Gait Speed, Survival, and Recommended Treatment Intensity in Older Adults with Blood Cancers Requiring Treatment

Abstract

BACKGROUND

Brief measures of physical function such as gait speed may be useful to optimize treatment intensity for older adults with blood cancers; however, little is known about whether such assessments are already captured within oncologists’ “gestalt” assessments.

METHODS

Gait speed was assessed in 782 patients ≥75 years with blood cancers at our institution, with results reported to providers after treatment decisions were made; 408 required treatment when different intensities were available per National Comprehensive Cancer Network (NCCN) guidelines. We performed structured abstractions of treatment intensity recommendations into standard intensity, reduced intensity, or supportive care, based on NCCN guidelines. We modeled gait speed and survival using Cox regression and performed ordinal logistic regression to assess predictors of NCCN-based categorizations of oncologists’ treatment intensity recommendations, including gait speed.

RESULTS

Median survival by gait speed category was 10.8 months (<0.4 m/s), 18.6 months (0.4–0.6 m/s) 34.0 months (0.6–0.8 m/s), and unreached (>0.8 m/s). Univariable hazard ratios for death increased for each lower category compared to ≥0.8 m/s (0.6–0.8 m/s: HR 1.76; 0.4–0.6 m/s: HR 2.30; <0.4 m/s: HR 3.31). Gait speed predicted survival in multivariable Cox regression (all p<0.05). In multivariable models including age, sex, and ECOG performance status, gait speed did not predict oncologists’ recommended treatment intensity (all p>0.05) and did not add to a base model predicting recommended treatment intensity.

CONCLUSION

In older adults with blood cancers presenting for treatment, gait speed predicted survival but not treatment intensity recommendation. Incorporating gait speed into decision-making may improve optimal treatment selection.

Lay Summary

We assessed if gait speed predicted survival in a population of older adults with blood cancers who required treatment when different intensities were available, and if gait speed was captured in oncologists’ gestalt assessments during treatment recommendation. Gait speed testing was performed on 408 older adults in this population; results were not reported to oncologists. We found that lower gait speeds predicted worse survival; however, clinicians’ recommended treatment intensities did not correlate with gait speed findings. As gait speed predicted survival but was not captured in treatment intensity recommendations, incorporating gait speed into decision-making may improve optimal treatment selection.

Precis

Gait speed predicts survival but not the intensity of recommended treatment in older adults with blood cancers. Incorporating gait speed into decision-making may improve treatment selection.

INTRODUCTION

The majority of patients with blood cancers are older adults.¹ When older adults present for cancer treatment, multiple factors are evaluated: their pre-existing medical problems, the specific features of their malignancy, their personal values, and the potential efficacy and toxicity—the intensity—of available treatments.¹ The increasing number of therapeutic options now allows hematologic oncologists to tailor treatment intensity for older adults. Paradoxically, the advent of tailored treatment intensity in this population has led to increased uncertainty in treatment selection² and greater concern regarding under- and overtreatment.³ Contributing to this uncertainty is the continued underrepresentation of older adults in the clinical trials investigating novel therapies.⁴

Recognizing this dilemma, geriatric oncology has investigated objective assessments to help patients and doctors better evaluate the risks and benefits of available treatment options.⁵ Comprehensive geriatric assessment employs a battery of cognitive, physical, psychological, and functional tests⁶ to detect age-related vulnerability and is now recommended for every older adult with cancer undergoing systemic therapy.⁷ Unfortunately, the availability of geriatric assessment is limited by the time and personnel required to complete it.^1,8,9 Recent evidence suggests that concise measures of physical function may predict treatment effects and survival in older adults as well as resource-intensive comprehensive assessment.^9–13

One such assessment is the 4-meter gait speed test:^11,14 a brief, integrative measure of physical function that does not require significant time, specialized equipment, or expert training to perform. In a prior study, our group evaluated gait speed in 448 older adults with blood cancers and found that reductions in gait speed were independently associated with higher mortality and carried similar predictive value as a more comprehensive assessment.¹¹ While these data demonstrated the predictive value of gait speed for survival in older adults both on and off treatment, it is unknown whether gait speed maintains this predictive power in a population that requires treatment when different intensities are available. Moreover, it is unclear if hematologic oncologists’ “gestalt”^15–17 assessment of physical function approximates formal gait speed assessment during therapeutic intensity selection, or if gait speed testing could provide additional discriminatory value at the time of treatment choice. To answer these questions, we analyzed gait speed, survival, and recommended treatment intensity for a large cohort of older adults with blood cancer who underwent treatment of varying intensities.

MATERIALS AND METHODS

Study Design and Patients

Our overall study design is as previously described^10,11,18 but with an extended enrollment period. From 1 February 2015 to 15 November 2019, all patients 75-years and older who presented for initial consultation in the leukemia, myeloma, or lymphoma clinics at the Dana-Farber Cancer Institute (DFCI), Boston, Massachusetts, were approached to participate in a frailty assessment, which was performed by a nonphysician research assistant. The assessment included the phenotype¹⁹ and cumulative deficit²⁰ frailty assessment methods, National Institutes of Health 4-meter gait speed test,¹⁴ cognitive testing of executive function using the clock-in-the-box (CIB) test,²¹ and the 5-word delayed recall from the Montreal Cognitive Assessment (MOCA).²² Frailty was grouped into robust, prefrail, and frail based on the worse of the two assessment scores. Based on results of our previous work,¹¹ gait speed was categorized in 0.2 m/s increments of <0.4, 0.4–0.6, 0.6–0.8, and >0.8 m/s. CIB results were grouped into scores of normal (7–8), possibly impaired (5–6), and probably impaired (<5) and MOCA word recall into normal (4–5), possibly impaired (3), and probably impaired (<2), both groupings were consistent with prior studies.^10,23 The assistant was trained in these geriatric screening methods by a board-certified geriatrician and observed for competency before screening independently. To ensure independent clinical decision making, treating oncologists were blinded to all assessment results for at least 48 hours. The study was approved by the DFCI Office for the Protection of Human Research Subjects. All participants signed informed consent.

Abstracting Recommended Treatment Intensity

Baseline characteristics including sex, age, cancer type, Eastern Cooperative Oncology Group (ECOG) performance status, recommended treatment, line of treatment, and date of death or last follow up were abstracted from the medical records (a cutoff date of 7 July 2020 was used for survival follow up). Cancers were grouped into nine categories: acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML), myelodysplastic syndrome (MDS), chronic myelomonocytic leukemia (CMML), Hodgkin lymphoma (HL), indolent non-Hodgkin’s lymphoma (e.g., follicular, marginal zone; iNHL), aggressive non-Hodgkin lymphoma (e.g., diffuse large B-cell, Burkitt; aNHL), multiple myeloma (MM), and lymphoplasmacytic lymphoma (LPL). Of the 909 patients approached over the enrollment period, 127 refused and 782 were assessed (overall participation rate, 86.0%). We reviewed the National Comprehensive Cancer Network (NCCN) guidelines for the Treatment of Cancer by Site²⁴ for all diseases seen in this cohort to evaluate if oncologists’ recommended treatments clearly varied by intensity and excluded those where recommendations did not (e.g., BCR-ABL tyrosine kinase inhibitors for chronic myeloid leukemia, ibrutinib for chronic lymphocytic leukemia).^25,26 NCCN guidelines current at the end of the study cohort (15 November 2019) were used. Medical records of the 595 remaining patients were reviewed by a hematologic oncologist (A.H.) to confirm oncologists’ treatment recommendations, excluding consultations where no treatment would have been recommended regardless of frailty status (e.g., low-risk MDS with mild cytopenias). A total of 408 patients were included in the final cohort. Oncologists’ treatment recommendations were then abstracted into standard intensity, reduced intensity, or supportive care following NCCN guidelines, as applicable. These NCCN-based categorizations of oncologists’ treatment intensity recommendations (“recommended treatment intensity”) were then independently recategorized, using the same NCCN guidelines, as applicable, by a disease-specific expert in lymphoma (O.O.), leukemia (M.L.), and myeloma (A.S.). Disputes were resolved by a third hematologic oncologist (G.A.). The mean [k] was 0.77 (mean SE, 0.07).

Statistical Analysis

We calculated descriptive means and proportions for baseline characteristics, including age, sex, specific cancer types, ECOG performance status, CIB score, MOCA word recall score, and gait speed categories. We performed Cox proportional hazards regression to evaluate the relationship between categorical gait speed, other baseline characteristics, and survival, using significant univariable predictors (p<0.05) to build a multivariable model. We compared the concordance statistics (c-statistic)²⁷ of baseline characteristics alone and with the addition of gait speed or frailty indices to compare the discrimination of various models and considered a difference of 0.025 clinically meaningful.¹¹ The proportionality of the hazards assumption was verified using Schoenfeld residual plots and testing for interaction with log person-time. We performed interaction analyses between gait speed and age, sex, ECOG performance status, line of treatment, and specific cancer type for the Cox regression models. Incident death, 1-year survival, 3-year survival, and median survival rates for categories of gait speed were calculated using Kaplan-Meier (KM) survival curves. Subgroup KM survival analysis was performed on patients with ECOG performance status of 0–1 and by the general disease categories of myeloma, lymphoma, and leukemia. We reported proportions for recommended treatment intensity by gait speed. We performed univariate ordinal logistic regressions to assess significant predictors of recommended treatment intensity and multivariate ordinal logistic regressions among significant predictors. To evaluate if treating oncologists’ recommendations already captured the information obtained through gait speed testing, we constructed receiver-operator curves (ROC) and compared areas under the curve (AUCs) between reduced intensity, supportive care treatment, and standard intensity treatment by age, sex, ECOG performance status, and cancer type, with and without gait speed; if gait speed testing information was captured by oncologists’ gestalt assessment, we would expect to see higher AUCs with gait speed added to the model. In all analyses, p<0.05 was considered significant.

RESULTS

Demographics

Mean age of the 408 participants was 79.4 years (95% CI [79.08, 79.87]), 64.2% of participants were male, and mean gait speed was 0.72 m/s (95% CI [0.69, 0.75]). 83.3% of participants had an ECOG performance status of 0–1, 65.4% had normal word recall, and 39.0% performed the CIB normally. Complete demographic characteristics are shown in Table 1.

Table 1.

Cohort Demographics

		N	%
Age	75–79	240	58.8
	80–84	120	29.4
	85+	48	11.8
Sex	Female	146	35.8
	Male	262	64.2
Cancer Type	ALL	5	1.2
	AML	40	9.8
	CMML	7	1.7
	MDS	66	16.2
	HL	6	1.5
	Indolent NHL	33	8.1
	Aggressive NHL	94	23.0
	MM	128	31.4
	LPL	29	7.1
Line of Treatment	First Line	96	23.5
	Second or Later	312	76.5
ECOG PS	0	214	52.5
	1	126	30.9
	2	35	8.6
	3	30	7.4
	4	3	0.7
Frailty Status	Robust	98	24.0
	Prefrail	223	54.7
	Frail	87	21.3
Gait Speed Category	<0.4 m/s	47	11.5
	0.4–0.6 m/s	52	12.8
	0.6–0.8 m/s	129	31.6
	>0.8 m/s	180	44.1
MOCA Word Recall Group	Normal (4–5)	267	65.4
	Possibly Impaired (3)	76	18.6
	Probably Impaired (≤2)	65	15.9
CIB Group	Normal (7–8)	159	39.0
	Possibly Impaired (5–6)	94	23.0
	Probably Impaired (≤4)	155	38.0

N: number. %: percentage. ALL: acute lymphoblastic leukemia. AML: acute myeloid leukemia. CMML: chronic myelomonocytic leukemia. MDS: myelodysplastic syndrome. HL Hodgkin’s lymphoma. NHL: non-Hodgkin’s lymphoma. MM: multiple myeloma. LPL: lymphoplasmacytic lymphoma. ECOG PS: Eastern Cooperative Oncology Group Performance Status. MOCA: Montreal Cognitive Assessment Test. CIB: Clock in the Box Test.

Gait Speed and Survival

Median survival of the entire cohort was 42.0 months; median survival by gait speed category was unreached for the >0.8 m/s group and was 34.0 months (0.6–0.8 m/s), 18.6 months (0.4–0.6 m/s), and 10.8 months (<0.4 m/s). The probability of 1-year and 3-year survival for the entire cohort was 89.1 and 43.3%, respectively. In a univariable Cox regression of survival by gait speed category, the HR for death increased for each lower gait speed category compared to the >0.8 m/s category (0.6–0.8 m/s: HR 1.76, 95% CI [1.22, 2.52]; 0.4–0.6 m/s: HR 2.30, 95% CI [1.47, 3.58]; <0.4 m/s: HR 3.31, 95% CI [2.12, 5.14]). Distinct KM curves for survival by gait speed category could be generated for the whole cohort (Figure 1; log rank test, p<0.001), as well as between the higher three gait speed categories for patients with an ECOG performance status of 0–1 (Figure 2; the <0.4 m/s category was excluded because few patients had low gait speed and preserved performance status). KM subgroup analysis for the myeloma, lymphoma, and myeloid disease categories are shown in the Supplemental Appendix.

Figure 1.

Kaplan-Meier Survival Curves by Gait Speed Category for the Entire Cohort.

Figure 2.

Kaplan-Meier Survival Curves by Gait Speed Category for Gait speeds >0.4 m/s among Patients with ECOG Performance Status 0–1.

In separate univariable Cox regression models, age, ECOG performance status, and line of treatment were predictive of survival, whereas MOCA word recall and CIB score groups were not. In a multivariable Cox regression with significant predictors from univariable analyses (gait speed category, age, ECOG performance status, line of treatment) and sex, adjusted for cancer type, categorical gait speed remained independently predictive of survival (Table 2). There were no significant interactions between covariates. Categorical gait speed remained predictive of survival when recommended treatment intensity was added to the model (all gait speed categories: p<0.05); recommended treatment intensity was not predictive (Supplemental Appendix). The c-statistics of models with baseline covariates with and without the addition of gait speed and frailty assessments are shown in the Supplemental Appendix; the model with gait speed had the highest c-statistic of models tested (0.7695). When assessed as a continuous variable, slower gait speed was predictive of survival in univariable (HR 3.45, 95% CI [2.17, 5.26]) and multivariable Cox regressions (HR 2.94, 95% CI [1.59, 5.56]).

Table 2.

Multivariable Cox Regression of Survival, Adjusted for Cancer Type

Variable (comparator)	Categories	N	Hazard Ratio	95% CI		p-value*
ECOG PS (0)	1	126	1.57	1.10	2.24	0.01
	2	35	1.15	0.67	1.98	0.61
	3	30	1.35	0.72	2.53	0.35
	4	3	1.62	0.43	6.09	0.48
Age	N/A	408	1.08	1.04	1.12	<0.001
Sex (Male)	Female	146	0.66	0.47	0.92	0.03
Line of Treatment (First Line)	≥2 Line	312	5.12	1.81	3.79	<0.001
Gait Speed Category (>0.8 m/s)	<0.4 m/s	47	3.77	2.10	6.77	<0.001
	0.4–0.6 m/s	52	1.75	1.07	2.88	0.03
	0.6–0.8 m/s	129	1.57	1.07	2.30	0.02

CI: confidence interval. ECOG PS: Eastern Cooperative Oncology Group performance status.

^*Significant p-values in bold

Gait Speed and Recommended Treatment Intensity

Oncologists recommended standard intensity treatments for 36.8% of participants and reduced intensity or supportive care treatments for 54.9 and 8.3% of participants, respectively. A two-way table showing the frequencies of treatment intensity categories and gait speed categories individually and together is shown in Table 3. Twenty patients (4.9%) with the lowest two gait speed categories were recommended standard intensity treatment, while 99 patients (24.3%) in the highest gait speed category were recommended reduced intensity treatment or supportive care only. Univariable ordinal logistic regression models identified age, ECOG performance status, and MOCA word recall score, and gait speed categories as potentially significant predictors of recommended treatment intensity. In multivariable regression adjusted for sex and with those univariably significant predictors, cancer type, age, and ECOG performance status were predictive of treatment intensity (Table 4); MOCA word recall score and gait speed category were not consistently predictive. There were no significant interactions between predictors. Models containing cancer type, age, sex and ECOG performance status classified patients by their intensity of treatment with fair to good discriminatory power²⁸ (AUC: 0.73 standard versus reduced intensity, 0.76 standard intensity versus supportive care, 0.81 reduced intensity versus supportive care); prediction did not improve for any comparison with the inclusion of gait speed (AUC 0.73, 0.76, 0.82, p>0.05 for all; Figure 3).

Table 3.

Recommended Treatment Intensity by Categorical Gait Speed

		Gait Speed Category (m/s)								Total N Per Treatment Intensity Category	% of Total
		<0.4		0.4–0.6		0.6–0.8		>0.8
		N	%*	N	%*	N	%*	N	%
Recommended Treatment Intensity Category	Supportive Care	10	21.3	7	13.5	13	10.1	4	2.2	34	8.3
	Reduced Intensity	27	57.5	35	67.3	67	51.9	95	52.8	224	54.9
	Standard Intensity	10	21.3	10	19.2	49	38.0	81	45.0	150	36.8
Total N Per Gait Speed Category		47		52		129		180
% of Total		11.5		12.8		31.6		44.1

m/s: meters per second. N: number. %: percentage.

^*Percentages are additive by column

Table 4.

Multivariable Regression of Recommended Treatment Intensity, Adjusted for Cancer Type and Sex

Variable (comparator)	Categories	N	Odds Ratio	95% CI		p-value*
ECOG PS (0)	1	126	0.49	0.29	0.82	0.006
	2	35	0.32	0.14	0.74	0.008
	3	30	0.32	0.11	0.93	0.04
	4	3	0.12	0.01	1.89	0.13
Age	N/A	408	0.94	0.89	0.99	0.03
Gait Speed Category (>0.8 m/s)	<0.4 m/s	47	0.45	0.18	1.12	0.09
	0.4–0.6 m/s	52	0.47	0.22	1.00	0.05
	0.6–0.8 m/s	129	0.91	0.55	1.53	0.73
MOCA Word Recall Score Group (Not Impaired)	Probably Impaired	76	0.41	0.21	0.78	0.007
	Possibly Impaired	65	1.22	0.69	2.15	0.49

CI: confidence interval. ECOG PS: Eastern Cooperative Oncology Group Performance Status. MOCA: Montreal Cognitive Assessment Test.

^*Significant p-values in bold

Figure 3.

ROC AUCs Comparing Recommendations for Standard Intensity versus Reduced Intensity (A), Standard Intensity versus Supportive Care (B), and Reduced Intensity versus Supportive Care (C) by Cancer type, Age, Sex, and ECOG Performance Status, With and Without Gait Speed.

DISCUSSION

In this cohort of older adults with blood cancers presenting for treatment when different intensities were available, gait speed was predictive of survival but was not captured by oncologists’ gestalt during treatment intensity decision-making. Almost a third of patients in this cohort had a faster gait speed (>0.8 m/s) and were recommended a lower intensity treatment or had a slower gait speed (<0.4 m/s) and were recommended a standard intensity treatment. As the predictive value of treatment intensity by baseline characteristics and oncologist gestalt did not capture the additional information provided by gait speed results, these findings suggest that incorporating gait speed testing into routine treatment decision-making may improve optimal therapeutic selection for this population.

There is a strong correlation between gait speed and survival across a wide variety of persons over age 65.^11,29,30 Survival findings from this cohort are consistent with our previous analysis, which focused on all patients with hematologic malignancies, including those without indications for active treatment and those without different possible treatment intensities. Both cohorts had slower mean gait speeds (0.72 and 0.73 m/s) than other studies of other older adult populations, which were made up of relatively younger participants without cancer diagnoses.^29,31 In the current cohort, gait speed categories distinguished distinct survival curves, which were maintained even when restricted to those with an ECOG performance status of 0–1. Moreover, gait speed retained its predictive value for survival in multivariable analysis, whereas performance status did not. These findings, the ease of measuring gait speed, and its objectivity relative to ECOG performance status, highlight its benefit in this population.

As oncologists in this cohort were not aware of gait speed findings when making treatment recommendations, treatment decisions were based on available information such as patient age, sex, cancer type, performance status, and perhaps “gestalt” physical function assessments. The treatment intensity prediction models suggest that neither clinician gestalt nor other available data captured the same information as gait speed. The multivariable model and AUC values we found (two of three comparisons were fair and one was good by traditional criteria²⁸) demonstrate that while baseline information and gestalt inform these recommendations, there is room for improvement that gait speed testing can offer.

Such information is especially important for the 29.2% of the cohort with potential mismatches between their recommended treatment intensity and their gait speed category. Some patients with fast gait speeds may desire or require reduced intensity therapy, and others with slow gait speeds may request higher intensity treatment. Similarly, reductions in gait speed may be disease-related and acute but reversible while others may be comorbidity-related, chronic, and irreversible.³² Nonetheless, the additional information this simple clinical test reveals about survival may help patients and physicians avoid inappropriately intensive or non-intensive treatment while allaying moral distress over therapeutic uncertainty.³ Although under and over-treatment concerns will not be resolved purely through more objective measures, minimizing the uncertainties that can be reduced can better match treatment decisions to personal values and preferences. Prior studies, which did not focus on blood cancers, have shown that providing oncologists with measures of function and other geriatric assessment domains is associated with appropriate decreases and increases in the intensity of treatment.^33–35 Our findings call for investigating how providing objective gait speed results to hematologic oncologists affects their treatment intensity recommendations in this population.

The strengths of this study include the size of the cohort, the high participation rate, and the completeness of the data. The limitations of the study are that the cohort was derived from a single academic institution in the Northeastern United States, where relatively few patients had ALL, CMML, or HL, and where there was a preponderance of male participants. The predictive value of sex on survival in this cohort has been seen in other studies of gait speed in older adults;^29,36 in this case, sex-related survival differences may also be driven by the relative difference in the proportions of each sex in the disease categories with the shortest and longest survival (30% of males each with myeloid and lymphoid malignancies, 25 and 36% of females each; see Supplemental Appendix). Though we attempted to control for the variation in treatment due to disease by controlling for nine cancer types, we could not control for the nuance of disease risk within each cancer type that may have also influenced treatment recommendations. While we tried to objectively characterize treatment intensity recommendations through our use of the NCCN guidelines as a standard, and by using two independent reviewers and a moderator, the categorization of all treatments into three discrete intensity categories is inherently reductionist and cannot capture the nuance of all possible variations; this includes variations in dose and tolerability-based treatment escalation or de-escalation that may have occurred. Additionally, the use of the NCCN guideline versions current at the end of the study period may have impacted categorizations of patients enrolled earlier. Finally, as this was not an interventional study, we were unable to assess the effect of reporting gait speed test results on clinicians’ recommendations, and if that information would impact survival.

The breadth of treatments available to older adults with blood cancers is rapidly expanding. Choosing the optimal treatment for each patient is difficult and multifaceted, requiring discussions of patients’ values as well as easy-to-use, discriminatory tests. This study demonstrates that gait speed predicts survival in older adults with blood cancer who require treatment when different intensities are available, but that the additive information gait speed testing generates is not captured in current treatment decision-making. As a significant proportion of this population may be recommended treatment intensities that do not match their physical function, as evidenced by their gait speed, hematologic oncologists may benefit from having gait speed data at the time when treatment decisions are considered. Indeed, informing patients of their gait speed, and its implications, may also help them make more informed decisions about their health.