Pancreatic cancer among solid organ transplant recipients in the United States

Pancreatic cancer (PC) in solid organ transplant (SOT) recipients is not well studied. Some PC cases may be incidentally detected during hepatobiliary imaging. We evaluated PC among 374,106 SOT recipients during 1995–2017 in the United States using linked data from the national transplant registry and multiple state/regional cancer registries. Standardized incidence ratios (SIRs) were used to compare PC risk in recipients to the general population. We used multivariate Poisson regression to identify independent risk factors for PC. We assessed survival after PC diagnosis using Kaplan–Meier curves and log-rank tests. SOT recipients had elevated incidence for PC compared with the general population (SIR 1.40, 95% CI 1.29–1.52), and this increase was strongest in liver recipients (1.65, 1.41–1.92). Among all recipients, PC incidence was especially increased for cases arising in the head of the pancreas (SIR 1.50, 95% CI 1.34–1.68) and for cases diagnosed at localized stage (1.85, 1.37–2.44). Among SOT recipients, factors independently associated with increased incidence were consistent with those in general population including male sex, older age, non-O blood type, and history of diabetes. Additionally, compared to other organ recipients, liver transplant recipients had higher PC incidence (adjusted incidence rate ratio 1.28; 95% CI 1.06–1.54). Overall survival after PC diagnosis was poor (median 4 months) and similar between liver and other organ transplant recipients (p = 0.08). PC incidence is elevated among SOT recipients, and more commonly diagnosed in liver transplant recipients perhaps related to incidental detection. However, survival is poor even in liver recipients, arguing against routine PC screening.


Introduction
Pancreatic cancer (PC) is the third leading cause of cancerrelated mortality in the United States, accounting for an estimated 48,220 deaths in 2021 (Siegel et al. 2021), and is projected to become the second leading cause of cancer death by 2030 (Rahib et al. 2014). According to data from US Surveillance, Epidemiology, and End Results (SEER) cancer registries for 2011-2017, the 5-year relative survival for PC overall was 11%, ranging from 42% for localized PC down to 3% for distant metastatic PC (Force et al. 2019). The current guideline from the US Preventive Services Task Force indicates that screening for PC is of unproven benefit for the general population, because there is a lack of evidence that screening reduces morbidity or mortality (Force et al. 2019) However, a recent study suggested the usefulness of early detection of PC as an incidental finding (i.e., detection of PC in the absence of symptoms or screening) during surveillance imaging in patients with chronic liver disease (Kumagi et al. 2019). A study in Japan found that over 50% of early-stage PCs are found incidentally during evaluations for other conditions (Kanno et al. 2018).
Several risk factors for PC have been identified from studies in the general population (Park et al. 2021). In addition to hereditary factors (such as genetic variants of BRCA1/2, ATM, MLH1, MSH2, MSH6, or PMS2) (Rainone et al. 2020), cigarette smoking has been reported to have the strongest association with PC among various modifiable risk factors (Maisonneuve and Lowenfels 2015). Other risk factors for PC include older age, chronic pancreatitis, long-standing history of diabetes, overweight and obesity, non-O blood type, and cystic fibrosis Lowenfels 2015, 2010;Genkinger et al. 2009;Iodice et al. 2010;Maisonneuve et al. 2007). Heavy alcohol consumption is a suggested risk factor (Lucenteforte et al. 2012), but the evidence remains inconclusive.
Solid organ transplantation (SOT, including liver, kidney, pancreas, heart, and lung transplantation) is a lifesaving procedure for individuals with end-stage organ dysfunction. In the US, 32,313 deceased donor transplants were performed in 2019; this number has been increasing from 28,463 transplants in 2009 (Israni et al. 2021). In addition, transplant outcomes have improved substantially over time due to improvements in surgical technique, donor and recipient conditioning, intensive care treatment, and immunosuppressive regimens (Rana et al. 2015). Following SOT, however, recipients are universally exposed to immunosuppressive medications used to prevent rejection, which places them at risk of complications including infection and malignancy (Engels et al. 2011).
Among SOT recipients, PC risk is modestly increased, with relative risks of up to 2.9 in studies with an adequate sample size (Engels et al. 2011;Grulich et al. 2007;Collett et al. 2010;Tsai et al. 2017). Risk factors for PC arising post-transplant are not well understood but could include recipient characteristics and factors linked to immunosuppression. There are few studies examining risk factors for PC among SOT recipients. Corral et al. recently found increased risk of PC in thoracic organ recipients using data from the US SOT network (Corral et al. 2021). However, other potential risk factors have not been assessed.
In the present study, we used data from the Transplant Cancer Match (TCM) Study to describe risk factors for PC among SOT recipients in the United States (US). We evaluated for evidence that medical examinations of SOT recipients, especially resulting from abdominal imaging procedures typically performed among liver recipients, are associated with earlier PC diagnosis. We also assessed survival following a PC diagnosis.

Methods
The TCM Study is described in detail elsewhere (https:// trans plant match. cancer. gov/) (Engels et al. 2011). The cohort consists of SOT recipients characterized through linkage of the Scientific Registry of Transplant Recipients (SRTR) and multiple state and regional cancer registries. The SRTR provides information on all US SOT recipients beginning in 1987 including demographics, medical characteristics, transplanted organs, and induction and baseline maintenance immunosuppressive medications. Record linkages were completed between the SRTR and 32 state cancer registries, covering the period 1995-2017. We included SOT recipients who resided in one of the cancer registry regions at the time of their transplant during the period of cancer registration (approximately 68% of all US transplants performed during this period). We excluded recipients who had a previous PC diagnosis before transplant. This study was exempted from human subjects' review at the National Institutes of Health and approved, as required, at participating cancer registries.
We evaluated PC incidence from the date of transplantation until the first of the following events: failure of a transplanted organ, a subsequent transplant, death, loss to follow-up by the SRTR, or end of cancer registry surveillance (December 2017).
We compared PC incidence in SOT recipients to the general population by calculating standardized incidence ratios (SIRs) as the number of PCs observed in the transplant cohort divided by the number expected in the general population. Expected counts were estimated using rates from 13 SEER cancer registries (https:// seer. cancer. gov/ regis tries/ terms. html), specific to attained age, sex, race/ ethnicity, and attained calendar year. We present SIRs for PC overall and according to transplanted organ (classified as liver with or without other organs [referred to simply as "liver"], kidney, pancreas and/or small intestine ["other abdominal organs"], or only heart and/or lung ["thoracic organ"]), cancer stage per SEER (localized, regional, distant, unstaged/missing), and anatomical subsite per SEER (head, body, tail, other/overlapping/unspecified).
We used Poisson regression to compare PC incidence among subgroups of SOT recipients, based on baseline information provided in the SRTR. We excluded SOT recipients with age < 18 at transplantation from these regression analyses because there were no PC cases in this age group. We conducted univariate analyses that evaluated demographic characteristics (age at transplantation, sex, race/ethnicity, body mass index [BMI], blood type), transplanted organ, living vs. deceased donor status for kidney transplants, induction and baseline maintenance immunosuppressive medications, time since transplantation, and history of diabetes mellitus. Factors significantly associated with PC incidence in univariate Poisson models, or otherwise believed to be clinically important (i.e., year of transplant), were included in multivariable models. In addition, we conducted similar analyses in which the outcomes were subsets of PC cases classified according to cancer stage.
We assessed overall survival among SOT recipients following PC diagnosis. We used Kaplan-Meier curves and log-rank tests to compare survival for cases according to transplanted organ and stage at diagnosis. For these analyses, we excluded cases identified on death certificate or autopsy and those with no follow-up time to assess survival. SAS (version 9.3, SAS Institute, Cary, NC) was used for statistical analyses. Kaplan-Meier curves were constructed with R software (version 4.1.2; R Foundation for Statistical Computing, Vienna, Austria). We used a p value of 0.05 to determine statistical significance; no adjustment for multiple comparisons was made because our analyses were considered exploratory.

Results
A total of 401,502 transplant procedures among 374,106 SOT recipients were included (Table 1). The majority of transplants occurred in males (61.8%) and non-Hispanic whites (60.1%). The median age at transplantation was 50 years. Kidney was the most commonly transplanted 1 3 organ (58.4%), followed by liver (21.1%), and heart and/or lung (14.3%). SOT recipients were followed for a total of 2,047,688 person-years (median follow-up 4.4 years, interquartile range 1.7-8.2 years). During the follow-up period, 574 PC cases were diagnosed (Table 2). PC incidence was 40% higher than in the general population (SIR 1.40, 95% CI 1.29-1.52). This elevation in risk appeared greatest in liver transplant recipients (SIR 1.65, 95% CI 1.41-1.92), followed by recipients of other abdominal organs (1.35, 1.21-1.51), while the elevation was smallest and nonsignificant in thoracic organ recipients (1.23, 0.99-1.52). When considered by anatomic subsite, PC incidence was most elevated for cases arising in the head of the pancreas (SIR 1.50, 95% CI 1.34-1.68) and also elevated for PC in the tail of the pancreas (1.35, 1.06-1.69) but not for cases in the body (1.09, 0.82-1.42).
The largest group of PC cases were diagnosed at distant stage (n = 271, 47.2%), but we saw the greatest elevation relative to the general population for local stage tumors (SIR 1.85, 95% CI 1.37-2.44) followed by regional stage tumors (1.57, 1.36-1.80) and then distant stage tumors (1.17, 1.03-1.31). Similarly, SIRs were consistently elevated both in liver recipients and recipients of other abdominal organs for local stage PC (SIR 2.71, 95% CI 1.58-4.33 and 1.77, 1.18-2.56, respectively) and regional PC (1.94, 1.49-2.47 and 1.47, 1.21-1.76, respectively); this pattern was not observed in thoracic transplant recipients. SIRs for distant stage PC were not significantly elevated in any of these groups (Table 2).
Univariate associations of demographic and transplantrelated factors with PC incidence are shown in Table S2. In multivariable models (Table 3), PC incidence was higher for males than females [adjusted incidence rate ratio (aIRR) 1.47, 95% CI 1.23-1.77] and increased with age at transplantation. Liver recipients showed a significantly increased incidence of PC relative to other organ recipients combined (adjusted IRR 1.28, 95% CI 1.06-1.54). In univariate analyses, there was an increase in PC incidence following the first year post-transplant, after which incidence appeared steady, and in the multivariable model (Table S2), we confirmed that PC incidence was higher after the first year post-transplant (Table 3). Hispanic race/ethnicity was associated with reduced risk in univariate analysis (p = 0.004, compared with non-Hispanic white race/ethnicity), but this association was not significant in multivariable analysis (not shown).
As shown in Table 3, blood type O was associated with reduced PC incidence (aIRR 0.73, 95% CI 0.61-0.86), while a history of diabetes mellitus was associated with increased incidence (1.30, 1.09-1.55). Neither baseline maintenance immunosuppressive regimen (tacrolimus and/or mycophenolate mofetil, cyclosporine and/or azathioprine, vs. others) nor induction therapy was associated with PC incidence. Maintenance immunosuppression with a mammalian target of rapamycin inhibitor (mTORi) was associated with reduced incidence of PC in univariate analysis (p = 0.02). However, this association was not significant in a multivariable analysis (aIRR 0.68, 95% CI 0.44-1.05, p = 0.066), so it was not included in the final model. In a univariate analysis, kidney recipients had lower PC incidence if they received grafts from living donors than from deceased donors (unadjusted incidence rate ratio 0.69, 95% CI 0.54-0.87, p = 0.002) (Table S2), but a significant association was not observed after adjustment (aIRR 0.88, 0.69-1.12, p = 0.30) so it was not included in the final model. When we examined predictors of PC incidence according to tumor stage in multivariable models, we observed that liver recipients had higher incidence than recipients of other organs for local/ regional stage PC (aIRR 1.45, 95% CI 1.11-1.89) but not for distant stage PC (1.11, 0.84-1.46) (Tables S3 and S4). Survival was assessed in 554 SOT recipients with PC. After PC diagnosis, the median overall survival was 4 months. As shown in Fig. 1, survival was better for patients with local/regional stage PC than for patients with distant stage PC (p < 0.0001) but was similar for liver recipients and recipients of other organs (p = 0.08). In addition, when we stratified the analysis by stage at diagnosis, survival did not differ according to the transplanted organ for local/regional stage PC (p = 0.08) or for distant stage PC (p = 0.78; Fig. 2).

Discussion
In this nationwide cohort of recipients of over 400,000 solid organ transplants in the US, we found a modestly increased incidence of PC compared to the general population (SIR 1.40, 95% CI 1.29-1.52). This finding is similar to the SIR of 1.5 observed in our prior overview of cancer risk in the TCM Study, which was based on follow-up through 2008 and included a much smaller number of PC cases (n = 157) (Engels et al. 2011). Of note, PC incidence was most elevated for cases arising in the head of the pancreas. Also, compared to other organ recipients, liver recipients had higher PC incidence, especially for local/regional stage PC. Demographic factors associated with increased incidence were consistent with those in the general population, including male sex, older age, blood type other than O, and history of diabetes mellitus. We also demonstrated that survival after PC detection was very poor among this study population.
In general, epidemiological studies have shown that cancer incidence is elevated in immunocompromised populations including SOT recipients exposed to longterm immunosuppressive therapy. Individuals with human immunodeficiency virus (HIV) infection have been studied as another immunocompromised population. A 2007 systematic review of cancer risk in people with HIV infection reported SIRs for PC ranging from 0.70 to 2.86 (Grulich et al. 2007). The most recent data from US population-based HIV registries demonstrated a slightly increased incidence of PC (SIR 1.13, 95% CI 1.01-1.26) (Hernandez-Ramirez et al. 2017). Other recent registry-based studies of SOT 1 3 recipients from Europe and Asia reported SIRs for PC that were similar to ours but with relatively wide confidence intervals (1.5, 1.0-2.1, and 1.58, 0.51-4.90, respectively) (Collett et al. 2010;Tsai et al. 2017). We observed the most elevated SIR in liver recipients, and PC incidence was higher in this group than other SOT recipients in multivariable analyses. Our result differs from those recently reported by Corral et al., who found a higher incidence of PC among heart and/or lung recipients than liver recipients (Corral et al. 2021). However, that study relied on cancer diagnoses reported by transplant centers, which may be inaccurate (Yanik et al. 2016), and did not formally test whether these incidence estimates differed or adjust for other factors.
Because SOT recipients (especially liver recipients) frequently undergo hepatobiliary-related laboratory and radiology tests, we hypothesized that some of the PC cases in SOT recipients were incidentally detected, in which case they would tend to be diagnosed at an early stage. This hypothesis was supported by the following study findings: (1) the SIR appeared more strongly elevated for PCs arising in the head of the pancreas than in the body/tail, and this pattern was more evident in liver recipients than others; (2) there existed a shift in stage to more localized PC tumors among SOT recipients overall, with the greatest elevation of SIR for localized stage tumors (SIR 1.85, 95% CI 1.37-2.44) followed by regional stage tumors (1.57, 1.36-1.80) and distant stage tumors (1.17, 1.03-1.31); and (3) liver recipients had higher incidence of local/regional stage PC than recipients of other organs, but this association was not seen for distant stage PCs.
Since long-term use of immunosuppressive medication is an important clinical characteristic of SOT recipients, we also evaluated duration of immunosuppression and immunosuppression regimen as risk factors for PC. We did not see a progressive increase over time following transplantation (Table S2), arguing against a role for long-term immunosuppression. Instead, incidence increased after the first year post-SOT and then remained constant, which may be a result of pretransplant medical evaluation. Specifically, prior to being placed on the transplant waitlist, individuals who receive SOT have undergone rigorous medical evaluation, including cancer screening, laboratory measurements, and probably, cross-sectional abdominal imaging (especially for liver and other abdominal organ recipients). It is therefore likely that early asymptomatic PC cases would have been detected, which would have led to removal from the waitlist of individuals whose PC might have otherwise become clinically evident in the first year following transplant. Additionally, induction and common baseline maintenance Results are shown as Kaplan-Meier curves with 95% CI confidence intervals for liver recipients and recipients of other organs. P values are from log-rank tests medication combinations were not associated with PC risk. Those findings are consistent with our hypothesis that liver recipients, who are usually treated with the least exposure to immunosuppression (Levitsky 2019), manifest the highest SIRs, again presumably due to the frequent laboratory and radiological tests performed posttransplant. Our analyses suggested that mTORis were associated with reduced PC incidence, which might be due to direct anticancer effects of mTORis (Meric-Bernstam and Gonzalez-Angulo 2009) or reduction in the use of calcineurin inhibitors. mTORis reduce the risk of skin cancer among SOT recipients and may be associated with reduced risk of other cancers (Yanik et al. 2015;Wolf et al. 2018).
We also found that other established factors were associated with PC incidence in SOT recipients. Blood type O has been identified as a protective factor for PC, although the mechanism is unknown (Jellas et al. 2017;Liu et al. 2019). Of note, the current prevalence of blood type O in SOT recipients is 44% (https:// optn. trans plant. hrsa. gov/ data/ view-data-repor ts/ natio nal-data/#, accessed on 12/7/2021), which is similar to the prevalence (38%) in the general population (American Red Cross; https:// www. redcr ossbl ood. org/ donate-blood/ blood-types/o-blood-type. html, accessed on 12/7/2021). Thus, it is unlikely that the blood type distribution in SOT recipients explains the elevated SIR for PC.
Diabetes mellitus is associated with development of PC in the general population (Kaleru et al. 2020) with a pooled relative risk of 2.1 reported in a 1995 meta-analysis (Everhart and Wright 1995). Diabetes at the time of transplantation was present in 28% of our cohort (data not shown). In comparison, approximately 13% of adults in the US general population have diabetes. Based on the prevalence of diabetes and its effect on PC risk, we would predict an SIR = [(1 − 0.28) + 2.1 × 0.28]/[(1 − 0.13) + 2.1 × 0.13] = 1.14, in the absence of other factors (Steenland and Greenland 2004), which is smaller than what we observed. Thus, diabetes probably does not by itself explain the elevated SIR. Epidemiological studies also show a link between obesity and PC (Xu et al. 2018). In a cohort study conducted by the US National Institutes of Health, overweight or obese patients with a BMI ≥ 30 kg/m 2 were more likely to develop PC than those with a normal BMI (hazard ratios 1.15-1.53 for BMI measured at different ages) (Stolzenberg-Solomon et al. 2013). Indeed, the higher PC incidence in diabetic patients may be related to an increased prevalence of obesity, or vice versa. In our study, however, obesity was not associated with elevated PC incidence in a multivariable model (data not shown).
A recent study of patients with chronic hepatobiliary disease outside the SOT setting suggested the usefulness of early and incidental PC detection during surveillance imaging for hepatic disease (Kumagi et al. 2019). However, our analysis showed that survival after a PC diagnosis was very poor, regardless of stage at diagnosis, with a median survival of 4 months for the overall group. Survival was similarly poor for liver and other organ recipients. This seems no better than that of PC patients in the general population (Force et al. 2019). Thus, our study failed to support the possibility that earlier PC detection in liver recipients would lead to improved survival.
A strength of this study is the evaluation of a large nationwide cohort representative of the SOT recipient population in the US. This allowed us to investigate PC in detail and to assess risk factors using multivariable analysis. One limitation is that the cohort lacked information on smoking, which is a risk factor for PC in the general population (Park et al. 2021). However, the prevalence of cigarette smoking among SOT recipients seems unlikely to explain the elevated SIR of PC by itself, as kidney recipients, who comprise the majority of SOT recipients, have a smoking prevalence that is comparable to the US population (Kasiske and Klinger 2000). Also, only 15% of liver transplant recipients are active smokers (Anis et al. 2019). Heart and/or lung transplantation is usually not considered for active smokers, although up to a third of thoracic organ recipients resume smoking after transplantation (Zmeskal et al. 2016;Mehra et al. 2005). Another important limitation is the lack of information on changes over time in maintenance immunosuppressive medications and medication dosages, which prevented us from fully assessing the impact of immunosuppression on PC incidence post-SOT.
To conclude, our study demonstrates that PC incidence in SOT recipients is modestly higher than in the general population. Novel findings include the increased incidence among liver recipients and elevated risks for PC cases with local and regional stages and for those located in the head of the pancreas, suggesting that some cases may be diagnosed earlier due to incidental detection, related to frequent laboratory and radiological monitoring. However, survival after the diagnosis of PC in SOT recipients is poor even in patients diagnosed at a local or regional stage, and liver transplant recipients did not have improved survival. Consequently, our findings fail to support a potential benefit for systematic screening for PC following SOT.