Study participants and data collection

BMTSS was established to examine the long-term outcomes of individuals who survived ≥2 years after undergoing BMT between 1974 and 2014 at the University of Alabama at Birmingham (UAB), City of Hope (COH) or University of Minnesota (UMN). BMTSS also examines outcomes in an unaffected comparison group drawn from the survivors’ siblings. Study participation consists of completion of the BMTSS survey by the BMT survivors and siblings. The survey captures sociodemographic characteristics (sex, race/ethnicity, education, annual household income, and health insurance), chronic health conditions (as diagnosed by their healthcare provider), history of chronic graft vs. host disease (cGvHD), relapse and subsequent neoplasms after BMT, health risk behaviors (smoking, alcohol consumption, and lack of exercise) and BMT-related anxiety (Supplementary Table 1). Chronic health conditions have been graded using the Common Terminology Criteria for Adverse Events, v5.0 from grade 1 (mild) to grade 5 (death due chronic health condition) [5]. Survivors’ age at BMT, primary diagnosis, type of BMT (autologous, allogeneic), risk of relapse at BMT (standard risk: first or second complete remission after acute lymphoblastic [ALL] or acute myeloid leukemia [AML], Hodgkin lymphoma [HL] or non-Hodgkin lymphoma [NHL], first chronic phase of chronic myeloid leukemia [CML], or severe aplastic anemia; high risk: all other patients) [6], stem cell source (bone marrow, peripheral blood stem cells [PBSCs], or cord blood), use of total body irradiation (TBI) for conditioning, conditioning intensity (myeloablative conditioning [MAC] or non-myeloablative/reduced-intensity conditioning [collectively termed NMA]), and pre-BMT therapeutic exposures were retrieved from institutional transplant databases and medical records. BMT survivors were placed into four groups based on TBI exposure and conditioning intensity: NMA/no TBI, NMA/TBI, MAC/no TBI, and MAC/TBI. The Institutional Review Board (IRB) at UAB serves as the single IRB of record; IRBs at UMN and COH approved the BMTSS protocol. Participants provided informed consent according to the Declaration of Helsinki. The present report includes survivors who received BMT at any age and were 18 years of age or older when they completed the BMTSS survey. Siblings were also 18 years or older at study participation.

Frailty phenotype

Frailty phenotype was constructed from responses provided by BMT survivors for the following five indices (Supplementary Table 2): clinically underweight, exhaustion, low energy expenditure, slowness, and muscle weakness. Participants were categorized as frail (≥3 indices), pre-frail (2 indices), or non-frail (0–1 indices) [1]. For this analysis, we excluded survivors and siblings who met the criteria for frailty and retained only those who were pre-frail or non-frail.

Late mortality

The primary outcomes of interest included all-cause and cause-specific late mortality (non-recurrence-related [NRM] and recurrence-related mortality [RRM]). National Death Index (NDI) Plus provided data regarding the date and cause of death through December 31, 2020 [7]. Additional data from the Accurint database [8] extended the vital status information through December 21, 2021. Suicides, homicides, and accidents were classified as external causes of death. A cause of death matching the pre-transplant diagnosis was classified as RRM. All other causes of death were classified as NRM (subsequent malignant neoplasms [SMNs], cardiovascular disease [CVD], infections, pulmonary, etc.).

Statistical analysisComparison with same-sex biological siblings

We paired closest-age same-sex biological siblings with BMT survivors. To estimate adjusted odds ratios (aOR) and 95% confidence intervals (95% CI) of pre-frailty in BMT survivors compared with their siblings we used logistic regression with generalized estimating equation [9] to account for the paired data. The following variables were evaluated for inclusion in the model: age at survey grades 3–4 chronic health conditions, socioeconomic status (SES: <college and <$50,000; <college and ≥$50,000; ≥college and <$50,000; ≥college and ≥$50,000), health insurance, alcohol consumption, lack of exercise and smoking.

Factors associated with pre-frailty among the BMT survivors

Risk factors examined for association with pre-frailty included time from BMT to survey, age at survey, sex, race/ethnicity, SES, health insurance, primary diagnosis, pre-BMT radiation, risk of relapse at BMT, BMT era (1974–1989; 1990–2004; 2005–2014) [10], BMT institution (UAB; COH; UMN), stem cell source, conditioning intensity/TBI, BMT type/cGvHD (autologous BMT; allogeneic BMT/no cGvHD; allogeneic BMT/cGvHD), grades 3–4 chronic health conditions, post-BMT relapse, BMT-related anxiety, smoking, alcohol consumption, and lack of physical activity. These demographic and clinical variables were examined for their possible inclusion in the multivariable analysis based on their significance in the unadjusted model and previous knowledge (Supplementary Table 3); associations between the risk factors and pre-frailty were reported as unadjusted OR with corresponding 95% CI, using non-frail as the reference group.

Pre-frailty status and subsequent mortality among BMT survivors

Kaplan–Meier methods were used to calculate overall survival. Cox proportional hazards models with time from survey as the time axis was used for identifying the association between pre-frailty and all-cause mortality. Demographic and clinical variables listed above were examined for their possible inclusion in the models (Supplementary Table 4). We examined statistical interaction between pre-frailty and key risk factors (primary disease, chronic health conditions, stem cell source, age at BMT [<45 years; ≥45 years], age at survey [<65 years; ≥65 years], and BMT type), to assess if the association between pre-frailty and subsequent mortality was modified by these variables.

Cumulative incidence of cause-specific mortality was calculated using competing risk methods; deaths attributed to RRM and external causes served as competing risks for NRM, and deaths attributed to NRM and external causes served as competing risks for RRM. Proportional sub-distribution hazard (Fine-Gray) models [11] were used to examine the association between pre-frailty and cause-specific mortality. Participants with missing or unknown cause of death were excluded from the proportional sub-distribution hazard models for RRM and NRM. Adjustment of the models with demographic and clinical factors was similar to that described in the Cox regression models above. Results were presented as adjusted hazard ratio (aHR) with 95% CI.

All analyses were performed using SAS v9.4 (SAS Institute, Cary, North Carolina). Findings with 2-sided tests were considered statistically significant at P < 0.05.