The TCM Study (www.transplantmatch.cancer.gov) links the US Scientific Registry of Transplant Recipients (SRTR) with 17 state and regional cancer registries (see Table 1 footnote).(6) The SRTR includes data on donors, candidates, and recipients in the US, submitted by the members of the Organ Procurement and Transplantation Network (OPTN), and is described elsewhere.(15) The Health Resources and Services Administration, US Department of Health and Human Services, provides oversight of the OPTN and SRTR. The TCM Study covers approximately half of US transplants during 1987–2012. We evaluated first-time lung recipients who resided in the area covered by a participating registry at waiting list entry or transplant for inclusion (n=11,571); after exclusions, the final cohort was comprised of 8,993 first-time lung recipients (see Figure 1).

Characteristics of recipients and donors, transplant characteristics, and baseline immunosuppression and induction therapy were obtained from SRTR records. Recipient cigarette use was assessed using two SRTR variables: history of cigarette use or history of smoking >10 pack-years. Donor cigarette use was only reported in the dataset if >20 pack-years of lifetime use. Lung disease group was based on recipients’ primary listing diagnosis and categorized into 8 groups (see Table 1). We also assessed grouped diagnoses based on the 4 Lung Allocation Score (LAS) groups. We chose the 8 clinical disease groups for our analyses as we felt these better reflected clinicopathological as LAS grouping is partially based on waitlist mortality.(16) Pre-transplant spirometry testing values (forced expiratory volume in 1 second [FEV1], forced vital capacity [FVC]) were obtained from SRTR.

Incident lung cancers were ascertained using International Classification of Diseases for Oncology, 3^rd edition codes (ICD-O-3) from the linked cancer registry records. Histologic subtype was classified as non-small cell lung cancer (NSCLC), small cell lung cancer (SCLC) and other. NSCLC was further subdivided into adenocarcinoma, squamous cell carcinoma, and other NSCLC for some analyses. Clinical stage at cancer diagnosis was based on summary stage (localized, regional or distant). Cancer treatment (surgery, radiation, chemotherapy) and cause of death were ascertained from cancer registries, as described below.

Follow-up of lung recipients for incident lung cancer began at 3 months after transplant (to avoid pre-existing cancers) and ended at the earliest of lung cancer diagnosis, death, graft failure/retransplant, loss to follow-up, or end of registry coverage (December 2012). We compared incidence among lung recipients to the general population using standardized incidence ratios (SIR, defined as observed cases/expected cases). Expected counts were obtained by applying general population lung cancer rates to person-time at risk, stratified by age, sex, race/ethnicity, calendar year, and registry area. We calculated overall SIRs and separate SIRs for cancers in the native and transplanted lungs by generating observed and expected counts separately for left and right lungs, then summing across recipients’ lungs based on the transplant procedure (left SLT, right SLT, bilateral transplant [BLT]). SIRs use the recipients’ demographic characteristics to calculate the expected number of cancers, but when the SIRs were calculated separately for transplanted lungs, we alternatively standardized by donor characteristics.

Poisson regression was utilized to calculate incidence rate ratios (IRRs) to examine risk factors for native lung cancers. Each SLT recipients’ relevant lung was used for this analysis. Risk factors assessed in univariate models for association with native lung cancer included recipient sex, age at transplant, education level, lung disease group, LAS diagnosis group, history of smoking, time since transplant, listing FEV1, listing FVC, baseline antimetabolite therapy, baseline calcineurin inhibitor, induction regimen and transplanted lung (left vs. right). LAS diagnosis group, baseline antimetabolite/calcineurin inhibitor and induction regimen were not included in our final multivariable model because of lack of significant association in univariate models. FEV1 and FVC demonstrated some differences in risk between categories of percent predicted values, but no significant trend. Thus, they were also not included. IRRs for lung disease group were based on effect coding, which compares incidence in each group to overall average incidence. In order to retain subjects with missing data in the models, “missing” was included as a category for each of these variables of interest. We also used Poisson regression to examine risk factors for lung cancers arising in donor lungs. Given a limited number of cases, we only assessed a few risk factors in univariate models including: donor and recipient age, single vs. bilateral lung transplant, donor and recipient smoking, and time since transplant.

We then compared lung cancer cases in lung recipients to cases in the general population, restricted to cancer registry areas that provided information on vital status (see Table 5 footnote). Hispanic ethnicity was inconsistently reported by cancer registries, so we excluded the few cases for which Hispanic ethnicity was indicated. We used chi-squared testing to compare characteristics of cases with or without a prior lung transplant. The probability of all-cause mortality following cancer diagnosis was calculated for transplant and non-transplant cases using Kaplan-Meier curves with log-rank testing. We used Cox proportional hazards regression to compare all-cause mortality and lung cancer-specific mortality following cancer diagnosis, adjusting for age at diagnosis, sex, race/ethnicity, year of diagnosis, stage, and treatment modality. Some multivariable models were restricted to registries that provided information on cause of death and cancer treatment (see Table 5 and 6 footnotes). Analyses are reported overall and for NSCLC cases, as SCLC is staged and treated differently.(17) A final survival model was created to compare transplant and non-transplant NSCLC cases with local or regional stage disease undergoing potentially curative therapy (either surgery or radiation).(18–20) To account for any differences in the relationship between lung transplant and survival over time, we developed survival models separately for 3 eras according to year of cancer diagnosis (1992–2002, 2003–2006, 2007–2012). We then tested an interaction term between year of cancer diagnosis and an indicator variable for having a lung transplant. Finally, among our cohort of lung transplant recipients, we compared survival between those who developed lung cancer and those who did not, treating the occurrence of lung cancer as a time-dependent risk factor.

The TCM Study was approved by human subjects committees of the National Cancer Institute and participating cancer registries. All statistical analyses were two-sided and performed using SAS (versions 9.3 and 9.4, SAS Institute, Cary, NC). A p-value of < 0.05 was considered statistically significant.

The cohort of 8,993 first-time lung recipients were near evenly split between males and females, and the median age at transplant was 54 years (Table 1). The leading indications for transplant were COPD with or without alpha-1 antitrypsin deficiency (COPD/A1AT, 34%), IPF (25%) and cystic fibrosis (CF, 16%). Of recipients with available smoking data (77%), 64% had pre-transplant cigarette use. Forty-five percent received SLT and 55% received BLT. Donors were more likely to be male, more likely to be non-smokers, and were much younger than recipients (Table 2).

There were 183 incident lung cancers during follow-up (528 per 100,000 person-years). The median time from transplant to cancer was 3.9 years (interquartile range [IQR] 1.8–6.2) and was similar between those who developed lung cancer in a native (4.0 years, IQR 1.8–6.2) or a donor lung (3.5 years, IQR 0.93–6.9). We excluded all cases occurring in the first 3 months, and only 4 of 183 cases occurred in the subsequent 3 months. Most lung cancers occurred in SLT recipients (n=169, 92%) and, within this group, in the native lung (n=153, 91%). Overall, lung cancer risk was elevated almost 5-fold in lung recipients compared to the general population (SIR 4.8, 95% confidence interval [CI] 4.1–5.5) (Table 3). Risk of lung cancer in the native lung in SLT recipients was markedly high (SIR 13, 95%CI 11–15). Risk was not elevated in transplanted lungs when standardized by recipient characteristics (SIR 1.0, 95%CI 0.69–1.5). However, given the younger age of donors, when standardized by donor characteristics, the risk was strongly increased (SIR 5.0, 95%CI 3.3–7.3).

Thirty-seven percent (n=67) of lung cancer diagnoses were squamous cell carcinoma, 32% (n=58) adenocarcinoma, and 22% (n=40) other NSCLC. Eight percent (n=15) were SCLC and 2% (n=3) were other histologic subtypes. The SIR for squamous cell cancer was elevated to 8.1 (95%CI 6.3–10.3) while the SIR for adenocarcinoma was 4.4 (95%CI 3.3–5.7). In contrast, the SIR for SCLC was lower at 2.6 (95%CI 1.5–4.2).

We examined risk factors for native lung cancer (n=153) in SLT recipients (n=4,072) (Table 4). Older age at transplant, lower education level, history of cigarette smoking prior to transplant, and longer time since transplant were significantly associated with increased risk in multivariable models. Ninety percent of cases of native lung cancers occurred in recipients with either COPD/A1AT or IPF. However, while an IPF diagnosis was associated with significantly higher risk of native lung cancer (adjusted IRR 1.6, 95%CI 1.0–2.6), risk in recipients with COPD/A1AT was similar to the average risk across disease groups. Lung cancer incidence was also higher in a native right lung compared to a native left lung. In unadjusted models to explore risk factors for lung cancer arising in the donor lung (n=26), donor and recipient age, recipient smoking, and having a single lung transplant were associated with incidence of donor lung cancer (Table S1).

Lung cancer cases in lung recipients in the restricted cohort (n=121) were compared to 1,168,149 other lung cancer patients from the general population. Lung cancer patients with a lung transplant were younger (p<0.001) compared to other lung cancer patients (Table 5). Based on summary stage, lung cancers in lung recipients were more likely than other cases to be localized, both overall and for the subset of NSCLC (28% vs. 20%, p=0.04). Lung cancer patients with a lung transplant were more likely to receive surgery compared to other lung cancer patients (37% vs. 26%, p=0.05). Among NSCLC patients with localized/regional disease, lung recipients were also more likely to receive surgery (93% vs. 69%, p=0.01) and less likely to receive radiation (11% vs. 43%, p=0.001). Despite being more likely to present in localized stage, lung transplant patients had shorter survival after cancer diagnosis than other lung cancer patients (median 153 vs. 245 days, p<0.001; Figure 2A). Survival was also worse in lung transplant recipients when analysis was stratified to NSCLC (p<0.001, Figure 2B) or SCLC (p=0.02, Figure 2C).

In Cox models adjusted for stage and cancer treatment, survival outcomes were worse, for all cancer cases and specifically for those with NSCLC (all-cause mortality: adjusted HR 1.84, 95%CI 1.46–2.33; cancer-specific mortality: adjusted HR 1.62, 95%CI 1.22–2.15) (Table 6). In analyses restricted to patients with NSCLC diagnosed at localized/regional stage who received potentially curative therapy (surgery and/or radiation), survival was also worse among lung recipients (p<0.001, Figure 1D; all-cause mortality: adjusted HR 2.18, 95%CI 1.54–3.10; cancer-specific mortality: adjusted HR 1.83, 95%CI 1.15–2.90). The association between transplantation and increased mortality appeared similar across the three time eras (Table S2), and interaction terms between year of lung cancer diagnosis and having a lung transplant were not significant for all-cause or lung cancer specific mortality.

Finally, among the cohort of lung transplant recipients, we compared survival between those who developed lung cancer in follow-up and those who did not, adjusted for age at transplant, sex, race/ethnicity and year of transplant. The risk of death was substantially higher among subjects who developed lung cancer compared to those who did not (unadjusted HR 7.25, 95% CI 6.17–8.53; adjusted HR 6.51, 95% CI 5.53–7.67).