries were divided into groups according to the dates on which they began collecting data (ie, 1973, 1992, and 2000). The at-risk time period extended from the date on which the registry started collecting data (or date of birth, if later) through the month before a patient’s GIST diagnosis. The cancer risk after GISTs was estimated with standardized incidence ratios (SIRs), which were defined as the number of observed cases divided by the number of expected cases of additional cancers.19,20 The number of expected cases was estimated by the multiplication of the total at-risk person-years by the cancer incidence specific to the patient’s age (in 5-year intervals), sex, race, year of GIST diagnosis (in 5-year intervals), and SEER registry grouping (as defined previously). The at-risk time period extended from the month of the GIST diagnosis through the date of death or last follow-up. The 2000 US standard population21 was used as the reference population when we were determining the expected prevalence (with the SPR) and incidence (with the SIR). We calculated SPRs and SIRs for any additional cancers as well as site-specific cancers. SPRs and SIRs greater than 1.0 indicated an excess in prevalence or incidence with respect to the general population. We assumed that the total number of observed events followed a Poisson distribution, and this allowed us to estimate the 95% confidence intervals (CIs) for SPRs and SIRs.22-25 We conducted subgroup analyses by stratifying patients by demographic and diseasespecific characteristics. The likelihood ratio test was used to assess heterogeneity in SIRs or SPRs between different subgroups of patient characteristics.20,26 In addition, we assessed the occurrence of cancer at varying times before and after the GIST diagnosis to elucidate the period of increased excess over time. All statistical tests were 2sided, and P values less than .05 were considered significant. Data were acquired directly from SEER and imported into SAS software, and all analyses were conducted with SAS software (version 9.4; SAS, Cary, NC). 2961
Original Article
RESULTS We identified 6112 patients diagnosed with GISTs between 2001 and 2011. The total observation period for additional cancers was 123,453 person-years, which included 101,551 years before the GIST diagnosis and 21,902 years after the GIST diagnosis. There were 1047 patients (17.1% or 1 in 5.8 patients) with a total of 1208 additional cancers. Specifically, 651 patients (62.2%) had other cancers diagnosed before GIST, 467 (44.6%) had other cancers diagnosed during the same month or after GIST, and 71 (6.8%) had other cancers diagnosed both before and after the GIST diagnosis. Among those with additional malignancies, 143 (13.7%) had more than 1 additional cancer, and 16 (1.5%) had more than 2 additional cancers. Table 1 presents demographic and clinical characteristics for the GIST cohort. Supporting Figure 1 (see online supporting information) shows the distribution of additional cancers by anatomic site and/or organ system. Less common malignancies, such as sarcomas, neuroendocrine-carcinoid tumors, and mesotheliomas, were placed in distinct categories because of their unique histologies and clinical significance. Overall, the most common neoplasms before and after GISTs were those of the genitourinary tract (35.8%; specifically, cancers of the prostate [57.4%], bladder [15.3%], and kidneys [14.1%]) and the GI tract (17.2%; specifically, colorectal adenocarcinomas [73.1%]; Supporting Fig. 1A). Breast cancer (11.9%) and respiratory cancer (8.2%) were also common, as were hematologic neoplasms (6.6%). Malignancies were then divided into those found before (Supporting Fig. 1B) and after GISTs (Supporting Fig. 1C). We then compared the occurrence of additional cancers among GIST patients and the US population. GIST patients in the SEER database had a 44% increased prevalence of cancers occurring before the GIST diagnosis (SPR, 1.44; 95% CI, 1.33-1.55) and a 66% increased relative risk of developing cancers after the GIST diagnosis (SIR, 1.66; 95% CI, 1.52-1.81; Fig. 1). Cancers with significantly increased occurrence both before and after the GIST diagnosis included sarcomas, neuroendocrinecarcinoid tumors, colorectal adenocarcinoma, and nonHodgkin lymphoma. Malignancies with a significantly elevated prevalence only before GISTs included esophageal adenocarcinoma, bladder adenocarcinoma, melanoma, and prostate adenocarcinoma. Malignancies with a significantly elevated incidence only after GISTs included small bowel adenocarcinoma, papillary thyroid cancer, renal cell carcinoma, gastric adenocarcinoma, hepatobiliary adenocarcinoma, pancreatic adenocarcinoma, non–small 2962
TABLE 1. Demographic and Clinical Characteristics of the GIST Cohort (n 5 6112) Characteristic Age at GIST diagnosis 20-39 y 40-49 y 50-59 y 60-69 y 70-79 y 80 y Sex Female Male Race White/unknown Black Other Ethnicity Hispanic Non-Hispanic/unknown Year of diagnosis 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 GIST location Esophagus Stomach Small intestine Colorectal Hepatobiliary Pancreas Retroperitoneum Peritoneum, omentum, and mesentery Other digestive organs GIST size 2 cm >2, 5 cm >5, 10 cm >10 cm Unknown
No.
%
325 739 1289 1553 1398 808
5.3 12.1 21.1 25.4 22.9 13.2
2860 3252
46.8 53.2
4320 1079 713
70.7 17.7 11.7
560 5552
9.2 90.8
408 522 505 521 531 494 517 577 604 730 703
6.7 8.5 8.3 8.5 8.7 8.1 8.5 9.4 9.9 11.9 11.5
33 3368 1762 343 5 23 57 126 384
0.5 55.1 28.8 5.6 0.1 0.4 0.9 2.1 6.3
430 1351 1847 1478 1006
7.0 22.1 30.2 24.2 16.5
Abbreviation: GIST, gastrointestinal stromal tumor.
cell lung cancer (NSCLC), and transitional cell carcinoma of the bladder. Among women, there was also an increased post-GIST incidence of ovarian carcinoma, uterine adenocarcinoma, and other genitourinary cancers. There were also several rare cancers that appeared significantly more often in the GIST cohort versus the general US population. However, because some had only 1 reported case in the SEER database, we elected to exclude them from Figure 1. These included prostatic transitional cell carcinoma before GISTs (SPR, 51.5) and squamous cell carcinoma of the breast after GISTs (SIR, Cancer
September 1, 2015
Increased Cancer Risk in GIST Patients/Murphy et al
Figure 1. Cancer development before and after the diagnosis of GIST. The occurrence of each cancer before the diagnosis of GIST is reported with the SPR (solid squares), and the occurrence after the diagnosis of GIST is reported as the SIR (white circles). Horizontal lines illustrate associated 95% CIs. Only cancers with statistically significantly elevated SPRs and SIRs (P 10 cm; in particular, patients with tumors 2 cm had the highest likelihood of having addiCancer
September 1, 2015
tional neoplasms before and after GISTs. Overall, these findings suggest that ethnicity and tumor size affect the risk of developing additional cancers. When we analyzed the period of increased occurrence of additional cancers, we found that the maximum increase occurred within 1 year before and after the GIST diagnosis (Fig. 2). The median latency period from the diagnosis of the first cancer to the diagnosis of GIST was 3.6 years for all patients, although it should be noted that SEER registries with longer follow-up (ie, earlier data collection dates) also had longer median latencies: the median latency was 6.1 years in the 1973 registries, 3.5 years in the 1992 registries, and 1.5 years in the 2000 registries. The median time from the diagnosis of GIST to the diagnosis of a subsequent cancer was 10 months for the entire cohort. 2963
Original Article TABLE 2. Occurrence of Additional Cancers Before and After the Diagnosis of GIST According to Select Patient Characteristics Cancers Before GIST
Age at GIST diagnosis 20-39 y 40-49 y 50-59 y 60-69 y 70-79 y 80 y Sex Female Male Race White/unknown Black Other Ethnicity Hispanic Non-Hispanic/unknown GIST size 2 cm >2, 5 cm >5, 10 cm >10 cm Unknown GIST site Stomach Small intestine Other Registry group 1973-2011 1992-2011 2000-2011
Cancers After GIST
SPR (95% CI)
Pa
Observed
Expected
SIR (95% CI)
Pa
1.8 11.8 48.1 125.5 179 118.5
2.82 1.87 2.16 1.53 1.33 1.20
(0.92-6.58) (1.17-2.83) (1.77-2.62) (1.32-1.76) (1.17-1.51) (1.01-1.41)
.11
6 24 76 144 176 79
2.1 13.7 48.4 97.5 101.3 38.6
2.86 1.75 1.57 1.48 1.74 2.05
(1.05-6.22) (1.12-2.61) (1.24-1.97) (1.25-1.74) (1.49-2.01) (1.62-2.55)
.71
280 423
199.4 285.2
1.40 (1.24-1.58) 1.48 (1.34-1.63)
.57
201 304
117.8 183.8
1.71 (1.48-1.96) 1.65 (1.47-1.85)
.89
526 126 51
369.3 74.8 40.5
1.42 (1.31-1.55) 1.68 (1.40-2.01) 1.26 (0.94-1.66)
.41
357 97 51
228.3 50.4 22.9
1.56 (1.41-1.73) 1.92 (1.56-2.35) 2.23 (1.66-2.93)
.26
33 670
37.6 447
0.88 (0.60-1.23) 1.50 (1.39-1.62)
.02
31 474
22.2 279.4
1.40 (0.95-1.98) 1.70 (1.55-1.86)
.47
91 187 182 126 117
37 109.8 144.7 113.4 79.8
2.46 1.70 1.26 1.11 1.47
.0009
66 119 154 94 72
19.1 71.3 96.8 69.9 44.6
3.46 1.67 1.59 1.34 1.61
(2.67-4.40) (1.38-2.00) (1.35-1.86) (1.09-1.65) (1.26-2.03)
.02
406 192 105
279 126.2 79.4
1.46 (1.32-1.60) 1.52 (1.31-1.75) 1.32 (1.08-1.60)
.73
303 140 62
171.6 86.9 43.1
1.77 (1.57-1.98) 1.61 (1.36-1.90) 1.44 (1.10-1.84)
.57
339 108 256
243.4 76 165.1
1.39 (1.25-1.55) 1.42 (1.17-1.71) 1.55 (1.37-1.75)
.81
183 71 251
106 42.3 153.3
1.73 (1.49-2.00) 1.68 (1.31-2.12) 1.64 (1.44-1.85)
.94
Observed
Expected
5 22 104 192 238 142
(1.98-3.02) (1.47-1.97) (1.08-1.45) (0.93-1.32) (1.21-1.76)
Abbreviations: CI, confidence interval; GIST, gastrointestinal stromal tumor; SIR, standardized incidence ratio; SPR, standardized prevalence ratio. a The P values represent a likelihood ratio test for assessing heterogeneity in SIRs or SPRs between different subgroups of patient characteristics.
DISCUSSION To the best of our knowledge, this is the first populationbased study with present-day SEER ICD-O-3 coding to quantify the occurrence of specific malignancies before and after GISTs. Although some of our results confirmed previous studies, we have also refuted many earlier findings, identified several novel associations, and provided a comprehensive description and statistical examination to support the existence of nonrandom associations between GISTs and other malignancies. Our results showed that the increased occurrence of other cancers was specific to the anatomic site and the histological type, and it had a temporal relation with the GIST diagnosis. In particular, the data pointed to the possibility of many new, clinically relevant associations between GISTs and cancers distinct from known disorders such Carney’s triad/tetrad (eg, gastric GISTs, pulmonary chondromas, extra-adrenal paragangliomas, and 2964
adrenal adenomas),27-29 familial GIST syndromes (eg, activating germline mutations in KIT and PDGFRa with multifocal GISTs),3,30-33 neurofibromatosis type 1 (eg, GISTs, sarcomas, periampullary/pancreatic neuroendocrine tumors, and pheochromocytomas),34-39 and Carney-Stratakis syndrome (eg, GISTs and paragangliomas).40-42 Because sarcomas and neuroendocrine tumors, which are found in neurofibromatosis, constituted only 4.0% and 3.1% of the 1208 additional cancers, respectively, this suggests that other etiologies might be responsible for the observed trends. In corroboration with single-institution reports, our study shows that 17.1% of the GIST patients developed additional cancers. Moreover, anatomic sites of additional malignancies were fairly similar to those reported in previous studies, including 1 multiseries analysis, which found second cancers in 13% of GIST patients and elevated risks of melanoma and GI, lung, and prostate cancers.8 A Cancer
September 1, 2015
Increased Cancer Risk in GIST Patients/Murphy et al
Figure 2. Timeline for the occurrence of additional cancers with respect to the diagnosis of GIST. The occurrence of additional cancers with respect to the time of the GIST diagnosis is represented. The occurrence of cancer before the diagnosis of GIST is quantified with the SPR, and the occurrence of cancer after the diagnosis of GIST is quantified with the SIR. Error bars represent the 95% confidence intervals, and asterisks represent P
