Exercise Duration During a Brief Step Test Independently Predicts Outcomes After Hematopoietic Cell Transplantation

Abstract

Background.

Pre-transplant pulmonary function testing (PFT) is standard before hematopoietic cell transplantation (HCT), but few centers complement it with objective functional capacity assessment. At our institution, a 3-minute step exercise test is performed at the time of PFTs, generating two distinct measures: exercise duration (a functional capacity surrogate) and oxygen desaturation (a gas exchange measure). Whether either component adds prognostic value beyond PFTs is unknown.

Objectives.

The primary objective of the study was whether step-derived variables added predictive value beyond a baseline model containing age, sex, transplant type, FEV₁ z-score, DLCO z-score, and hemoglobin. Outcomes of interest were overall survival, ICU admission within 120 days, and invasive mechanical ventilation within 120 days.

Study Design.

This is a single-center retrospective cohort of 3,458 adult HCT recipients (2,648 autologous, 810 allogeneic) who underwent a 3-minute step exercise test before HCT between 2005 and 2024. Pulse oximetry and heart rate were recorded during the step test. Pre-transplant PFTs were interpreted using Global Lung Initiative reference equations and reported as z-scores. Overall survival was analyzed using Cox proportional hazards regression and ICU admission and invasive mechanical ventilation were analyzed using logistic regression. Additive value of exercise variables was evaluated with a pre-specified baseline model containing established predictors of HCT outcomes.

Results.

Median follow-up was 4.5 years. During follow-up, 1,149 patients (33.2%) died, 495 (14.3%) were admitted to the ICU within 120 days, and 131 (3.8%) required invasive mechanical ventilation. Of 3,431 patients with exercise time recorded, 718 (20.9%) stopped before completing the 3-minute protocol. Exercise time predicted all three outcomes after full adjustment: each additional minute was associated with lower mortality (HR 0.90, 95% CI 0.82–0.99, p = 0.025), lower odds of ICU admission (OR 0.80, 95% CI 0.69–0.92, p = 0.003), and lower odds of invasive mechanical ventilation (OR 0.70, 95% CI 0.55–0.89, p = 0.004). Model fit improved with the addition of exercise time for all three outcomes. In contrast, oxygen desaturation did not predict any outcome after hemoglobin adjustment (overall survival HR 0.99, 95% CI 0.96–1.02, p = 0.53). An apparent association between desaturation and mortality without hemoglobin adjustment (HR 0.96, p = 0.003) was fully absorbed by measured hemoglobin, confirming confounding by oxygen-carrying capacity.

Conclusions.

Exercise duration during a brief pre-transplant step test independently predicts overall survival, ICU admission, and invasive mechanical ventilation after HCT, capturing a dimension of pre-transplant fitness not measured by resting PFTs or hemoglobin. Oxygen desaturation does not add prognostic value, and the apparent signal in unadjusted analyses reflects hemoglobin confounding. These findings support emphasizing exercise capacity assessment before HCT.

Introduction

Pre-transplant risk assessment for hematopoietic cell transplantation (HCT) relies on a small set of established tools, each with recognized limitations. The Hematopoietic Cell Transplantation-Specific Comorbidity Index (HCT-CI) catalogues organ-specific comorbidities.¹ The Karnofsky Performance Status (KPS) is physician-assessed and has well-documented interobserver variability. Pulmonary function tests (PFTs) measure resting pulmonary mechanics, with FEV₁ and DLCO each retaining independent prognostic value across the full HCT population.² None of these instruments directly measures the integrated cardiopulmonary-musculoskeletal capacity that determines whether a patient can withstand the physiologic stress of conditioning, prolonged neutropenia, and recovery.³ As HCT is offered to progressively older patients with greater comorbidity burden, this gap has become harder to ignore.^4,5

In solid organ transplantation, the six-minute walk distance (6MWD) is a validated, integrative measure of functional capacity that predicts mortality independently of disease severity scores and is incorporated into lung allocation algorithms and liver transplant risk assessment. ^6,7 In HCT, however, formal functional testing is not part of standard pre-transplant evaluation at most centers. A prior study examining 6MWD before allogeneic HCT found that it did not retain prognostic value after adjustment for KPS, though that analysis was limited to 243 patients with a single outcome.⁸ Despite growing consensus that frailty assessment is useful in transplant candidate evaluation, objective functional measurement before HCT remains the exception rather than the rule.^9,10

At a smaller number of centers, including ours, a brief step exercise test has been added to the pre-transplant PFT visit. The test emulates the Master’s two-step cardiac screening test and was adopted at our institution to screen for exercise-induced desaturation that may not be apparent at rest¹¹. It requires no additional clinic visit, corridor or specialized equipment beyond a standard step and pulse oximeter already present in the PFT laboratory. The protocol generates two distinct types of information. The first is a gas exchange measure: the change in oxygen saturation (SpO₂) from rest to nadir during exercise, with the rationale that exercise stress would unmask abnormal gas exchange not apparent at rest. The second is a functional capacity measure embedded inadvertently in the protocol: whether the patient completes the 3-minute test, and if not, how long they sustained exercise. Whether either component adds prognostic information beyond PFTs and hemoglobin has not been evaluated. We conducted a large retrospective cohort study to test whether exercise capacity and oxygen desaturation add prognostic value beyond age, sex, transplant type, FEV₁ z-score, DLCO z-score, and hemoglobin in HCT candidates. We hypothesized that exercise capacity, as a functional capacity surrogate, would carry independent prognostic information.

Methods

Study Design and Population

We conducted a retrospective cohort study of adult patients (≥18 years) who underwent HCT at Mayo Clinic, Rochester, Minnesota between January 2005 and December 2024 and underwent pre-transplant step exercise testing as part of their pulmonary function evaluation. The study was approved by the Mayo Clinic Institutional Review Board, and reporting follows the STROBE guidelines.¹² Step exercise testing has been performed at our institution as part of the pre-transplant PFT visit since 2005, with implementation expanding over the study period. The test is omitted when patients decline, when there is perceived falls risk, when musculoskeletal limitation (e.g. knee pain) prevents step exercise, when irregular heart rate is noticed during PFT or when scheduling or logistical constraints intervene (including inadvertent technician omission of the step exercise protocol). Patients who underwent multiple transplants during the study period contributed one observation per transplant episode.

Step Exercise Protocol

The protocol consists of three minutes of stepping on a 9-inch steps (5-inch steps available for patients with shorter height or orthopedic limitations) at a self-selected pace, performed immediately after standard PFTs. Pulse oximetry is recorded continuously, with resting and exercise SpO₂ captured. Heart rate is recorded at rest and during exercise. The test is continuous without rest breaks; if a patient cannot continue, the test is terminated and elapsed time recorded. The test is terminated early if heart rate reaches age-dependent thresholds (125–160 bpm), if oxygen saturation falls below 85%, or at the patient’s request. Oxygen desaturation was defined as the absolute fall in SpO₂ from resting to nadir exercise values. Clinically significant desaturation was defined a priori as a fall of ≥4 percentage points, adapted from the ATS six-minute walk test guidelines.¹³ Heart rate change was calculated as the difference between exercise and resting heart rate. Exercise duration was recorded for patients who did not complete the full protocol. No adverse events attributable to the step test, including falls, syncope, or arrhythmias, were reported during the study period.

Pulmonary Function Testing

Pre-transplant spirometry and DLCO were performed according to ATS/ERS technical standards.^14,15 Reference values for spirometry were calculated using the Global Lung Initiative (GLI) Global 2022 equations and DLCO reference values used the GLI 2017 equations.^16,17 FEV₁ and DLCO are reported as z-scores throughout.¹⁸

Covariates and Outcomes

The analytic framework was designed to test whether each step exercise variable added prognostic information beyond a pre-specified baseline model containing age (per year), sex, transplant type (allogeneic vs. autologous), FEV₁ z-score, DLCO z-score, and hemoglobin (g/dL). These covariates were selected a priori as established predictors of HCT outcomes.

Hemoglobin was included as a covariate since it reflects marrow reserve, the recency and intensity of cytotoxic chemotherapy, and independently predicts post-transplant mortality.¹⁹ It also directly influences SpO₂ response to exercise: with reduced oxygen-carrying capacity, the alveolar-to-tissue oxygen gradient must steepen to meet a given metabolic demand, lowering exercise SpO₂ even in the absence of pulmonary disease. Hemoglobin values were obtained from the laboratory value closest to the date of exercise testing within a 30-day window. Hemoglobin was not available for all patients, predominantly those transplanted before an institutional electronic health record (EHR) migration in 2018 that resulted in incomplete transfer of laboratory data. The primary analysis was restricted to patients with measured hemoglobin, with sensitivity analyses in the full cohort as described below. Transplant type was categorized as autologous or allogeneic. Age at transplant and sex were abstracted from the electronic medical record. Pulmonary function tests were performed for over 95% of HCT recipients at our institution over the study period, with exceptions typically being high-risk patients who remained inpatient for the duration of their pre-transplant evaluation.

The primary outcome was overall survival from the date of transplant. Secondary outcomes were ICU admission within 120 days and invasive mechanical ventilation within 120 days. The 120-day window was chosen to capture early peri-transplant morbidity while the patient remained close to the transplant center before being released home. Later events may have been more likely to occur locally and not be fully captured. Non-relapse mortality was not readily available in our data sources.

Statistical Analysis

Overall survival was analyzed using Cox proportional hazards regression. ICU admission and invasive mechanical ventilation were analyzed using logistic regression. Additive value was assessed by adding each exercise variable individually to the baseline model and evaluating the likelihood ratio test and the change in Akaike Information Criterion (ΔAIC), along with the magnitude and direction of the effect estimate. Two categories of step-derived variables were examined: functional capacity (exercise time as continuous minutes and as a binary completion variable) and gas exchange (desaturation as continuous percentage points, desaturation ≥4% as binary, and heart rate change per beat per minute).

To characterize the role of hemoglobin in the desaturation-outcome relationship, three analytic approaches were used. The primary analysis included those patients with hemoglobin available within 30 days of testing. A secondary complete-case analysis included all patients with exercise testing but omitted hemoglobin from the baseline model. A third analysis used multiple imputation by chained equations (MICE) with predictive mean matching across 10 imputed datasets to impute hemoglobin in the full cohort, including the outcome and event indicator in the imputation model.^20,21 The hemoglobin-missing population related to missing data for some patients due to an EHR merger in 2018. While the missingness pattern reflects EHR migration rather than clinical selection, it is not necessarily missing-at-random, motivating the primary analysis restriction with sensitivity analyses to confirm robustness. A pre-specified subgroup analysis was performed in allogeneic recipients only, where event rates and clinical stakes are highest.

The proportional hazards assumption was assessed using Schoenfeld residuals. Where violations were detected, stratified Cox models with transplant type as the stratifying variable were fitted to confirm that conclusions were unchanged. Linearity of continuous predictors was assessed using restricted cubic splines (3 degrees of freedom) compared to the linear term by likelihood ratio test. Continuous variables are presented as mean ± standard deviation or median (interquartile range) as appropriate. Categorical variables are presented as frequencies and percentages. All analyses were performed using R version 4.3 (R Foundation for Statistical Computing, Vienna, Austria); key packages were survival, mice, and rspiro. All tests were two-sided with p < 0.05 considered significant.

Results

Study Population

During the study period, 7,514 adult HCT were performed at our institution, of which 5,364 (71.4%) had step exercise testing. The step exercise protocol was technician-initiated rather than physician-ordered, and its inclusion in the pre-transplant PFT visit was inconsistent, particularly in the earlier study years. Among authorized patients without pre-transplant step testing in our database, overall mortality was 54.9% compared to 43.8% in the tested cohort. The groups were similar in age (mean 57.7 vs. 57.1 years) and proportion of allogeneic transplants (20.7% vs. 21.3%) but differed by era (median transplant year 2013 vs. 2017). After excluding 872 episodes in which testing was performed only during post-transplant PFTs and 13 with missing DLCO data, 4,479 pre-transplant episodes among 4,472 unique patients remained (3,496 autologous, 983 allogeneic). Of these, 3,458 had hemoglobin available within 30 days of exercise testing and constituted the primary analytic cohort (2,648 autolo34gous, 810 allogeneic). Of 3,458 patients in the primary cohort, 3,431 (99.2%) had exercise duration recorded. The remaining 27 had pulse oximetry data (resting and exercise SpO2) but exercise duration was not documented, presumed secondary to documentation error. These patients were included in desaturation analyses but excluded from exercise time analyses. The 1,021 patients without hemoglobin data were all transplanted before the 2018 EHR migration (median transplant year 2008 vs. 2019 in the primary cohort and had incomplete laboratory data transfer into the new EHR, Figure 1).

Figure 1.

Study flow diagram. HCT, hematopoietic cell transplantation; PFT, pulmonary function test; DLCO, diffusing capacity of the lung for carbon monoxide; EHR, electronic health record; ICU, intensive care unit.

Cohort Characteristics

Mean age in the primary cohort was 57.8 ± 12.7 years and 37.8% were female (Table 1). Pre-transplant pulmonary function was generally well-preserved, with a mean FEV₁ z-score of −0.2 ± 1.1 and a mean DLCO z-score of −1.2 ± 1.2. Mean hemoglobin was 11.5 ± 2.1 g/dL, substantially lower in allogeneic than autologous recipients (10.1 ± 2.1 vs. 12.0 ± 1.8 g/dL). During follow-up, 1,149 patients (33.2%) died, 495 (14.3%) were admitted to the ICU within 120 days, and 131 (3.8%) required invasive mechanical ventilation within 120 days; all three outcomes were more frequent among allogeneic recipients.

Table 1.

Baseline characteristics of the primary cohort (n = 3,458) stratified by transplant type.

Variable	All (n = 3,458)	Autologous (n = 2,648)	Allogeneic (n = 810)
Demographics
Age, years	57.8 ± 12.7	58.8 ± 11.8	54.3 ± 14.7
Female sex, n (%)	1,307 (37.8)	981 (37.0)	326 (40.2)
Underlying diagnosis, n (%)
Plasma cell disorder	1,726 (49.9)	1,724 (65.1)	2 (0.2)
Non-Hodgkin lymphoma	684 (19.8)	668 (25.2)	16 (2.0)
AML	401 (11.6)	1 (< 0.1)	400 (49.4)
MDS/MPN	237 (6.9)	2 (0.1)	235 (29.0)
Hodgkin lymphoma	162 (4.7)	161 (6.1)	1 (0.1)
Other	248 (7.2)	92 (3.5)	156 (19.3)
Donor type (allogeneic), n (%)
Unrelated	449 (55.4)	—	449 (55.4)
Related	285 (35.2)	—	285 (35.2)
Haploidentical	48 (5.9)	—	48 (5.9)
Cord blood	28 (3.5)	—	28 (3.5)
Pulmonary function
FEV1 z-score	−0.2 ± 1.1	−0.3 ± 1.1	−0.1 ± 1.1
FEV1, % predicted	95.9 ± 17.9	95.1 ± 18.1	98.6 ± 16.8
DLCO z-score	−1.2 ± 1.2	−1.2 ± 1.1	−1.2 ± 1.2
DLCO, % predicted	83.2 ± 16.1	83.3 ± 16.1	83.0 ± 16.2
Step exercise test
Hemoglobin, g/dL	11.5 ± 2.1	12.0 ± 1.8	10.1 ± 2.1
Resting SpO2, %	97.8 ± 1.5	97.7 ± 1.5	98.3 ± 1.5
Exercise SpO2, %	96.4 ± 2.2	96.2 ± 2.2	97.1 ± 2.1
Desaturation, percentage points	1.4 ± 1.8	1.5 ± 1.9	1.2 ± 1.7
Desaturation ≥ 4%, n (%)	328 (9.5)	272 (10.3)	56 (6.9)
Heart rate change, bpm	32.1 ± 12.8	31.2 ± 12.4	35.2 ± 13.7
Exercise time recorded, n (%)	3,431 (99.2)	2,632 (99.4)	799 (98.6)
Early stop (< 3 min), n (%)†	718 (20.9)	491 (18.7)	227 (28.4)
Transplant and outcomes
Death, n (%)	1,149 (33.2)	820 (31.0)	329 (40.6)
ICU admission 0–120 d, n (%)	495 (14.3)	300 (11.3)	195 (24.1)
Invasive ventilation 0–120 d, n (%)	131 (3.8)	51 (1.9)	80 (9.9)

Continuous variables are presented as mean ± standard deviation or median (interquartile range); categorical variables as n (%).

^†Among those with exercise time recorded.

Abbreviations: FEV₁, forced expiratory volume in 1 second; DLCO, diffusing capacity of the lung for carbon monoxide; SpO2, oxygen saturation by pulse oximetry; AML, acute myeloid leukemia; MDS, myelodysplastic syndrome; MPN, myeloproliferative neoplasm; ICU, intensive care unit; IQR, interquartile range.

Step Exercise Findings

Of 3,431 patients with exercise time recorded, 718 (20.9%) stopped before completing the 3-minute protocol, with early stopping more common in allogeneic than autologous recipients (28.4% vs. 18.7%) (Figure 2A). The protocol elicited modest gas exchange stress: mean resting SpO2 was 97.8 ± 1.5%, mean exercise SpO2 was 96.4 ± 2.2%, and mean desaturation was 1.4 ± 1.8 percentage points. Only 328 patients (9.5%) experienced clinically significant desaturation defined as ≥4 percentage points. Mean heart rate change was 32.1 ± 12.8 bpm

Figure 2.

Exercise capacity findings. (A) Distribution of exercise time among early stoppers (n = 718), stratified by transplant type. The 2,713 patients who completed the full 3-minute protocol are not shown. (B) Kaplan-Meier curves for overall survival by protocol completion status (log-rank p < 0.0001). (C) Adjusted dose-response relationship between exercise time and mortality, modeled with restricted cubic splines (3 knots). The dashed line indicates HR = 1.0. Nonlinearity test p = 0.14. (D) Forest plot of the adjusted effect of exercise time (per 1 minute) on overall survival (HR), ICU admission (OR), and invasive mechanical ventilation (OR). All models adjusted for age, sex, transplant type, FEV₁ z-score, DLCO z-score, and hemoglobin.

Exercise Capacity Independently Predicted Outcomes

Exercise time independently predicted all three outcomes after full covariate adjustment (Table 2, Figure 2B-D). Each additional minute of step exercise completed was associated with a 10% lower hazard of death (HR 0.899, 95% CI 0.820–0.985, p = 0.025), 21% lower odds of ICU admission (OR 0.795, 95% CI 0.686–0.921, p = 0.003), and 30% lower odds of invasive mechanical ventilation (OR 0.698, 95% CI 0.550–0.885, p = 0.004). The dose-response relationship between exercise time and mortality was adequately modeled as linear (nonlinearity test p = 0.14). Model fit improved with the addition of exercise time across all three outcomes (ΔAIC −3.1 for overall survival, −7.0 for ICU admission, −6.3 for invasive mechanical ventilation). Kaplan-Meier analysis confirmed that patients who stopped early had significantly worse overall survival than those who completed the protocol (log-rank p < 0.0001; Figure 2B). When exercise time was dichotomized as early stop versus completion, the adjusted effect on mortality attenuated to borderline significance (HR 1.144, 95% CI 0.993–1.318, p = 0.064), indicating that the binary classification discards prognostically meaningful information across the zero to three-minute range.

Table 2.

Additive value of step exercise variables beyond the pre-specified baseline model (age, sex, transplant type, FEV₁ z-score, DLCO z-score, hemoglobin).

Outcome	Variable / Analysis	Effect (95% CI)	p	ΔAIC
Exercise time (functional capacity)
Overall survival	Per 1-minute increment	HR 0.90 (0.82–0.99)	0.025	−3.1
Overall survival	Early stop (< 3 min) vs. completed	HR 1.14 (0.99–1.32)	0.064	−1.4
ICU admission 0–120 d	Per 1-minute increment	OR 0.80 (0.69–0.92)	0.003	−7.0
Invasive ventilation 0–120 d	Per 1-minute increment	OR 0.70 (0.55–0.89)	0.004	−6.3
Oxygen desaturation (gas exchange) — primary model with hemoglobin
Overall survival	Per 1 percentage point	HR 0.99 (0.96–1.02)	0.531	+1.6
ICU admission 0–120 d	Per 1 percentage point	OR 1.01 (0.96–1.07)	0.673	+1.8
Invasive ventilation 0–120 d	Per 1 percentage point	OR 1.03 (0.93–1.14)	0.569	+1.7
Desaturation — sensitivity analysis, hemoglobin omitted (full cohort, n = 4,479)
Overall survival	Per 1 percentage point	HR 0.96 (0.94–0.99)	0.003	—
ICU admission 0–120 d	Per 1 percentage point	OR 0.98 (0.94–1.03)	0.386	—
Invasive ventilation 0–120 d	Per 1 percentage point	OR 0.98 (0.90–1.06)	0.622	—
Desaturation — sensitivity analysis, MICE-imputed hemoglobin (full cohort, n = 4,479)
Overall survival	Per 1 percentage point	HR 0.97 (0.95–1.00)	0.033	—
ICU admission 0–120 d	Per 1 percentage point	OR 0.99 (0.95–1.04)	0.786	—
Invasive ventilation 0–120 d	Per 1 percentage point	OR 1.00 (0.92–1.08)	0.922	—
Desaturation — pre-specified allogeneic subgroup (n = 810)
Overall survival	Per 1 percentage point	HR 0.99 (0.93–1.06)	0.819	—
ICU admission 0–120 d	Per 1 percentage point	OR 1.02 (0.93–1.13)	0.644	—
Invasive ventilation 0–120 d	Per 1 percentage point	OR 1.06 (0.91–1.23)	0.453	—
Heart rate response
Overall survival	Per 1 bpm	HR 1.00 (0.99–1.00)	0.131	−0.3

Baseline model: age (per year), sex, transplant type (autologous vs. allogeneic), FEV₁ z-score, DLCO z-score, and hemoglobin (g/dL). Hazard ratios (HR) from Cox proportional-hazards models for overall survival; odds ratios (OR) from logistic regression for ICU admission and invasive mechanical ventilation within 120 days. ΔAIC denotes the change in Akaike Information Criterion relative to the baseline model after adding the listed variable; negative values indicate improved fit, positive values indicate worse fit (a value of approximately −2 or lower supports meaningful additive value). Sensitivity analyses for desaturation (without hemoglobin, with MICE-imputed hemoglobin in the full cohort of n = 4,479, and in the allogeneic-only subgroup) are shown to demonstrate hemoglobin confounding of the desaturation signal.

Abbreviations: HR, hazard ratio; OR, odds ratio; CI, confidence interval; AIC, Akaike Information Criterion; MICE, multiple imputation by chained equations.

Patients who stopped early were similar in age to those who completed the protocol (58.5 vs. 57.6 years, p = 0.10) but were significantly more likely to be female (56.3% vs. 32.8%, p < 0.001), to undergo allogeneic transplantation (31.6% vs. 21.1%, p < 0.001), and to have diagnoses associated with allogeneic transplantation such as AML (18.4% vs. 9.7%) rather than plasma cell disorders (36.5% vs. 53.7%, p < 0.001 for overall diagnosis distribution). Early stoppers also had lower DLCO z-scores (−1.39 vs. −1.13, p<0.001), lower FEV₁ z-scores (−0.35 vs. −0.20, p =0.002), and lower hemoglobin (10.4 vs. 11.8 g/dL, p<0.001) than those who completed the protocol (Supplemental Table 1). Among allogeneic recipients, conditioning intensity (25.9% myeloablative vs. 30.0% reduced-intensity/nonmyeloablative, p = 0.23) and donor type distribution did not differ significantly between groups (p = 0.09).

The exercise time effect was consistent across transplant types. Among autologous recipients, each additional minute of exercise was associated with 14% lower mortality (HR 0.86, 95% CI 0.77–0.97, p = 0.013), 18% lower odds of ICU admission (OR 0.82, 95% CI 0.68–1.00, p = 0.053), and 44% lower odds of invasive mechanical ventilation (OR 0.56, 95% CI 0.39–0.82, p = 0.004). Among allogeneic recipients, point estimates were consistent but did not all reach significance, with the strongest effect for ICU admission (OR 0.76, 95% CI 0.61–0.95, p = 0.018). Formal interaction testing confirmed that the effect of exercise time did not differ significantly between transplant types (interaction p = 0.11, 0.84, and 0.12 for OS, ICU, and IMV, respectively). Exercise capacity therefore captures a dimension of pre-transplant fitness that is not explained by resting pulmonary function or oxygen-carrying capacity alone.

In a sensitivity analysis adjusting for transplant year in addition to the baseline covariates, exercise time effects were minimally attenuated and remained significant for all three outcomes (Supplemental Table 2). Transplant year was independently associated with improved OS (HR 0.93 per year, p < 0.001) and lower ICU admission (OR 0.96, p < 0.001), reflecting secular improvements in transplant outcomes over the study period.

Oxygen Desaturation Did Not Add Predictive Value

Oxygen desaturation added no predictive value beyond the baseline model for any outcome (Table 2, Figure 3A-B). Per 1 percentage point of desaturation, the hazard ratio for overall survival was 0.990 (95% CI 0.959–1.022, p = 0.53), the odds ratio for ICU admission was 1.012 (95% CI 0.959–1.067, p = 0.67), and the odds ratio for invasive mechanical ventilation was 1.030 (95% CI 0.932–1.137, p = 0.57). AIC increased with the addition of desaturation for all three outcomes, indicating that the variable adds noise rather than signal. Results were unchanged when desaturation was modeled as the binary clinical threshold of ≥4% (overall survival HR 0.85, 95% CI 0.70–1.05, p = 0.12). Heart rate change during exercise was also not associated with overall survival (HR 0.996 per bpm, 95% CI 0.992–1.001, p = 0.13). In the allogeneic subgroup, where event rates were highest, desaturation showed no association with any outcome (overall survival HR 0.99, p = 0.82; ICU OR 1.02, p = 0.64; invasive mechanical ventilation OR 1.06, p = 0.45).

Figure 3.

Oxygen desaturation. (A) Distribution of desaturation (absolute fall in SpO2 with exercise) in the primary cohort, stratified by transplant type. The dashed line indicates zero change. (B) Forest plot of the adjusted effect of desaturation (per 1 percentage point) on each outcome in the primary analysis (n = 3,458). (C) Sensitivity analyses for the desaturation effect on overall survival, ICU admission, and invasive mechanical ventilation: without hemoglobin adjustment (n = 4,479), with MICE-imputed hemoglobin (n = 4,479), and in the allogeneic subgroup (n = 810). The dashed line indicates HR or OR = 1.0. All primary models adjusted for age, sex, transplant type, FEV₁ z-score, DLCO z-score, and hemoglobin.

The sensitivity analyses revealed a confounding pattern that explains the null finding (Figure 3C). Pre-transplant hemoglobin reflects marrow reserve and the intensity of prior cytotoxic chemotherapy, and it was the strongest continuous predictor of mortality in the baseline model (HR 0.902 per g/dL).

Hemoglobin also directly determines the SpO2 response to exercise: reduced oxygen-carrying capacity requires a steeper alveolar-to-tissue oxygen gradient to meet metabolic demand, lowering exercise SpO₂ even without pulmonary disease. In the complete-case analysis without hemoglobin adjustment (n = 4,479), desaturation was associated with a 3.6% lower hazard of death per percentage point (HR 0.964, 95% CI 0.941–0.988, p = 0.003), an apparent protective effect that is biologically implausible. With MICE-imputed hemoglobin, the association attenuated but remained nominally significant (HR 0.973, 95% CI 0.949–0.998, p = 0.033), consistent with incomplete confounding control from imputation error. With measured hemoglobin in the primary analysis, the association was eliminated (HR 0.990, p = 0.53). This gradient confirms that hemoglobin simultaneously drives the tendency to desaturate during exercise and the risk of post-transplant mortality, and that desaturation itself carries no independent prognostic information once this confounder is accounted for.

Discussion

Exercise duration during a 3-minute step test predicts overall survival, ICU admission, and invasive mechanical ventilation after HCT, with consistent dose-response relationships and improvement in model fit across all three outcomes. These effects persist after adjustment for resting pulmonary function and hemoglobin, indicating that even a crude, time-capped measure of exercise capacity captures a dimension of pre-transplant fitness that static measurements miss. In contrast, oxygen desaturation during the same test carried no prognostic value in any analytic approach, and the apparent association with survival observed without hemoglobin adjustment was a confounding artifact. The two components of the step test therefore carry fundamentally different prognostic information.

The construct most likely captured by exercise time is functional capacity: the integrated cardiopulmonary, musculoskeletal, and metabolic reserve required to sustain physical effort. This is the same construct that 6MWD measures formally across solid organ transplantation and that geriatric assessments capture through gait speed, grip strength, and instrumental activities of daily living.^6,7,10,22,23 The novel observation is that even a 3-minute step test, performed without specialized equipment and with no additional clinic visit, captures enough of this construct to add prognostic information. Each minute of exercise completed was associated with 10% lower mortality, 21% lower odds of ICU admission, and 30% lower odds of mechanical ventilation, and these effects were not explained by resting PFTs or hemoglobin. The fact that the binary classification (early stop vs. completion) attenuated the effect while continuous exercise time retained it reinforces that there is meaningful prognostic gradient across the full zero to three-minute range, not merely a threshold effect. Recent advances in transplant risk stratification provide important context for these findings. The Composite Health Assessment Risk Model (CHARM), developed in BMT CTN 1704 among allogeneic recipients aged ≥60 years, independently predicts 1-year non-relapse mortality (NRM) and, as recently reported, geriatric morbidity including frailty, disability, and physical function impairment through one year after transplant^24,25. CHARM integrates comorbidity burden, laboratory markers, patient-reported performance status, cognition, and age into a single composite score. However, CHARM does not include any direct performance-based functional capacity measure. The functional assessments evaluated within BMT CTN 1704, including patient-reported IADL and a 4-meter gait speed component of the Physical Frailty Phenotype, capture self-reported or instantaneous functional status rather than sustained exercise capacity. A 4-meter walk has a well-recognized ceiling effect in ambulatory patients: nearly all transplant candidates cleared for HCT can walk 4 meters at near-normal speed. In contrast, sustaining step exercise for 3 minutes, or walking for 6 minutes, stresses the integrated cardiopulmonary-musculoskeletal system sufficiently to reveal functional limitations invisible to brief assessments. Exercise time and composite risk models such as CHARM likely capture complementary constructs, comorbidity burden and global health status on one hand, and performance-based functional reserve on the other. Whether combining them improves prediction beyond either alone is an empirical question that prospective studies with both measures are needed to address.

The dissociation between exercise capacity and gas exchange as prognostic constructs has a clear physiologic basis. The 3-minute step protocol imposes only modest cardiopulmonary stress: mean desaturation was 1.4 percentage points and only 9.5% of patients desaturated by ≥4%. In diseases where exercise desaturation carries strong prognostic value, such as interstitial lung disease and pulmonary arterial hypertension, patients have intrinsic parenchymal or pulmonary vascular disease that produces gas exchange impairment under stress, whereas HCT candidates generally do not.^26,27 Their impaired transplant tolerance is driven by deconditioning, sarcopenia, marrow failure, cumulative toxicity from prior therapy, and frailty: constructs that manifest as reduced exercise capacity rather than abnormal oxygen transfer.

The hemoglobin confounding finding has implications beyond this study. Any analysis of exercise oximetry in HCT candidates that omits hemoglobin will tend to attribute a hemoglobin-driven mortality signal to desaturation, because hemoglobin simultaneously determines both the tendency to desaturate (through oxygen-carrying capacity) and post-transplant mortality (through marrow reserve and chemotherapy intensity). The systematic gradient we observed across analytic approaches, significant without hemoglobin, attenuated with imputed hemoglobin, null with measured hemoglobin, is a clean empirical demonstration of this confounding. This has practical relevance: if other centers are interpreting exercise desaturation as a prognostic marker without hemoglobin adjustment, those associations are likely artifactual. It also illustrates a broader methodological point about the interpretation of exercise oximetry in populations where anemia is prevalent and prognostically significant.

These findings have practical implications for centers performing or considering pre-HCT exercise testing. The 3-minute step test is not a substitute for formal functional capacity assessment such as 6MWD, which provides a validated, continuous, uncapped measure with established reference equations, a defined minimal important difference in the HCT population, and decades of cross-disease prognostic validation.^26,28–31 Where 6MWD is performed, it should be the preferred measure. However, 6MWD is not part of standard pre-HCT evaluation at most centers, and its role in HCT is comparatively underdeveloped relative to its established place in solid organ transplantation. In this practical landscape, a 3-minute step test that takes a few minutes, requires no additional clinic visit, uses equipment already present in the PFT lab, and adds independent prognostic information beyond PFTs and hemoglobin represents a low-burden complement to standard pre-transplant assessment. The exercise duration component is the part that adds value; the gas exchange component, on current evidence, does not.

Several design features strengthen the findings. The consistency of exercise time effects across three outcomes, with monotonically improving model fit, and the corresponding consistent null for desaturation across three analytic approaches and an allogeneic subgroup analysis, argue against either result being a chance finding. The pre-specified analytic framework with explicit hemoglobin sensitivity analyses produces an interpretable confounding demonstration rather than a single point estimate. The PFT data were analyzed using contemporary GLI reference equations, ensuring that the baseline model reflects current standards rather than legacy reference values that varied across the 20-year study period.

Limitations include the single-center retrospective design, which may limit generalizability to centers with different patient populations or step protocols. The step exercise protocol itself is not standardized, so variations in step height, duration, and supervision could affect specific effect estimates. Pace and number of steps were not recorded as part of the step protocol which limits the test’s ability to discriminate patients who completed the protocol, since it cannot distinguish a patient who stepped slowly from one who stepped vigorously. Validated functional capacity measures such as the 6MWD may provide finer prognostic discrimination. Omission of step testing was itself associated with higher mortality (54.9% vs. 43.8%), though the groups were similar in age and transplant type. The higher mortality among untested patients likely reflects both the earlier transplant era (median year 2013 vs. 2017) and exclusion of patients too frail to attempt the test, suggesting that our cohort underestimates the true prognostic gradient by truncating the lower end of the functional spectrum. Importantly, this selection likely biases our findings toward underestimation of the exercise capacity effect, since the prognostic gradient is truncated at its lower end. It also compresses the variance of desaturation in the study population relative to the full HCT population, though this would not change the confounding conclusion. Exercise time is a capped and crude functional measure, and its effect estimate cannot be directly compared to formal functional testing such as 6MWD or cardiopulmonary exercise testing. We did not directly compare step exercise time to 6MWD in the same patients, and such a comparison would clarify how much of the 6MWD prognostic signal a brief step test can recover. HCT-CI and Karnofsky performance status were inconsistently recorded across the 2005–2024 study period. HCT-CI was introduced in 2005 and institutional adoption was gradual. Data completeness was further impacted by an EHR migration in 2018 with substantial loss of prior data. Our study therefore tests whether exercise time adds prognostic value beyond resting PFTs and hemoglobin, which are the standard pre-transplant physiologic measures available across the full study period. In a small subgroup analysis of 483 allogeneic HCT recipients between 2018 and 2024 with HCT-CI data, exercise time showed a directionally consistent but non-significant association with overall survival after HCT-CI adjustment (HR 0.86 per minute, 95% CI 0.75–1.04, p = 0.11), likely due to capped exercise time limiting differentiation, overlapping assessment of comorbidity burden and small sample size. Testing additive value beyond HCT-CI is a primary aim of our planned multicenter prospective validation. We did not analyze non-relapse mortality (NRM), which would require historic relapse date ascertainment not available in our data sources. However, the consistency of exercise time effects across OS, ICU admission, and mechanical ventilation, combined with the absence of a plausible mechanism linking a brief exercise capacity measure to relapse biology, suggests a treatment-tolerance rather than disease-driven signal. NRM is a primary endpoint in a prospective study currently in progress (NCT05766332). Hemoglobin missingness was non-random with respect to era and outcome, but the consistency of findings across the primary, complete-case, and MICE analyses argues that the conclusions are robust to this missingness pattern. The desaturation null finding should be interpreted in the context of this specific low-intensity step protocol, as higher-intensity exercise protocols or formal cardiopulmonary exercise testing may unmask prognostic desaturation that a 3-minute step test does not elicit.

Conclusion

Exercise time during a 3-minute step protocol performed at the time of PFT independently predicts overall survival, ICU admission, and invasive mechanical ventilation after HCT, with consistent effects across all three outcomes. Oxygen desaturation during the same test does not add prognostic value, and the apparent desaturation signal observed without hemoglobin adjustment is a confounding artifact. Among centers performing pre-transplant step exercise testing, exercise duration should be emphasized as a low-burden source of prognostic information. Where formal functional capacity testing such as the six-minute walk test is available, it should be preferred. Whether a brief step test can recover a clinically useful fraction of the 6MWD signal in HCT candidates is a question worth direct comparison.

Highlights.

• Exercise time during a 3-minute step protocol predicts overall survival after HCT

• Exercise time predicts ICU admission and mechanical ventilation need in 120 days

• Oxygen desatuation during exercise was not prognostic

• 3-minute step protocol is easy to do and provide prognostic information