Literature Evaluation Table

ORIGINAL ARTICLE

Factors associated with falls in older adults with cancer: a validated model from the Cancer and Aging Research Group

Tanya M. Wildes1 & Ronald J. Maggiore2 & William P. Tew3 & David Smith4 & Can-Lan Sun4 & Harvey Cohen5 & Supriya G. Mohile2 & Ajeet Gajra6 & Heidi D. Klepin7 & Cynthia Owusu8 & Cary P. Gross9 & Hyman Muss10 & Andrew Chapman11 & Stuart M. Lichtman3 & Vani Katheria4 & Arti Hurria4 & On behalf of the Cancer and Aging Research Group

Received: 16 January 2018 /Accepted: 13 April 2018 /Published online: 28 April 2018 # Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract Background Falls in older adults with cancer are common, yet factors associated with fall-risk are not well-defined and may differ from the general geriatric population. This study aims to develop and validate a model of factors associated with prior falls among older adults with cancer. Methods In this cross-sectional secondary analysis, two cohorts of patients aged ≥ 65 with cancer were examined to develop and validate a model of factors associated with falls in the prior 6 months. Potential independent variables, including demographic and laboratory data and a geriatric assessment (encompassing comorbidities, functional status, physical performance, medica- tions, and psychosocial status), were identified. A multivariate model was developed in the derivation cohort using an exhaustive modeling approach. The model selected for validation offered a low Akaike Information Criteria value and included dichoto- mized variables for ease of clinical use. This model was then applied in the validation cohort. Results The development cohort (N = 498) had a mean age of 73 (range 65–91). Nearly one-fifth (18.2%) reported a fall in the prior 6 months. The selected model comprised nine variables involving functional status, objective physical performance, depression, medications, and renal function. The AUC of the model was 0.72 (95% confidence intervals 0.65–0.78). In the validation cohort (N = 250), the prevalence of prior falls was 23.6%. The AUC of the model in the validation cohort was 0.62 (95% confidence intervals 0.51–0.71). Conclusion In this study, we developed and validated a model of factors associated with prior falls in older adults with cancer. Future study is needed to examine the utility of such a model in prospectively predicting incident falls.

Keywords Neoplasms . Geriatric assessment . Accidental falls . Polypharmacy . Activities of daily living

Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00520-018-4212-3) contains supplementary material, which is available to authorized users.

* Tanya M. Wildes [email protected]

1 Division of Medical Oncology, Washington University School of Medicine, 660 South Euclid Ave, Campus Box 8056, St Louis, MO 63110, USA

2 University of Rochester Medical Center, Rochester, NY, USA 3 Memorial Sloan Kettering Cancer Center, New York, NY, USA 4 City of Hope, Duarte, CA, USA 5 Duke University, Raleigh, Durham, NC, USA

6 Upstate University Hospital, Syracuse, NY, USA 7 Wake Forest University, Wake Forest, NC, USA 8 Case Western Reserve University School of Medicine,

Cleveland, OH, USA 9 Yale University, New Haven, CT, USA 10 Lineberger Cancer Center, University of North Carolina, Chapel

Hill, NC, USA 11 Thomas Jefferson University, Philadelphia, PA, USA

Supportive Care in Cancer (2018) 26:3563–3570 https://doi.org/10.1007/s00520-018-4212-3

Introduction

Falls are a common, costly, and serious event in the lives of older adults. In the general population, one-in-three older adults fall each year [1]. The CDC estimates that, in the USA, the direct annual healthcare cost of falls was $30 bil- lion in 2010 and is expected to reach $54.9 billion by 2020 [2]. The consequences of falls are manifold: falls may result in severe injuries, including fractures and intracranial hem- orrhage, nursing home placement, functional decline, and fear-of-falling [3–5]. Older adults with cancer are at greater risk of serious injuries related to falls than matched controls without cancer [6]. Further, falls are associated with in- creased risk of death and lower quality of life. [7, 8] In older adults with cancer, falls are associated with increased risk of toxicity of chemotherapy [9]. Functional decline subse- quent to a fall may make an older patient less able to tolerate subsequent cancer therapy.

Falls are preventable [10]. Because of the great risk and cost of falls in older adults with cancer, it is imperative that we understand what factors are associated with falls and iden- tify individuals at greater risk for incident falls. Results have been inconsistent across studies, with some studies identifying no factors predictive of falls and others finding no association with factors considered strongly predictive of falls in non- cancer populations. These inconsistencies are likely due to different populations studied, varying definitions of falls used, and differing methods of fall ascertainment [11–15]. Reliable identification of individuals at greater risk for falls will allow targeted intervention to ultimately prevent falls in this vulner- able population.

In this study, we sought to develop and validate a model of factors associated with a higher likelihood of prior falls in the last 6 months in a cohort of older adults with cancer who underwent geriatric assessment prior to initiation of chemotherapy.

Methods

This cross-sectional study is a secondary analysis of baseline data from a previously published, IRB approved, prospective cohort study which aimed to develop and validate a predictive model of chemotherapy toxicity in older adults [16, 17]. Eligible patients were age 65 or older, with a diagnosis of cancer, who were to begin, but had not yet begun, a new course of chemotherapy. Five hundred patients participated in the development cohort in the initial study (enrolled 2006–2009), and 250 participated in the validation cohort in the initial study (enrolled 2008–2012). The development co- hort in the chemotherapy toxicity study was used as the de- velopment cohort in this analysis [9], and the initial validation cohort was used as the validation cohort in this analysis [17].

Participants completed a baseline geriatric assessment [9], in- cluding a self-report of prior falls in the past 6 months, before initiation of a new course of chemotherapy. Cases were ex- cluded if they were missing data on falls.

Clinical data available included demographics, laboratory values, cancer type, stage, and receipt of prior chemotherapy. The geriatric assessment included an assessment of perfor- mance status (both self-reported and clinician-reported), func- tional status, comorbidities, medications, social support, psy- chological state, cognition, and hearing/visual impairment. Medications were categorized by the number of prescription and nonprescription medications as well as the specific classes of drugs. Participants were asked if they had fallen in the 6 months prior to assessment, and, if so, how many times. Data on fall-related injuries or healthcare utilization (e.g., emergency department visits) were not available.

The primary outcome was self-reported falls in the 6 months prior to assessment. It was dichotomized into a yes/no variable and logistic regression was used for analysis. Independent variables examined included demographics, can- cer type and stage, prior chemotherapy, hemoglobin, creati- nine clearance, and all geriatric assessment parameters. For model selection, R glmulti package with exhaustive modeling approach was used to select the best set of independent vari- ables with the best model fitting. [18] An exhaustive modeling approach fits models for every combination of all terms, from smallest (one term versus the outcome) to the largest (all terms included). The Akaike’s information criterion (AIC) and re- sidual were calculated for each potential model. The five best models with lowest AIC and residual were selected. All five models were similarly low in terms of AIC and residual; we chose the model with the fewest variables to move forward. Once the model variables were selected, cut-points for the continuous variables were selected using the Standardized Wilcoxon statistic to ease clinical comprehension and utility of the model. A point scale was developed based on the z statistic for each variable in the final model and rounded up to the nearest integer. The Cancer and Aging Research Group fall score was created by summing up all the points. The area under the curve (AUC) was calculated to determine the dis- crimination of the model. Calibration was examined by com- paring between the observed rate of prior falls versus expected rate of prior falls as estimated by the model for each fall score. Youden index was used to select the best cutoff point for fall score with the highest sensitivity and specificity in classifying the presence or absence of prior falls.

The model derived from the development cohort was then applied to the validation cohort. The fall scores were calculat- ed using the same scoring system. As in the development cohort, the AUC of the model was determined and calibration in the validation set was examined by comparing the observed versus expected rate of prior falls for each fall score. R and SAS 9.3 (SAS institute, Cary, NC) were used for data analysis.

3564 Support Care Cancer (2018) 26:3563–3570

Results

In the development cohort, two participants in the original 500 patient cohort were not evaluable due to missing data on falls, resulting in a cohort of 498 patients. Table 1 pre- sents the baseline data of the patients in the development cohort. In the development cohort, the prevalence of falls in the 6 months prior to assessment was 18.2%. Among those with falls, there were 59 patients who reported 1 fall (11.8%), 13 patients who reported 2 falls (2.6%), and 19 patients who reported 3 falls (3.8%).

The relationship between prior falls and baseline demo- graphic, geriatric assessment, and laboratory characteristics were examined on univariate analysis (Table 2). Factors signif- icantly associated with prior falls comprised a number of vari- ables in different domains, including functional status, comor- bidities, medications and laboratory data. Exhaustive modeling of candidate independent variables resulted in 8 potential models (Supplementary Table). The best model selected to move forward contained nine variables: (1) any dependence in instrumental activities of daily living [IADLs]), (2) slower

objective physical performance (Timed Up and Go test > 13.5 s), (3) polypharmacy (≥ 4 medications), (4) creatinine clearance <30 ml/min), (5) depressed mood (Hospital Anxiety and Depression scale ≥ 8,) (6) selective serotonin reuptake inhibitor (SSRI) use, (7) benzodiazepine use, (8) proton pump inhibitor (PPI) use, and (9) self-report of BI feel like I am slowed down.^ The discrimination of the model was fair, with an AUC of 0.72 (95% confidence intervals 0.65– 0.78). Scores were assigned to each variable as described in the statistical section and presented in Table 3. The fall score was then calculated as the sum of each individual score. The median fall score was 4, ranged from 0 to15 in the development cohort. Comparing to patients with a fall score of 6 or under, those with fall scores 7 or above were 4 times more likely to report falls in the past 6 months (OR 3.86, 95%CI 2.21–6.74, p < 0.001).

The model was then applied to the validation cohort. In the validation cohort, all 250 patients were evaluable. Baseline characteristics of the validation cohort are provided in Table 1. The prevalence of prior falls in the validation cohort was 23.6%, with 15.6% reporting one fall, 4.8% reporting 2 falls, and 3.2% reporting 3 or more falls. The fall score was calculated using the same scoring system as the development cohort. The median fall score was 5 (range 0–16). For the validation cohort, the AUC of the model was 0.62 (95% con- fidence intervals 0.51–0.71). Figure 1 demonstrates the cali- bration plots of the observed prevalence of prior falls versus the expected rate of prior falls associated with the fall score in the development and validation cohorts. Comparing to pa- tients with a fall score of 6 or under, those with fall scores 7 or above were two times more likely to report falls in the past 6 months (OR = 2.34, 95%CI 1.21–4.59, p = 0.01).

Discussion

In this study, we developed and validated a model of factors associated with prior falls within the last 6 months in older adults with cancer. Because prior falls are strong- ly predictive of future falls [19, 20], with prospective validation, this model may inform approaches to assessing fall-risk that can be employed clinically to determine which older adults with cancer are at greater risk for in- cident falls, rather than waiting for them to have their first fall and then reacting. Identification of patients at higher risk is essential, as falls are potentially preventable through fall-prevention interventions, particularly when targeted to those at higher risk [10].

The factors associated with prior falls in our model have face validity, based on associations seen in other studies of older adults with cancer or in general geriatric populations. For example, the need for assistance with IADLs is strongly associated with falls both in community-dwelling older adults

Table 1 Baseline characteristics

Characteristic Development cohort (N = 498)

Validation cohort (N = 250)

Age (median, range) 72 (65–91) 72 (65–94)

Sex

Female 280 (56.2%) 135 (54.0%)

Male 218 (43.8%) 115 (46.0%)

Race

White 425 (85.3%) 208 (83.2%)

Black 41 (8.2%) 19 (7.6%)

Asian 26 (5.2%) 10 (4.0%)

Other 6 (1.2%) 13 (5.2%)

Cancer type

Lung 143 (28.7%) 64 (25.6%)

Gastrointestinal 135 (27.1%) 68 (27.2%)

Gynecologic 86 (17.3%) 18 (7.2%)

Breast 57 (11.4%) 58 (23.6%)

Genitourinary 50 (10.0%) 30 (12.0%)

Other 27 (5.4%) 11 (4.4%)

Cancer Stage

I 23 (4.6%) 9 (3.6%)

II 58 (11.6%) 40 (16.0%)

III 111 (22.3%) 62 (24.8%)

IV 306 (61.4%) 133 (53.2%)

Other 6 (2.4%)

Chemotherapy setting

First-line therapy 353 (70.9%) 175 (70%)

Second or greater line of therapy

145 (29.12%) 75 (30%)

Support Care Cancer (2018) 26:3563–3570 3565

and in older adults with cancer [12, 21]. The report of feeling Bslowed down^ was also associated with prior falls in our model. Interestingly, this association was independent of both depression, antidepressant use, and slow performance on the Timed Up and Go test. Hence, the feeling of being Bslowed down^ is possibly more reflective of fatigue, which has been associated with falls [13]. Self-reported slowing is also report- ed as a component of phenotypic frailty [22], which is

associated with a greater than twofold increased risk of falls in community-dwelling older adults without cancer [23].

Symptoms of depression are associated with prior falls in our model, as seen in other studies of community-dwelling older adults [24]. Of note, use of antidepressants was also significant in our model, independent of symptoms of depres- sion. Others have found SSRI use to be associated with falls [25, 26], even after controlling for residual symptoms of

Table 2 Factors associated with falls on univariate analysis in development cohort

Parameter Fallers (N = 91) Nonfallers (N = 407) P

Age 74.69 ± 0.64 72.79 ± 0.30 0.0077

Function

Instrumental activities of daily living total score 12.32 ± 0.19 13.03 ± 0.09 0.0024

MOS physical score 62.36 ± 2.71 69.74 ± 1.28 0.019

MD-reported Karnofsky performance status 82.16 ± 1.21 85.26 ± 0.57 0.034

Timed Up and Go 14.02 ± 0.68 11.70 ± 0.32 0.039

Comorbidities

Number of comorbidities 3.09 ± 0.17 2.31 ± 0.08 0.0066

Arthritis 52/91 (57.1%) 177/407 (43.5%) 0.020

Medications

Number of daily meds 6.07 ± 0.38 4.95 ± 0.18 0.019

Proton pump inhibitor 36/91 (39.6%) 93/407 (22.8%) 0.0015

Anti-emetics 22/91 (24.2%) 56/407 (13.8%) 0.017

Anti-diarrheal 5/91 (5.5%) 6/407 (1.5%) 0.034

Bowel aid (fiber supplement, stool softener, laxative) 22/91 (24.2%) 58/407 (14.3%) 0.026

Selective serotonin reuptake inhibitor 16/91 (17.6%) 26/407 (6.4%) 0.0014

Benzodiazepine 20/91 (22.0%) 40/407 (9.8%) 0.0024

Sleep aid (e.g., zolpidem) 11/91 (12.1%) 20/407 (4.9%) 0.016

Prescription mineral supplement 40/91 (44.0%) 125/407 (30.7%) 0.029

Other medication 32/91 (35.2%) 93/407 (22.9%) 0.016

Laboratory values

Creatinine clearance 66.57 ± 26.29 73.00 ± 26.58 0.039

Hemoglobin 11.93 ± 0.16 12.31 ± 0.078 0.037

Table 3 Cancer and Aging Research Group fall score Variable Adjusted odds ratio (95%

confidence interval)* |Z| Score**

Dependence in any instrumental activity of daily living 0.53 (0.29, 0.96) 2.09 3

BI feel as if I am slowed down^: Yes 0.95 (0.40, 2.11) 0.13 1

Depression: hospital anxiety and depression scale score ≥ 8 2.09 (0.93, 4.55) 1.83 2 Selective seratonin reuptake inhibitor use 2.57 (1.14, 5.60) 2.34 3

Benzodiazepine use 2.58 (1.17, 5.56) 2.40 3

Proton pump inhibitor use 1.90 (1.03, 3.49) 2.06 3

Polypharmacy: 4 or more daily medications 0.91 (0.49, 1.70) 0.30 1

Creatinine clearance < 30 ml/min 0.78 (0.21, 3.81) 0.35 1

Timed Up and Go > 13.5 s 1.67 (0.88, 3.11) 1.59 2

*Note model includes variables that are not independently significant because the modeling approach used selects the best overall model, even if individual variables are not significant

**Total score is the sum of individual factor scores

3566 Support Care Cancer (2018) 26:3563–3570

depression [25]. The mechanism for this association is not well-understood; however, one study suggested an association between paroxetine use and impaired balance which, hence, increased the risk of falls [27].

Several other medications, in addition to antidepres- sants, were associated with prior falls in our study, includ- ing benzodiazepines and proton pump inhibitors. Benzodiazepines are well-established to be considered po- tentially inappropriate in older adults [28]. They are strongly associated with falls [29, 30] in a dose- dependent manner [31]. We also found an association be- tween proton pump inhibitor (PPI) use and prior falls. Literature is emerging regarding an association between PPI use and falls, with proposed potential mechanisms re- lated to altered calcium homeostasis or impaired B12 ab- sorption with long-term PPI use [32]. Lastly, we found that the use of four or more medications was associated with prior falls, as has been seen in other populations [33]. Polypharmacy is extremely common in older adults with cancer, with 85% of patients in one cohort taking five or more medications [34]. Turner et al. found a statistical cut point of 5.5 medications to be associated with repeated falls [35]. Others have found the impact of polypharmacy on fall-risk to be greatest in the presence of benzodiaze- pines or antidepressants [36].

Impaired renal function was associated with prior falls in our study, an association noted in populations without cancer [37]. Potential explanations for a mechanism of the associa- tion could be related to altered Vitamin D metabolism, with impaired conversion of 25-hydoxy vitamin D to 1,25 dihy- droxy vitamin D [37]. Alternate potential explanations for

the association between falls and impaired renal function in- clude muscle weakness associated with chronic kidney dis- ease [38], increased risk of adverse drug reactions [39], or supratherapeutic drug levels [40], which are areas to explore in future research.

Finally, slowed performance on the Timed Up and Go (TUG) was associated with prior falls in our study; mea- sures of physical performance have been associated with falls in other studies of populations with cancer [41]. Some studies of community-dwelling older adults suggest that alternate tests, such as gait speed, may have greater utility, and that the TUG should not be used in isolation for predicting fall-risk [42]. Indeed, in our study, the TUG was only one of nine factors comprising the fall score, supporting the multifactorial nature of falls that physical function alone does not entirely encompass.

Strengths of our study include a large, diverse cohort of older adults with multiple types of malignancy, allowing application to a broad population of older adults with cancer. In addition, the model was developed and validat- ed in separate populations, supporting its generalizability. We used advanced statistical techniques to create a robust, generalizable, clinically useful model of the likelihood of prior falls. While the AUC of the model was lower in the validation cohort than in the development cohort, the dec- rement in the AUC related to the optimism of the model in the development cohort is on par with that seen in simulation models using bootstrapping in development and validation cohorts [43].

Nonetheless, this study has several limitations. First, the cross-sectional design does not allow for prospective

Fig. 1 Plots of observed and expected fall rate by fall score. a) Development Cohort (N=498) b) Validation Cohort (n=250)

Support Care Cancer (2018) 26:3563–3570 3567

assessment of incident falls and their relationship with baseline factors. Retrospective recall of falls has been shown to be unreliable, resulting in misclassification bias. Consensus groups have recommended that, optimally, a primary outcome of falls should be ascertained prospec- tively, grounded with a validated definition of falls to limit subjectivity in participants’ appraisal of what events constitute a fall [44]. Recall bias may result in differential reporting of prior falls and spurious associations.

Second, the cross-sectional design only allows us to iden- tify factors associated with prior falls, rather than factors pro- spectively predictive of incident falls; it is possible that some of the associations actually resulted from the prior fall, such as slowed gait speed or increased symptoms of depression. It is unknown how this model, prospectively applied, will compare with other approaches to fall-risk prediction. Finally, the avail- able data did not allow us to explore some candidate factors associated with falls in other studies of older adults with can- cer, such as pain, brain metastases or prior falls. [31] We are also unable to explore consequences of falls, such as whether falls resulted in injury. Finally, the AUC of the model in the validation set was only fair, at 0.62, with 95% confidence intervals of 0.51–0.71, the lower bound approaching chance. Unfortunately, this highlights global limitations in our ability to predict falls in older adults, as model of falls in community- dwelling older adults using commonly used predictors, in- cluding prior falls, gait speed, and the Short Physical Performance Battery, have AUCs of 0.57–0.59 [45]. A pro- spective study to validate the CARG fall score and explore time-varying predictors is underway [NCT02912273].

Another limitation is that factors that may be associated with falls in older adults with cancer, particularly peripheral neuropathy, sarcopenia, or brain metastases, are not available in our dataset. For example, data on the presence of pre- existing neuropathy or chemotherapy-induce peripheral neu- ropathy for those receiving second or subsequent lines of ther- apy are not available. Peripheral neuropathy is a well- established risk factor for falls in older adults with cancer [46–49]. Similarly, we have data on neither the presence of sarcopenia, which is associated with falls in community- dwelling older adults [50], nor the presence of brain metasta- ses, which was associated with falls in older adults with ad- vanced cancer [31].

In conclusion, the Cancer and Aging Research Group fall score provides insight into factors associated with prior falls in older adults with cancer. The factors associated with prior falls in our study include functional dependence, self-reported slowing, depression, use of SSRIs, benzodiazepines or PPIs, polypharmacy (i.e., use of ≥ 4 medications), impaired physical performance on the TUG and impaired renal function. In the future, we anticipate that more robust models will be devel- oped and validated, incorporating the factors we have identi- fied, along with other potentially relevant factors such as

chemotherapy-induced peripheral neuropathy, brain metasta- ses, and other cancer-specific variables, to improve our ability to identify older adults with cancer at increased risk for falls in order to intervene and prevent falls.

Funding Sources This publication was made possible by Grant Number 1K12CA167540 through the National Cancer Institute (NCI) at the National Institutes of Health (NIH) and Grant Number UL1 TR000448 through the Clinical and Translational Science Award (CTSA) program of the National Center for Advancing Translational Sciences (NCATS) at the National Institutes of Health. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of NCI, NCATS or NIH. (T.W.) This work was also supported by the National Institutes of Health, National Institute on Aging Grant No. K23- AG026749–01 (A.H.), Paul Beeson Career Development Award in Aging Research, and American Society of Clinical Oncology, Association of Specialty Professors, Junior Development Award in Geriatric Oncology (A.H.).

Research reported in this publication included work performed in the Survey Research Core supported by the National Cancer Institute of the National Institutes of Health under award number P30CA33572. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Funded in part through the Memorial Sloan Kettering Cancer Center Support Grant P30 CA008748 (S.L.)

Compliance with ethical standards

Conflict of interest The authors declare that they have no conflict of interest.

References

1. (2008) WHO global report on falls Prevention in older age. 1–53 2. Englander F, Hodson TJ, Terregrossa RA (1996) Economic dimen-

sions of slip and fall injuries. J Forensic Sci 41:733–746 3. Owens PL, Russo CA, Spector W, Mutter R (2006) Emergency

Department Visits for Injurious Falls among the Elderly, 2006: Statistical Brief #80. Agency for Health Care Policy and Research (US), Rockville (MD)

4. Collerton J, Kingston A, Bond J, Davies K, Eccles MP, Jagger C, Kirkwood TBL, Newton JL (2012) The personal and health service impact of falls in 85 year olds: cross-sectional findings from the Newcastle 85+ cohort study. PLoS One 7:e33078. https://doi.org/ 10.1371/journal.pone.0033078

5. Tinetti ME, Williams CS (1997) Falls, injuries due to falls, and the risk of admission to a nursing home. N Engl J Med 337:1279–1284. https://doi.org/10.1056/NEJM199710303371806

6. Toomey A, Friedman L (2014) Mortality in cancer patients after a fall-related injury: the impact of cancer spread and type. Injury 45: 1710–1716. https://doi.org/10.1016/j.injury.2014.03.008

7. Huang MH, Lytle T, Miller KA, Smith K, Fredrickson K (2014) History of falls, balance performance, and quality of life in older cancer survivors. Gait & Posture 40:451–456. https://doi.org/10. 1016/j.gaitpost.2014.05.015

8. Sylliaas H, Idland G, Sandvik L, Forsen L, Bergland A (2009) Does mortality of the aged increase with the number of falls? Results from a nine-year follow-up study. Eur J Epidemiol 24:351–355. https://doi.org/10.1007/s10654-009-9348-5

9. Hurria A, Togawa K, Mohile SG, Owusu C, Klepin HD, Gross CP, Lichtman SM, Gajra A, Bhatia S, Katheria V, Klapper S, Hansen K,

3568 Support Care Cancer (2018) 26:3563–3570

Ramani R, Lachs M, Wong FL, Tew WP (2011) Predicting chemo- therapy toxicity in older adults with Cancer: a prospective multi- center study. J Clin Oncol 29:3457–3465. https://doi.org/10.1200/ JCO.2011.34.7625

10. Gillespie LD, Robertson MC, Gillespie WJ, Sherrington C, Gates S, Clemson LM, Lamb SE, Cochrane Bone, Joint and Muscle Trauma Group (2012) Interventions for preventing falls in older people living in the community. Cochrane Database Syst Rev 9:CD007146. https://doi.org/10.1002/ 14651858.CD007146.pub3

11. Wildes TM, Dua P, Fowler SA, Miller JP, Carpenter CR, Avidan MS, Stark S (2015) Systematic review of falls in older adults with cancer. J Geriatr Oncol 6:70–83. https://doi.org/10.1016/j.jgo.2014. 10.003

12. Williams GR, Deal AM, Nyrop KA, Pergolotti M, Guerard EJ, Jolly TA, Muss HB (2015) Geriatric assessment as an aide to un- derstanding falls in older adults with cancer. Support Care Cancer 23:2273–2280. https://doi.org/10.1007/s00520-014-2598-0

13. Vande Walle N, Kenis C, Heeren P, van Puyvelde K, Decoster L, Beyer I, Conings G, Flamaing J, Lobelle JP, Wildiers H, Milisen K (2014) Fall predictors in older cancer patients: a multicenter pro- spective study. BMC Geriatr 14:135. https://doi.org/10.1186/1471- 2318-14-135

14. Sattar S, Alibhai SMH, Spoelstra SL, Fazelzad R, Puts MTE (2016) Falls in older adults with cancer: a systematic review of prevalence, injurious falls, and impact on cancer treatment. Support Care Cancer 24:4459–4469. https://doi.org/10.1007/s00520-016-3342-8

15. Puts MTE, Monette J, Girre V, Wolfson C, Monette M, Batist G, Bergman H (2013) The fall rate of older community-dwelling can- cer patients. Support Care Cancer 21:775–783. https://doi.org/10. 1007/s00520-012-1579-4

16. Hurria A, Cirrincione CT, Muss HB, Kornblith AB, Barry W, Artz AS, Schmieder L, Ansari R, Tew WP, Weckstein D, Kirshner J, Togawa K, Hansen K, Katheria V, Stone R, Galinsky I, Postiglione J, Cohen HJ (2011) Implementing a geriatric assessment in cooperative group clinical cancer trials: CALGB 360401. J Clin Oncol 29:1290–1296. https://doi.org/ 10.1200/JCO.2010.30.6985

17. Hurria A, Mohile S, Gajra A, Klepin H, Muss H, Chapman A, Feng T, Smith D, Sun CL, de Glas N, Cohen HJ, Katheria V, Doan C, Zavala L, Levi A, Akiba C, Tew WP (2016) Validation of a prediction tool for chemotherapy toxicity in older adults with cancer. J Clin Oncol 34:2366–2371. https://doi.org/10. 1200/JCO.2015.65.4327

18. Calcagno V, de Mazancourt C (2010) glmulti: an R package for easy automated model selection with (generalized) linear models. J Stat Softw 34:1–29

19. Ganz DA, Bao Y, Shekelle PG, Rubenstein LZ (2007) Will my patient fall? JAMA 297:77–86. https://doi.org/10.1001/ jama.297.1.77

20. Stone C, Lawlor PG, Nolan B, Kenny RA (2011) A prospective study of the incidence of falls in patients with advanced cancer. J Pain Symptom Manag 42:535–540. https://doi.org/10.1016/j. jpainsymman.2011.01.006

21. Chen T-Y, Janke MC (2013) Predictors of falls among community- dwelling older adults with cancer: results from the health and retire- ment study. Support Care Cancer 22:479–485. https://doi.org/10. 1007/s00520-013-2000-7

22. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, Seeman T, Tracy R, Kop WJ, Burke G, McBurnie MA (2001) Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 56:M146–M156

23. Ensrud KE, Ewing SK, Taylor BC, Fink HA, Cawthon PM, Stone KL, Hillier TA, Cauley JA, Hochberg MC, Rodondi N, Tracy JK, Cummings SR (2008) Comparison of 2 frailty indexes for predic- tion of falls, disability, fractures, and death in older women. Arch

Intern Med 168:382–389. https://doi.org/10.1001/archinternmed. 2007.113

24. Menant JC, Wong AKW, Trollor JN, Close JCT, Lord SR (2016) Depressive symptoms and orthostatic hypotension are risk factors for unexplained falls in community-living older people. J Am Geriatr Soc 64:1073–1078. https://doi.org/10.1111/jgs.14104

25. Carrière I, Farré A, Norton J, Wyart M, Tzourio C, Noize P, Pérès K, Fourrier-Réglat A, Ancelin ML (2016) Patterns of selective se- rotonin reuptake inhibitor use and risk of falls and fractures in community-dwelling elderly people: the Three-City cohort. Osteoporos Int 27:3187–3195. https://doi.org/10.1007/s00198- 016-3667-7

26. Gebara MA, Lipsey KL, Karp JF, Nash MC, Iaboni A, Lenze EJ (2015) Cause or effect? Selective serotonin reuptake inhibitors and falls in older adults: a systematic review. Am J Geriatr Psychiatry 23:1016–1028. https://doi.org/10.1016/j.jagp.2014.11.004

27. Hegeman J, van den Bemt B, Weerdesteyn V, Nienhuis B, van Limbeek J, Duysens J (2011) Unraveling the association between SSRI use and falls: an experimental study of risk factors for acci- dental falls in long-term paroxetine users. Clin Neuropharmacol 34: 210–215. https://doi.org/10.1097/WNF.0b013e31823337d1

28. The American Geriatrics Society 2015 Beers Criteria Update Expert Panel (2015) American Geriatrics Society 2015 updated beers criteria for potentially inappropriate medication use in older adults. J Am Geriatr Soc 63:2227–2246. https://doi.org/10.1111/ jgs.13702

29. Woolcott JC, Richardson KJ, Wiens MO, Patel B, Marin J, Khan KM, Marra CA (2009) Meta-analysis of the impact of 9 medication classes on falls in elderly persons. Arch Intern Med 169:1952–1960

30. Leipzig RM, Cumming RG (1999) Drugs and falls in older people: a systematic review and Meta-analysis: I. Psychotropic drugs. J Am Geriatr Soc 47:30–39. https://doi.org/10.1111/j.1532-5415.1999. tb01898.x

31. Stone CA, Lawlor PG, Savva GM, Bennett K, Kenny RA (2012) Prospective study of falls and risk factors for falls in adults with advanced cancer. J Clin Oncol 30:2128–2133. https://doi.org/10. 1200/JCO.2011.40.7791

32. Lewis JR, Barre D, Zhu K, Ivey KL, Lim EEM, Hughes J, Prince RL (2014) Long-term proton pump inhibitor therapy and falls and fractures in elderly women: a prospective cohort study. J Bone Miner Res 29:2489–2497. https://doi.org/10.1002/jbmr.2279

33. Huang ES, Karter AJ, Danielson KK, Warton EM, Ahmed AT (2010) The association between the number of prescription medi- cations and incident falls in a multi-ethnic population of adult type- 2 diabetes patients: the diabetes and aging study. J Gen Intern Med 25:141–146. https://doi.org/10.1007/s11606-009-1179-2

34. Nightingale G, Hajjar E, Swartz K, Andrel-Sendecki J, Chapman A (2015) Evaluation of a pharmacist-led medication assessment used to identify prevalence of and associations with polypharmacy and potentially inappropriate medication use among ambulatory senior adults with cancer. J Clin Oncol 33:1453–1459. https://doi.org/10. 1200/JCO.2014.58.7550

35. Turner JP, Jamsen KM, Shakib S, Singhal N, Prowse R, Bell JS (2016) Polypharmacy cut-points in older people with cancer: how many medications are too many? Support Care Cancer 24:1831– 1840. https://doi.org/10.1007/s00520-015-2970-8

36. Richardson K, Bennett K, Kenny RA (2014) Polypharmacy includ- ing falls risk-increasing medications and subsequent falls in community-dwelling middle-aged and older adults. Age Ageing 44:90–96. https://doi.org/10.1093/ageing/afu141

37. Gallagher JC, Rapuri P, Smith L (2007) Falls are associated with decreased renal function and insufficient calcitriol production by the kidney. J Steroid Biochem Mol Biol 103:610–613. https://doi. org/10.1016/j.jsbmb.2006.12.082

38. Boudville N, Inderjeeth C, Elder GJ, Glendenning P (2010) Association between 25-hydroxyvitamin D, somatic muscle

Support Care Cancer (2018) 26:3563–3570 3569

weakness and falls risk in end-stage renal failure. Clin Endocrinol 73:299–304. https://doi.org/10.1111/j.1365-2265.2010.03821.x

39. Corsonello A, Pedone C, Corica F, Mussi C, Carbonin P, Antonelli Incalzi R, Gruppo Italiano di Farmacovigilanza nell'Anziano (GIFA) Investigators (2005) Concealed renal in- sufficiency and adverse drug reactions in elderly hospitalized patients. Arch Intern Med 165:790–795. https://doi.org/10. 1001/archinte.165.7.790

40. Hartholt KA, Becker ML, van der Cammen TJM (2016) Drug- induced falls in older persons: is there a role for therapeutic drug monitoring? Ther Adv Drug Saf 7:39–42. https://doi.org/10.1177/ 2042098615627806

41. Overcash JA, Rivera HR (2008) Physical performance evalu- ation of older cancer patients: a preliminary study. Crit Rev Oncol Hematol 68:233–241. https://doi.org/10.1016/j. critrevonc.2008.07.001

42. Barry E, Galvin R, Keogh C, Horgan F, Fahey T (2014) Is the Timed Up and Go test a useful predictor of risk of falls in commu- nity dwelling older adults: a systematic review and meta- analysis. BMC Geriatr 14:14. https://doi.org/10.1186/1471-2318-14-14

43. Steyerberg EW, Bleeker SE, Moll HA (2003) Internal and external validation of predictive models: a simulation study of bias and precision in small samples. J Clin Epidemiol 56:441–447. https:// doi.org/10.1016/S0895-4356(03)00047-7

44. Lamb SE, Jørstad-Stein EC, Hauer K et al (2005) Development of a common outcome data set for fall injury prevention trials: the pre- vention of falls network Europe consensus. J Am Geriatr Soc 53: 1618–1622. https://doi.org/10.1111/j.1532-5415.2005.53455.x

45. Palumbo P, Palmerini L, Bandinelli S, Chiari L (2015) Fall risk assessment tools for elderly living in the community: can we do better? PLoS One 10:e0146247. https://doi.org/10.1371/journal. pone.0146247

46. Ward PR, Wong MD, Moore R, Naeim A (2014) Fall-related inju- ries in elderly cancer patients treated with neurotoxic chemothera- py: a retrospective cohort study. J Geriatr Oncol 5:57–64. https:// doi.org/10.1016/j.jgo.2013.10.002

47. Tofthagen C, Overcash J, Kip K (2012) Falls in persons with chemotherapy-induced peripheral neuropathy. Support Care Cancer 20:583–589. https://doi.org/10.1007/s00520-011-1127-7

48. Gewandter JS, Fan L, Magnuson A, Mustian K, Peppone L, Heckler C, Hopkins J, Tejani M, Morrow GR, Mohile SG (2013) Falls and functional impairments in cancer survivors with chemotherapy-induced peripheral neuropathy (CIPN): a University of Rochester CCOP study. Support Care Cancer 21: 2059–2066. https://doi.org/10.1007/s00520-013-1766-y

49. Winters-Stone KM, Horak F, Jacobs PG (2017) Falls, functioning, and disability among women with persistent symptoms of chemotherapy-induced peripheral neuropathy. J Clin Epidemiol 35:2593–2594. https://doi.org/10.1200/JCO.2017.73.6207

50. Scott D, Hayes A, Sanders KM, Aitken D, Ebeling PR, Jones G (2014) Operational definitions of sarcopenia and their associations with 5-year changes in falls risk in community-dwelling middle- aged and older adults. Osteoporos Int 25:187–193. https://doi.org/ 10.1007/s00198-013-2431-5

3570 Support Care Cancer (2018) 26:3563–3570

Supportive Care in Cancer is a copyright of Springer, 2018. All Rights Reserved.