Evaluation of Second Solid Cancers After Hematopoietic Stem Cell Transplantation in European Patients

Key Points

Question  What is the clinical outcome for patients who developed a second solid cancer after a stem cell transplant, when compared with a de novo cancer of the same subtype?

Findings  In this cohort study of 4065 European patients with 18 different second solid cancers after a stem cell transplant, 5-year overall survival was 47%, and after solid second cancer diagnosis, survival was poor in pancreas, lungs, hepatobiliary, esophageal, brain, and gastric cancers.

Meaning  This study provides observational evidence of the utility of screening and counseling for long-term survivors of second cancers.

Importance  Incidence and risk factors of second solid cancers (SSCs) that occur after hematopoietic stem cell transplantation (HSCT) are well documented. However, clinical outcome data of patients who developed an SSC after HSCT are limited.

Objective  To assess the outcome of patients with an SSC occurring after HSCT from the time of SSC diagnosis.

Design, Setting, and Participants  This cohort study used data of 4065 patients from 26 countries registered with the European Society for Blood and Marrow Transplantation, which has maintained clinical data since 1977 of patients who received a transplant. Information from all patients who underwent a transplant in Europe and had an SSC diagnosis between January 1, 2000, and December 31, 2014, was extracted. The cohort included patients with 18 different cancers. Data analysis was conducted from September 3, 2017, to March 17, 2018.

Main Outcomes and Measures  Median and 5-year age-standardized overall survival, causes of death, risk factor multivariate analysis using a clustered Cox proportional hazard regression model, and standardized mortality ratio were calculated for each of the 18 types of SSC.

Results  In total, 220 617 patients underwent a transplant, of whom only 4065 (1.8%) patients with a second solid cancer after HSCT were included in the study. Among the 4065 patients, 2321 (57.1%) were men and 1744 (42.9%) were women, with a mean (range) age of 59.1 (3.2-82.3) years at diagnosis of second solid cancer. The 5-year age-standardized overall survival was 47% (95% CI, 45%-49%). The 5-year overall survival rate after SSC diagnosis was poor for pancreas, lung, hepatobiliary, esophageal, brain, and gastric cancers, with a median survival between 0.6 and 1 year. The 5-year overall survival was intermediate for endometrial, colorectal, sarcomas, ovarian, bladder, oropharyngeal, and kidney cancers, with a median survival between 2 and 10 years. The 5-year overall survival was more favorable for melanoma, breast, prostate, cervix, and thyroid cancers, with a median survival of 10 or more years. Additional transplant-associated factors for mortality for patients treated with allogeneic HSCT were age at transplant, donor type, conditioning regimen, and graft-vs-host disease. In total, 1777 patients (43.7%) died, of which 1256 (74.8%) were from SSC, 344 (20.5%) from primary disease, and 79 (4.7%) from other causes. Standardized mortality ratio was higher, compared with de novo solid cancers, for melanoma, prostate, breast, kidney, bladder, colorectal, and endometrial cancers but not for the other cancers.

Conclusions and Relevance  The outcome of SSC is mainly dependent on the type of second cancer; thus, future studies should investigate the reasons the standardized mortality ratio is higher for some cancers to identify whether patients with these cancers should be treated differently and to help in screening and counseling patients who developed an SSC after HSCT.

Hematopoietic stem cell transplantation (HSCT) is a reasonable treatment option for a variety of malignant and nonmalignant disorders. In the past decades, the clinical outcomes of patients treated with HSCT have substantially improved.¹ With an increasing number of transplants performed yearly and with better outcomes, the number of patients surviving the acute phase of a transplant is steadily growing. By 2030, an estimated half a million people are expected to survive long term from HSCT in the United States and probably more than 1.5 million worldwide.²

Late complications are frequent in HSCT survivors, and life expectancy is not fully restored.^3-5 Second solid cancers (SSCs) are among the most feared late complications in HSCT survivors and are associated with substantial morbidity and mortality.^6-10 Among patients who survive 5 years or longer after allogeneic HSCT, second cancer is the main cause of death.¹¹ Incidence and risk factors for SSC after HSCT have been evaluated.^7,12-16 The cumulative incidence of developing SSC after transplant increases continually between 2% and 6% at 20 years and is substantially higher when compared with the incidence in the general population. Nonsquamous cell cancers, such as breast and thyroid, are strongly associated with exposure to radiation,^7,15,17 whereas squamous cell carcinoma of the skin and oropharyngeal region is associated with chronic graft-vs-host disease (GVHD).¹² However, outcome data of long-term survivors from the time of diagnosis of an SSC are scant.^18-20

The aim of the present study was to evaluate the outcome of patients with an SSC that occurred after HSCT, using a large cohort of patients from the 26 countries that participate in the European Society for Blood and Marrow Transplantation (EBMT) registry. We estimated overall survival (OS) of the different types of SSC from the time of diagnosis, evaluated the causes of death, investigated risk factors affecting OS, and compared the outcome of the different SSCs with a cohort of patients who developed the same cancer type but were not pretreated with HSCT.

This retrospective cohort study from the Late Effects Subcommittee of the Transplant Complications Working Party of the EBMT was based on the mandatory minimum data set reported to the EBMT Registry. This registry contains clinical data since 1977, including diagnosis, transplant, complications, and outcome, of patients who received a transplant. Patient data are reported exhaustively, and patients are followed up indefinitely. This study received approval from the Transplant Complications Working Party Review Board of the EBMT. Patient informed consent was not obtained because, according to the guidelines of the registry, it is the transplant center's responsibility to ensure that the patient has consented to the data collection before data are forwarded to the registry. Data analysis was conducted from September 3, 2017, to March 17, 2018.

We extracted data from the EBMT registry on all patients who received a transplant in Europe (including Turkey) and had an SSC diagnosis between January 1, 2000, and December 31, 2014. Cancers were defined by site according to the International Classification of Diseases for Oncology, 3rd Edition.²¹ We evaluated all patients together as well as patients treated with allogeneic and autologous HSCT separately. We restricted the analysis to patients who developed an SSC; nonmelanoma skin cancers were not included because these cancers in patients with a transplant, as in the general population, are nonlethal malignant neoplasms (thus, standardized mortality cannot be computed). Patients with congenital bone marrow failure or other congenital diseases, histiocytic disorders, autoimmune diseases, undefined cancer, and cancer diagnosis at autopsy as well as patients with missing data on type of transplant were excluded. For patients who developed more than 1 SSC after HSCT, the outcome was assessed from time of diagnosis of first SSC. No independent central review on diagnosis of SSC was performed.

We estimated the outcome of SSC occurring after HSCT from the time of cancer diagnosis, irrespective of cancer stage. The end points were OS since SSC diagnosis, cumulative incidence and risk factors of death from SSC, death rates, causes of death, and standardized mortality ratio (SMR) and absolute excess risk (AER) compared with de novo cancers.

Causes of death were defined as the primary disease, the SSC, or any other cause. Because of a large number of different primary diseases for which patients received a transplant, we used the Disease Risk Index (DRI) described by Armand et al,²² which combines both diagnosis and status prior to transplant, to risk-stratify the patients into 3 groups: low, intermediate, or high risk.

Patient demographics and disease characteristics were described using median and range for continuous variables and counts and percentages for categorical variables. Follow-up was calculated using the reverse Kaplan-Meier method, with the event indicator reversed (the end point being the loss of follow-up instead of death). For the estimation of survival outcome, 5-year age-standardized OS was calculated individually for each type of SSC after diagnosis. To estimate risk factors for survival following a diagnosis of SSC, a multivariate risk factor analysis was performed to calculate the hazard ratio (HR) with 95% CI; a clustered Cox proportional hazard regression model allowed us to take into account geographic heterogeneity. Variables evaluated were the site of SSC (thyroid cancer as the reference), age at transplant, DRI of the primary disease, donor type, interval between HSCT and diagnosis of SSC, calendar year of SSC diagnosis, conditioning, and GVHD before the diagnosis of SSC. The risk factor analysis was done for allogeneic and autologous HSCT and was done separately for men and women (different types of SSC). Only cancers with at least 20 cases in each group were evaluated in the multivariate analysis.

For each type of cancer, we calculated the death rate because of the SSC after HSCT, compared with the expected death rate for the corresponding cancer in the general population. The estimates of the expected death rates are based on data extracted from population-based cancer registries in Europe, members of the European Network of Cancer Registries (https://www.encr.eu/).²³ We obtained the SMR by calculating the ratio of observed to expected number of cases by 1000 person-years. The 95% CI were estimated using the Mantel-Haenszel test. The AER, an estimate in absolute number per 1000 person-years of observed excess deaths that can be attributed to SSC, was calculated by subtracting the number of expected deaths from observed deaths, dividing by person-years of follow-up for the HSCT cohort. For sarcomas, SMR and AER could not be calculated because no data on the European reference cancer population were available.

The cumulative incidence of death from SSC after HSCT was calculated by considering any other cause of death, such as primary disease and toxic effect, as a competing event. The time to risk was computed from the date of diagnosis of SSC to the date of death or the date of last contact. Cumulative incidence of death from cancer was estimated for all patients and for each cancer type.

Group differences were analyzed using the unpaired Mann-Whitney test for continuous and the χ² test for categorical variables. All P values were 2-sided, with a statistical significance set at P = .05. Statistical analyses were performed with SPSS Statistics, version 22 (IBM Corp), and R, version 3.2.3 (R Development Core Team).

In total, 220 617 patients were treated with a transplant, with 83 834 (37.9%) receiving allogeneic HSCT and 138 988 (62.9%) autologous HSCT. Among the 220 617 patients, 4065 (of whom 2321 [57.1%] were men and 1744 [42.9%] were women, with a mean [range] age of 59.1 [3.2-82.3] years at diagnosis of second solid cancer and 53.2 [0.7-70.0] years at transplant) had 18 different types of SSC and were included in the study: 1443 patients (35.5%) were treated with allogeneic HSCT and 2622 (64.5%) with autologous HSCT. The 5 most frequent SSCs were lung (n = 597), breast (n = 547), colorectal (n = 446), prostate (n = 410), and melanoma (n = 343) (Table). The median (range) time of follow-up after SSC diagnosis was 3.65 (3.48-3.88) years: 3.86 years for the allogeneic HSCT group, and 3.58 years for the autologous HSCT group. The completeness of follow-up was 68%; 202 patients (4.9%) were lost to follow-up. The characteristics of the 4065 patients are shown in eTable 1 in the Supplement.

The OS at 5 years (Figure, A) after a diagnosis of SSC was 47% (95% CI, 45%-49%); it was 51.5% (95% CI, 48.4%-54.8%) in the allogeneic group and 44.5% (95% CI, 42.1%-47.0%) in the autologous group. The OS was mainly dependent on the type of SSC. The 5-year age-standardized survival rates of the different SSCs are shown in the Table and eFigure 1A in the Supplement. The 5-year OS after SSC diagnosis was poor for pancreas (8%; 95% CI, 3%-18%), lung (14%; 95% CI, 11%-19%), hepatobiliary (18%; 95% CI, 11%-31%), esophageal (21%; 95% CI, 13%-36%), brain (21%; 95% CI, 15%-30%), and gastric (29%; 95% CI, 21%-39%) cancers, with a median survival between 0.6 and 1 year. The 5-year OS was intermediate for endometrial (40%; 95% CI, 26%-63%), colorectal (41%; 95% CI, 36%-48%), sarcomas (42%; 95% CI, 34%-51%), ovarian (43%; 95% CI, 32%-58%), bladder (49%; 95% CI, 39%-62%), oropharyngeal (53%; 95% CI, 46%-62%), and kidney (55%; 95% CI, 47%-65%) cancers, with a median survival between 2 and 10 years. Five-year OS was more favorable for melanoma (68%; 95% CI, 62%-74%), breast (69%; 95% CI, 64%-74%), prostate (69%; 95% CI, 64%-75%), cervix (70%; 95% CI, 57%-86%), and thyroid (83%; 95% CI, 76%-92%) cancers, with a median survival of 10 or more years. The outcome of SSC after allogeneic HSCT did not essentially differ from the outcome of patients treated with autologous HSCT (Table), with the exception of endometrial cancer and, to a lesser extent, sarcomas (eFigure 1B in the Supplement).

In total, 1777 (43.7%) of 4065 patients with an SSC died; information on cause of death was missing for 99 patients (2.4%). Of the remaining 1678 deaths, 1256 (74.8%) were from SSC, 344 (20.5%) from primary disease, and 79 (4.7%) from other causes (ie, 13 cardiac, 2 hemorrhage, 2 failure or rejection, 33 infection, 14 GVHD, 2 transplant-associated, and 13 other causes).

The percentage of deaths was strongly associated with the type of SSC. It ranged from 12.8% to 28.1% in the favorable outcome group, from 37.3% to 51.9% in the intermediate outcome group, and from 59.5% to 76.6% in the poor outcome group (eTable 2 in the Supplement). Overall, 1256 deaths (74.9%) were from the SSC itself. However, deaths from SSC were less frequent in the cancers with favorable outcome (191 [55.2%] of 346 deaths) compared with the intermediate outcome group (361 [69.2%] of 522 deaths) and poor outcome group (704 [86.9%] of 810 deaths).

Age at HSCT and type of SSC were risk factors for death in all patients (allogeneic and autologous as well as men and women). In the allogeneic group, additional risk factors in male patients were intermediate DRI, unrelated matched donor, conditioning with total body irradiation, and GVHD; in female patients, the risk factor was shorter interval between HSCT and SSC diagnosis (eTable 3 in the Supplement). In the autologous group, additional risk factors were high DRI at HSCT and longer interval between HSCT and SSC diagnosis in men and diagnosis of SSC between 2000 and 2008 (compared with 2009 and later) in women.

The observed and expected cancer death rates per 1000 person-years, SMR, and AER for the different types of SSC are shown in eTable 4 in the Supplement. A significant increase in SMR with excess mortality was observed for prostate, breast, melanoma, kidney, bladder, colorectal, and endometrial cancers, compared with de novo cancers in the general population. Meanwhile, SMR equal to 1 was seen for thyroid, cervix, oropharyngeal, ovarian, gastric, brain, and lung cancers. Decreased SMR was observed for esophageal, hepatobiliary, and pancreatic cancers (eFigure 2 in the Supplement).

Few differences were seen between patients treated with allogeneic HSCT and those who received autologous HSCT (eTable 4 in the Supplement). In contrast to the whole cohort, patients treated with allogeneic HSCT had no excess of deaths linked to prostate (SMR, 1.49; 95% CI, 0.79-2.37) and kidney (SMR, 0.92; 95% CI, 0.4-1.62) cancers; autologous HSCT recipients had no excess of deaths linked to breast (SMR, 1.26; 95% CI, 0.85-1.75) and bladder (SMR, 1.48; 95% CI, 0.81-2.32) cancers. A decreased SMR in autologous HSCT was only seen for hepatobiliary (SMR, 0.60; 95% CI, 0.41-0.83) and pancreas (SMR, 0.44; 95% CI, 0.34-0.55) cancers but not esophageal (SMR, 0.99; 95% CI, 0.59-1.49) cancer.

The cumulative incidence of death from SSC at 5 years was 39% (95% CI, 37%-40%) (Figure, B). The cumulative incidence of all SSC types is shown in the Table, and the cumulative incidence of death from breast (13%; 95% CI, 10%-17%), colorectal (39%; 95% CI, 33%-44%), and lung (74%; 95% CI, 69%-78%) cancers are shown in Figure, C.

Our data from a large cohort of European patients demonstrate that outcome of SSC after HSCT depends strongly on the type of cancer. We were able to identify 3 prognostic groups of SSC diagnosis after HSCT. Thyroid, cervix, prostate, and breast cancers as well as melanoma had a more favorable outcome, with a median survival of 10 years or longer and a mean 5-year age-standardized OS between 68% and 83%. Sarcomas, kidney, oropharyngeal, bladder, ovarian, colorectal, and endometrial cancers had an intermediate outcome, with a median survival longer than 1 year but shorter than 10 years and a mean 5-year OS between 40% and 55%. Gastric, brain, esophageal, hepatobiliary, lung, and pancreas cancers had a poor outcome, with a median survival of 1 year or less and a mean 5-year OS between 8% and 21%. The main cause of death was SSC after diagnosis, accounting for 75% of all deaths. However, depending on the type of SSC, considerable variation was observed. Cancer was the cause of death in approximately half the patients with a favorable SSC, but the percentage approached 90% in patients with a poor outcome SSC. The main transplant-associated risk factors for mortality after SSC diagnosis were age at transplant, donor type, conditioning, and GVHD for patients treated with allogeneic HSCT. Solid cancers, which are common in the nontransplant population, are also the most common in the posttransplant population.

In the European general population, important disparities in survival exist among different de novo cancer types, with a mean 5-year relative survival ranging from 86% for thyroid cancers to 5.5% for pancreas cancers.^24,25 The ranking of survival of the SSC types from our cohort was similar, with a mean 5-year age-standardized survival of 83% for thyroid cancers and 8% for pancreas cancers. For a number of SSCs, including cervix, oropharyngeal, ovarian, gastric, brain, and lung, the death rate did not exceed that expected in a nontransplant cancer population. Mortality excess was found for prostate, breast, melanoma, kidney, bladder, colorectal, and endometrial cancers, with a 1.4- to 5.5-fold increased risk of death. These findings are in line with published results from the Center for International Blood and Marrow Transplant Research on survival in a cohort of 112 patients who developed a solid tumor after HSCT.¹⁹ Of note, the outcome of many SSC types was similar for patients treated with autologous and allogeneic HSCT. A discrepancy was found, however, with better outcome after allogeneic HSCT for endometrial, sarcoma, gastric, melanoma, and kidney cancers. A possible explanation could be the increased adherence to recommended guidelines for health care practices after allogeneic transplant.²⁶

This study was not designed to evaluate the reasons for the differences in mortality rate as compared with de novo cancers, but several explanations for an increased SMR can be considered. First, the biologic behavior of an SSC and its responsiveness to treatment might differ from the behavior of de novo cancers. In a large population-based study of adolescents and young adults who did not receive a transplant, survival was substantially worse for patients with SSC compared with patients with a similar primary cancer at the same age.²⁷ Keegan et al²⁷ presumed a relative aggressiveness of the SSC over the primary cancer. High-risk histologic subtypes of cancers occurring after radiation therapy have been reported.²⁸ Second, patient preferences can change over time in case of an SSC diagnosis after HSCT, with orientation to less aggressive or palliative treatment strategies and movement away from curative options. Changes in patient expectations, including acceptance of shortened life expectancy to reduce severe adverse effects, have been reported for patients with advanced-stage cancers.^29-31 Third, transplant-associated factors, such as GVHD and its immunosuppressive treatment, donor type, and altered cancer immunity after HSCT, could interfere with the prognosis of an SSC. After solid organ transplant, the risk of SSC is increased and presents a major cause of death, especially because of the decreased immunologic reactivity during chronic immunosuppression.^32,33 Patients exposed to chronic immunosuppression after solid organ transplant developed colon cancer at a substantially younger age and had worse outcome compared with patients with the same cancer in the general population.³⁴ Reasons for the decreased SMR in pancreatic, hepatobiliary, or esophageal cancers, all of which are associated with a poor prognosis, are more difficult to explain. Immunomodulatory properties of the allogeneic cells do not seem to be involved, given that no difference exists between allogeneic and autologous HSCT. For such aggressive cancers, survival is greatly affected by disease stage. We cannot exclude that diagnosis of SSC was done earlier, at a lower stage.

Thus far, an increase in gastrointestinal cancers has not been reported after HSCT. In patients with cancer who did not receive a transplant, however, an excess of second colon cancer has been repeatedly reported. Childhood cancer survivors who were treated with radiation on the abdomen are at greater risk for colorectal cancer.³⁵ The risk of second colorectal cancer is also higher among Hodgkin lymphoma survivors who had received infra-diaphragmatic radiation and high doses of procarbazine hydrochloride.³⁶ Moreover, an excess of second colon cancer has been reported for prostate cancer after external radiation but not after treatment without radiation.³⁷ To our knowledge, we reported for the first time not only a high number of second colorectal cancers after HSCT but also that these second colorectal cancers had a worse prognosis (increased SMR) compared with de novo cancers in the general population.

The study has several limitations, mainly because of its retrospective design and the follow-up time from diagnosis of SSC. We lacked detailed information specific to the SSC, namely their stage and treatment. In addition, the length of follow-up from SSC diagnosis was short at 3.65 years. This study was deliberately restricted to patients who developed SSC after 2000 to enable us to evaluate outcomes with modern cancer treatments, because these data are of particular value to the hematology-oncology community. The short follow-up also reflects, in part, the number of deaths after SSC diagnosis; 30% of patients had poor-risk cancers associated with a median survival of less than 1 year, and nonmelanoma skin cancers, which do not affect survival, were excluded. This is not the case for other studies that have examined survival from cancer after HSCT.¹⁹ Despite this, completeness of follow-up in this study was nearly 70%, with only 5% of patients lost to follow-up. To our knowledge, no other transplant registry exists that combines such a high number of patients with a high degree of relevant information, allowing the consideration of the outcome of an SSC and transplant-associated factors.

This study demonstrated that the outcome of an SSC after HSCT is mainly dependent on the type of malignant neoplasm. Demographics and transplant-associated factors are also relevant, however. Survival is worse for some specific types of SSC, but overall these patients have comparable chances of survival to those of patients with the same de novo cancer. In light of increased SMR for some cancers, further investigations should consider whether patients with SSC after HSCT must be treated differently from the general population. This study will require close cooperation between oncologists and hematologists.

Accepted for Publication: August 23, 2018.

Corresponding Author: André Tichelli, MD, Hematology, University Hospital Basel, Petersgraben 4, 4031 Basel, Switzerland (tichelli@datacomm.ch).

Published Online: November 21, 2018. doi:10.1001/jamaoncol.2018.4934

Author Contributions: Dr Tichelli had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Tichelli, Beohou, Labopin, Socié, Rovó, van Biezen, Duarte, Basak, Salooja.

Acquisition, analysis, or interpretation of data: Tichelli, Beohou, Labopin, Rovó, Badoglio, van Biezen, Bader, Salooja.

Drafting of the manuscript: Tichelli, Socié, Rovó, van Biezen.

Critical revision of the manuscript for important intellectual content: Tichelli, Beohou, Labopin, Rovó, Badoglio, Bader, Duarte, Basak, Salooja.

Statistical analysis: Beohou, Labopin, Rovó, van Biezen, Salooja.

Administrative, technical, or material support: Tichelli, Rovó, Badoglio, van Biezen, Basak, Salooja.

Supervision: Tichelli, Socié, Rovó, Duarte, Basak, Salooja.

Conflict of Interest Disclosures: None reported.

The Transplant Complications Working Party of the EBMT: Grzegorz W. Basak (chairman), Zinaida Peric (secretary), Mahmoud Aljurf, Manueal Badoglio, Eric Benohou, Raphael F. Duarte, Enric Carreras, Corien Eeltink, Raffaella Greco, Kate Hill, Diana Greenfield, Hildegard Greinix, Alenca Harrington, Myriam Labopin, Maria Teresa Lupo-Stanghellini, Jacopo Mariotti, Carmen Martinez, Anne Nihtinen, Olaf Penack, Alicia Rovó, Tapani Rutuu, Nina Salooja, Helene Schoemans, Gérard Socié, Teresa Solano Moliner, André Tichelli, Rosario Varella, Anke Verlinden, and Daniel Wolff.

Additional Contributions: We thank all the transplant centers of the European Society for Blood and Marrow Transplantation (EBMT) that regularly submit their patients’ data and the EBMT office that provided the megafile of all registered patients.