Frailty as an Effect Modifier in Randomized Controlled Trials: A Systematic Review

Background The effect of clinical interventions may vary by patients’ frailty status. Understanding treatment effect heterogeneity by frailty could lead to frailty-guided treatment strategies and reduce overtreatment and undertreatment. This systematic review aimed to examine the effect modification by frailty in randomized controlled trials (RCTs) that evaluate pharmacological, non-pharmacological, and multicomponent interventions. Methods We searched PubMed, Web of Science, EMBASE, and ClinicalTrial.gov, from their inception to 8 December 2023. Two reviewers independently extracted trial data and examined the study quality with senior authors. Results Sixty-one RCTs that evaluated the interaction between frailty and treatment effects in older adults were included. Frailty was evaluated using different tools such as the deficit accumulation frailty index, frailty phenotype, and other methods. The effect of several pharmacological interventions (e.g., edoxaban, sacubitril/valsartan, prasugrel, and chemotherapy) varied according to the degree of frailty, whereas other treatments (e.g., antihypertensives, vaccinations, osteoporosis medications, and androgen medications) demonstrated consistent benefits across different frailty levels. Some non-pharmacological interventions had greater benefits in patients with higher (e.g., chair yoga, functional walking, physical rehabilitation, and higher dose exercise program) or lower (e.g., intensive lifestyle intervention, psychosocial intervention) levels of frailty, while others (e.g., resistance-type exercise training, moderate-intensive physical activity, walking and nutrition or walking) produced similar intervention effects. Specific combined interventions (e.g., hospital-based disease management programs) demonstrated inconsistent effects across different frailty levels. Discussion The efficacy of clinical interventions often varied by frailty levels, suggesting that frailty is an important factor to consider in recommending clinical interventions in older adults. Registration PROSPERO registration number CRD42021283051. Supplementary Information The online version contains supplementary material available at 10.1007/s11606-024-08732-8.


BACKGROUND
Frailty is a clinical state of reduced physiologic reserve and increased vulnerability to poor health outcomes.The prevalence of frailty is 12 to 24% in community-dwelling older adults 1 and almost 50% in hospitalized patients 2 and nursing home residents. 3In the United States, 15% of the older population was frail and 45% was pre-frail. 4hose with frailty are predisposed to adverse health events, including falls, disability, dementia, hospitalization, institutionalization in long-term care, and death. 5,6 mpared with a robust group, pre-frail and frail older adults incurred more healthcare costs. 7linicians increasingly consider frailty in treatment decision-making due to its association with poor treatment outcomes. 1Nonetheless, our understanding on how the benefits and risks of clinical interventions vary by patients' frailty remains limited.A more nuanced understanding of treatment effect heterogeneity by frailty could lead to frailty-guided treatment strategies and reduce overtreatment and undertreatment, which could lead to improved health outcomes, better quality of life, and more targeted use of healthcare resources.For instance, a robust patient might tolerate a more aggressive treatment regimen potentially leading to improved disease control or even cure.In contrast, a frail patient with depleted physiologic reserve might benefit more from a less invasive, more supportive approach focused on symptom management.To answer this question, frailty subgroup analyses of randomized controlled trials (RCTs) are increasingly conducted to investigate treatment effect heterogeneity by frailty. 8his systematic review was conducted to synthesize the findings from RCTs that assessed treatment effects stratified by participants' frailty levels.We examined how frailty was assessed in RCTs and whether the efficacy and safety of interventions varied by frailty category.

METHODS
We followed the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) statement to conduct a systematic review. 9The protocol was registered in the PROSPERO international prospective register of systematic reviews (registration number CRD42021283051, https:// www.crd.york.ac.uk/ PROSP ERO/ displ ay_ record.php?Recor dID= 283051) on 2 November 2021.
Our review addressed a key question: Did frailty modify intervention effects in RCTs of pharmacological, non-pharmacological, and multicomponent interventions?

Role of the Funding Source
This study was supported by the National Institute on Aging of the National Institutes of Health.The funding source had no role in the design, collection, analysis, or interpretation of the data, or the decision to submit the manuscript for publication.

Study Selection and Characteristics of the Included Trials
Figure 1 details study selection.Our database searches yielded 5917 references, of which 61 articles  met the inclusion criteria. The icluded trials are summarized by the type of frailty assessment: frailty index (24 trials) (Table 1), frailty phenotype (17 trials) (Table 2), and other assessments (20 trials) (Table 3).Further details of each trial can be found in Appendix Table .The mean age of trial populations ranged from 58.7 11 to 87.1 12 years and the proportion of women ranged from 23.4 29 to 93.3% 14 (in particular, one RCT was 100% male 15 and the other 100% female 16 ).There were 26 trials evaluating pharmacological interventions (sample size 40 17 to 31,989 18 ), 27 trials on non-pharmacological interventions (sample size 30 14,18 to 5145), 11 and 8 trials on multicomponent interventions (sample size 173 19 to 1464).20

Quality of the Included Trials
Most of the included articles have low likelihood of bias.Risk of bias related to blinding of participants and personnel and incomplete outcome data were identified in 18 and 12 trials, respectively.The risk of bias for each included trial is shown in both the Appendix Figure and Table.

Evaluation of Treatment Effect by Deficit Accumulation Frailty Index
Pharmacologic Interventions.[34] Anticoagulants In a trial of 20,867 adults with atrial fibrillation conducted over 34 months across 46 countries, 21 edoxaban was associated with lower rates of major bleeding compared with warfarin in patients with frailty index 0.12 to <0.36 (edoxaban 60 mg) and patients with frailty index <0.36(edoxaban 30 mg), but not in patients with frailty index <0.12(edoxaban 60 mg) and frailty index 0.36 to <1.0 (edoxaban 30 mg or 60 mg) (p-for-interaction=not reported [NR]).There was a significant treatment effect of edoxaban compared with warfarin on stroke or systemic embolism, but the treatment effect was not different across the frailty spectrum (p-for-interaction=NR).

Systolic blood pressure control
Modified Katz-15 score: NR There was no evidence that the effect of frailty screening alone or with nurse-led care program compared with usual care on modified Katz-15 score was difference across the frailty spectrum (p-for-interaction = NR) natriuretic peptide, 29 dapagliflozin was associated with lower rates of cardiovascular death (p-for-interaction=NR) and allcause death (p-for-interaction=NR) compared with placebo in patients with frailty index ≤0.210,but not in patients with frailty index >0.210.Besides, dapagliflozin was associated with lower rates of cardiovascular outcome (p-for-interaction=NR) compared with placebo in patients with frailty index ≤0.210or frailty index ≥0.311, 29,31 ut not in patients with frailty 0.210 to <0.311.Another trial found dapagliflozin was associated with lower rates of HF hospitalization or cardiovascular death (p-for-interaction=NR) compared with placebo in patients with frailty index <0.311,but not in patients with frailty index ≥0.311. 30There was significant treatment effect of spironolactone compared with placebo 27 on HF hospitalization or cardiovascular death, but the treatment effect was not different across frailty spectrum (p-for-interaction = 0.

Multicomponent intervention
Intervention: GA-intervention a vs usual-care Population: 541 patients with incurable cancer and impairment on ≥ 1 GA domain Follow-up: NR Reference: Gilmore, 2021, US 39 50-item frailty index (range, 0-1) A: frailty index (0 ≤ to < 0.2) B: frailty index (0.2 ≤ to < 0.35) C: frailty index (0.35 ≤ to 1) There was no evidence that the effect of GA-intervention compared with usual care on conversations (p-forinteraction = 0.6111), concerns acknowledged (p-forinteraction = 0.7397), concerns addressed (p-for-interaction = 0.9403) was different across the frailty spectrum There was no evidence that the effect of pharmacist-led deprescribing intervention compared with usual care on changes in DBI 0.5 or more, ACB was different across the frailty spectrum (p-for-interaction = NR) Abbreviations: HR, hazard ratio; NR, not reported; RR, relative risk; MD, mean difference; HFpEF, HF with preserved ejection fraction; HF, heart failure; WOMAC, Western Ontario and McMaster Universities; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; MACE, major adverse cardiovascular events; HFrEF, heart failure with reduced ejection fraction; GA, geriatric assessment; DBI, Drug Burden Index; ACB, anticholinergic cognitive burden * Data were extracted from supplemental files † Data were extracted by GetData Graph Digitizer software a The geriatric assessment (GA) is a validated multidisciplinary evaluation of the functional, psychosocial, physical, and cognitive abilities of older adults, as well as their comorbidities and medication use.Only patients and oncologists in the GA-intervention arm received a summary of the GA plus a list of GA-guided recommendations to address specific impairments (i.e., GA-interventions) b The pharmacist recorded all medications and supplements the person was taking.However, the pharmacist did not have access to the participants' clinical notes or medication records and all clinical decision-making, including prescribing, remained with the GP Fried frailty criteria (range, 0-5) A: Frailty phenotype (1 to 2) B: Frailty phenotype (3 to 5)

Refer to Appendix Table
There was no evidence that the effect of LSRT and HSRT compared with control on physical performance was different across the frailty spectrum (p-for-interaction = NR)   Fried frailty criteria (range, 0-5) A: frailty phenotype (1 to 2) B: frailty phenotype (3 to 5)

Changes in physical function:
NR There was no evidence that the effect of multimodal intervention compared with usual care on physical function was different across the frailty spectrum (p-for-interaction = 0.49) p-for-interaction: an indicator used to infer whether the effect of an intervention on outcome measures is significant at different levels of frailty * Data were extracted from supplemental files † Data were extracted by GetData Graph Digitizer software ‡ Calculations are based on published data in the literature a A multimodal intervention composed of (i) an individualized and progressive resistance exercise program for 16 weeks; (ii) a structured diabetes and nutritional educational program over seven sessions; and (iii) investigator-linked training to ensure optimal diabetes care b The multidomain intervention (MI) consisted of 2-h group sessions focusing on three domains (cognitive stimulation, physical activity, and nutrition) and a preventive consultation (at baseline, 12 months, and 24 months) and for omega-3 polyunsaturated fatty acids supplementation, participants took two capsules of either placebo or polyunsaturated fatty acids daily c The n-3 PUFA supplementation group consumed a daily dose of DHA (800 mg) and EPA (a maximum amount of 225 mg) for 3 years d Intervention was coordinated by two physiotherapists with extensive relevant experience and delivered by an interdisciplinary team comprising the physiotherapists, a dietician, a geriatrician, a rehabilitation physician and a nurse.Intervention was delivered primarily in the participants' homes, with additional outpatient appointments (for example, dietician, continence clinic), occupational therapy, and community exercise programs offered as indicated e Mobility during the preceding month was quantified in terms of distance and frequency of travel and degree of independence using the University of Alabama at Birmingham Life Space Assessment.Scored on a continuous scale from 0 to 120, a higher score illustrates greater life space  There was no evidence that the effect of high-dose inactivated influenza vaccine compared with standarddose vaccine on laboratoryconfirmed influenza caused by any viral type/subtype (regardless of similarity to the vaccine) was different across the frailty spectrum (p-for-interaction = 0.838) Study of Osteoporotic Fractures frailty index (range, 0-3 criteria) A: class 1 (0 to 1) B: class 2 (2 to 3) 400-m gait-speed (6-month) A: mean difference 0.029 (0.017, 0.041) B: mean difference 0.027 (− 0.001, 0.055) (12-month) A: mean difference 0.023 (0.011, 0.035) B: mean difference 0.014 (− 0.015, 0.042) (24-month) A: mean difference 0.023 (0.010, 0.035) B: mean difference 0.010 (− 0.020, 0.039) 4-m gait-speed (6-month) A: mean difference − 0.004 (− 0.018, 0.011) B: mean difference − 0.011 (− 0.041, 0.018) (12-month) A: mean difference − 0.002 (− 0.016, 0.013) B: mean difference − 0.008 (− 0.039, 0.012) (24-month) A: mean difference − 0.001 (− 0.014, 0.016) B: mean difference − 0.010 (− 0.022, 0.41) There was no evidence that the effect of physical activity compared with health education on 400-m gaitspeed, 4-m gait-speed was different across the frailty spectrum (p-forinteraction = NR)   ) is a person-centered and integrated care service for community-living older adults.A multidisciplinary care team consisting of the older adults' GP, a nursing home physician, and two case managers (district nurse and social worker) provides care and support to older adults e A physical therapist assessed each participant for potential impairments in physical abilities and assessed the participant's home environment.Detailed algorithms and decision rules were developed to link the results of the assessment with the recommended interventions.The program was designed to include an average of 16 visits over a 6-month period f A comprehensive solution for the care and monitoring of chronic patients, modelled and tested in patients with CHF that enables the provision of multichannel service and patient tracking through patient monitoring of biometric data (weight, heart rate, and blood pressure), symptoms reporting (seven questions to capture worsening symptoms of the cardiac condition, mainly worsening heart failure, and one question to capture general worsening), generation and management of warning alarms (biometrics out of range), and alerts (information related to the function of the household devices) g Combining five domains of functioning (age over 80 years, cognitive impairment defined as a MMSE score 24 or less, reduced mobility, urinary incontinence, and physical impairment defined as a NYHA functional class III-IV) into six stages of increasing impairment.Frailty score 1 included patients less than 80 years in NYHA class II and without impairments in mobility, continence, or cognitive function.Frailty scores 2-6 included patients with the presence of one, two, three, four, or all of the specific items, respectively h Within the PoC approach, the general practitioner and practice nurse cooperate closely with occupational and physical therapists.If needed, other inpatient and outpatient healthcare professionals, such as a pharmacist or a geriatrician, are involved as well

Groningen Activity Restriction Scale (range total scale 18-78) at 6, 12, and 24 months NR
There was no evidence that the effect of POC approach compared with usual care on Groningen Activity Restriction Scale was different across the frailty spectrum (p-forinteraction > 0.05) does not differ across frailty levels for two specific interventions (p-for-interaction=NR): adjuvanted recombinant zoster vaccine compared with placebo 32 and 23-valent polysaccharide vaccine compared with 23-valent polysaccharide vaccine with 7-valent pneumococcal conjugate vaccine. 33,34 i-interleukin 1 Monoclonal Antibody Canakinumab had a significant treatment effect on incident major adverse cardiovascular events compared with placebo, but the effect was not different across frailty levels (p-for-interaction=NR). 13 Non-pharmacological Intervention.Six trials assessed whether the effect of following treatment was different by frailty levels: diabetes management, 11 physical activity and exercise, 12,35,37,38 and other.36 Diabetes Management In a trial of 5145 adults with type 2 diabetes and overweight or obesity conducted over 118 months in the US, 11 intensive lifestyle intervention was associated with lower cardiovascular events compared with diabetes support and education in patients with frailty index <0.178,not in patients with frailty index ≥0.178(p-for-interaction = 0.01).
Physical Activity and Exercise Increasing frailty index was associated with greater reductions in pain score (p-for-interaction = 0.02) and pain interference (p-for-interaction = 0.01) with chair yoga compared with health education. 35Aerobic exercise training was associated with lower rates of composite of all-cause hospitalization or all-cause mortality compared with usual care in patients with frailty index >0.21,but not in patients with frailty index ≤0.21(p-for-interaction=NR). 37 Similarly, physical activity was associated with lower rates of major mobility disability compared with health education in patients with frailty index ≥0.15,but not in patients with frailty index <0.15(p-for-interaction=NR) 38 .There was significant treatment effect of intensive exercise compared with usual care on mortality, but the effect was not different across the frailty spectrum (p-for-interaction=NR). 12 Other In a trial conducted in the Netherlands, 36 involving 3092 adults aged 60 and older over a 12-month follow-up, there was evidence to suggest that the effect of frailty screening alone or when combined with a nurse-led care program compared with usual care was beneficial on the modified Katz-15 function score, but the effect was different across the frailty spectrum (p-for-interaction=NR).
Multicomponent Intervention.We identified three trials that assessed the effects of multicomponent interventions among frailty subgroups.
Geriatric Assessment In a trial of 541 patients with incurable cancer and impairment in one or more geriatric assessment domains, 39 the effect of geriatric assessment compared with usual care on conversations (p-for-interaction = 0.611), concerns acknowledged (p-for-interaction = 0.740), and concerns addressed (p-for-interaction = 0.940) was beneficial, but the effect was similar across the frailty spectrum.
Pharmacist-Led Deprescribing Intervention In a trial of 363 older adults living in the community conducted over 6 months in the US, the effect of pharmacist-led deprescribing intervention compared with usual care on changes in drug burden index 0.5 or more, anticholinergic cognitive burden was beneficial, but the effect was not different across the frailty spectrum (p-for-interaction=NR). 40,41 Evaluation of Treatment Effect by Frailty Phenotype Pharmacology Intervention.Four trials assessed whether the effect of the following treatments was different by frailty levels: antiplatelet medications, 42 anticoagulants, 43 osteoporosis medications, 16 and androgen medications. 15tiplatelet Medications In a trial 42 conducted across 52 countries with 9326 adults aged 65 or older suffering from unstable angina, prasugrel was associated with higher rates of the composite outcome of cardiovascular death, myocardial infarction, or stroke compared to clopidogrel in patients with one to two components of the frailty phenotype, but not in patients with zero or three to five components of frailty phenotype (p-for-interaction = 0.032).
Other The remaining three interventions (including anticoagulants, 43 osteoporosis medications, 16 and androgen medications) 15 were beneficial, but all concluded that intervention effect was not different across the frailty spectrum.Specifically, the effect of edoxaban, strontium, or testosterone gel compared with placebo on stroke or systemic embolism (p-for-interaction = 0.55), all-cause death (p-for-interaction = 0.06), net clinical composite outcome (p-for-interaction = 0.42), vertebral fracture (p-for-interaction = 0.11), and isometric knee extension peak torque (p-for-interaction = 0.68) was not different across the frailty spectrum.
Non-pharmacological Intervention.The included 10 trials assessed whether the effects of the following treatments were different by frailty levels: diabetes management 44 and physical activity and exercise. 14,37,45,46,47,49,50,51,52 Diabeteanagement In a trial conducted 44 by Rodriguez-Manas in 2019 across 7 European countries, involving 964 adults aged over 70 with type 2 diabetes mellitus and functional impairment, multimodal intervention was compared with usual care over a 12-month follow-up.The results indicated diabetes management was beneficial, but no evidence of a differential effect of the multimodal intervention on Short Physical Performance Battery (SPPB) score across the frailty spectrum (p-for-interaction = 0.49).
Physical Activity and Exercise There was significant effect of functional walking compared with usual pattern of activities on persistent mobility disability, 46 physical rehabilitation intervention compared with attention control on SPPB score. 45,46The majority of the trials revealed no significant differences across the frailty spectrum concerning the effects of various physical interventions.These interventions, when compared to controls or usual care, consistently improved outcomes such as physical performance, 37,47 Mini-BESTest score, 14 functional gait assessment, 14 SPPB score, 47 days at home, 49 dominant handgrip strength, 50 major mobility disability, 51 persistent mobility disability 51 , skeletal muscle mass index, 52 insulin-like growth factor, 52 dehydroepiandrosterone sulfate, 52 and 25-hydroxy vitamin D 52 (p-for-interaction = NR or >0.05).Exceptions to this pattern were identified in two areas: (1) functional walking interventions were associated with higher risks of falls in patients with three to five components of the frailty phenotype compared to usual activities (p-for-interaction = 0.002); 46 (2) physical rehabilitation intervention was associated with greater improvement in SPPB score in patients with three to five components of the frailty phenotype (p-for-interaction = 0.03). 37lticomponent Intervention.We identified three trials that assessed the effects of multicomponent interventions among frailty subgroups.
Multifactorial, Interdisciplinary Intervention Two articles 53,54 from the same RCT illustrated varying effects of multifactorial, interdisciplinary interventions (details shown in the footnote of Table 2) compared to usual care on patients with three to five components of the frailty phenotype, but not in patients with zero to two components of the frailty phenotype.A significant association between multidomain intervention and usual care was observed with greater Life Space Assessment score in patients with three components of the frailty phenotype at 3 months (p-for-interaction = 0.03), but not in patients with four to five components of the frailty phenotype and this effect attenuated by 12 months (p-for-interaction = 0.4).There was an increase in gait speed in patients with four to five components of the frailty phenotype at 12 months, but not in patients with three components of the frailty phenotype (p-for-interaction = 0.03).No significant differences were found in the effects of the intervention on SPPB and Physiological Profile Assessment across the frailty spectrum (p-for-interaction=NR).

Multidomain Intervention and Omega-3 Polyunsaturated Fatty
Acids (n3 PUFA) In the trial 20 conducted by Tabue-teguo in 2018 in France, involving three different interventions (multicomponent + n3 PUFA vs multicomponent alone vs n3 PUFA alone, details shown in the footnote of Table 2) on a population of 1680 older adults over a 3-year follow-up, partial measures of cognitive function (Trail-Making Test A and B) showed significant improvement.However, no significant differences in cognitive function across the frailty spectrum were found (p-for-interaction > 0.05).

Evaluation of Treatment Effect by Other Frailty Tools
Of the included trials, 20 trials measured frailty using other frailty assessment tools, for example, INTERMED for the Elderly Self-Assessment (INTERMED-E-SA), 55,56 Groningen Frailty Indicator (GFI), 36,[55][56][57] frailty-associated conditions, 18 modified frailty score, 19 two tests of physical abilities, 58 simplified Eastern Cooperate Oncology Group (ECOG)-based frailty assessment, 59,60 International Myeloma Working Group geriatric score, 17,61,62 and the Study of Osteoporotic Fractures frailty index. 63armacology Intervention.Seven trials assessed whether the effect of the following treatments were different by frailty levels: anti-neoplastic therapy 17,[59][60][61][62]64 and vaccinations. 18 nti-neoplastic Therapy for Multiple Myeloma Six trials assessed the effects of various treatments on progressionfree survival (PFS) and overall survival (OS) in patients with different frailty levels.Four trials found that their interventions, including melphalan-prednisone-lenalidomide, 59 ixazomib, 62 and lenalidomide and dexamethasone 57 were potentially associated with more prolonged PFS and OS in fit patients or patients with lower levels of frailty. Twrials found significant effect of lenalidomide compared to placebo, 17 or daratumumab plus lenalidomide/dexamethasone compared with lenalidomide/dexamethasone 58 on PFS 58 or OS, 17 but the effect was not different across the frailty spectrum (measured by International Myeloma Working Group geriatric score, simplified ECOG-based frailty assessment, or frailty assessment based on four components, separately) (p-for-interaction=NR in all trials).
Vaccinations There was significant treatment effect of highdose inactivated influenza vaccine compared with a standard-dose vaccine on laboratory-confirmed influenza, but the treatment effect was not different across the frailty spectrum measured by frailty-associated conditions (details shown in the footnotes of Table 3) (p-for-interaction = 0.838). 18 Non-pharmacological Intervention.The included 11 trials assessed whether the effects of the following treatment were different by frailty levels: radiation therapy, 65 surgical procedures, 66 physical activity and exercise, 63,67,68 psychosocial intervention, 69 and others. 55,56,58,70,71 Radion Therapy There was no significant treatment effect of 1-week course radiation therapy compared with a 3-week course on OS, 65 and the effect was not different across the frailty spectrum measured by the Karnofsky Performance Status (p-for-interaction=NR).
Surgical Procedures There was no significant treatment effect of anterior minimally invasive hemiarthroplasty compared with lateral Hardinge hemiarthroplasty on timed up and go duration, 66 and the impact was not different across the frailty spectrum measured by the Functional Independence Measure, Charlson Index, and Medication score (p-for-interaction=NR).
Physical Activity and Exercise A higher dose exercise program led to a greater improvement in health-related quality of life compared to no exercise in patients meeting at least two of three frailty assessments (including Combined Fried frailty phenotype, modified Physical Performance Test, and 7-point Clinical Frailty Scale) (p-for-interaction = 0.037). 67Additionally, physical activity was associated with higher total cardiovascular event rates compared to health education in patients with an SPPB score of 8 to 9, but not in those with an SPPB score <8 (p-for-interaction = 0.006). 68However, there was no evidence that the effect of physical activity compared with health education on 400-m gait-speed, 4-m gait-speed was different across the frailty spectrum (p-for-interaction=NR). 63 Psychosocial Intervention One psychosocial intervention was associated with improvement in instrumental activities of daily living (p-for-interaction < 0.01), potential enhancement in physical performance (p-for-interaction = 0.02), and a reduction in mortality (p-for-interaction = 0.01) in patients with a summary frailty index ≤3, but not in patients with a summary frailty index 4 to 5 (details shown in the footnotes of Table 3).However, these effects were not observed in patients with a frailty index of 4-5, indicating that the benefits of psychosocial intervention may be specific to lower levels of frailty. 69hers No evidence that the effects of various interventions (for HF, stroke, physical frailty, or others), such as home telemedicine, 71 home-based physiotherapy, 58 and embrace program (details shown in the footnotes of Table 3), 55,56 55,56 Multicomponent Intervention.We identified two trials 19,57 that assessed the effect of multicomponent interventions among frailty subgroups.
Hospital-Based Disease Management Programs (DMP) In one RCT 19 , 173 older patients with HF were randomly assigned to DMP or the usual care group.During a 2-year follow-up, a statistically significant effect of DMP compared to usual care was found in lowering all-cause admissions in patients with a higher modified frailty score, but not in patients with a lower frailty score (p-for-interaction = 0.018).However, the effect on death and/or HF admissions was not different across the frailty spectrum (p-for-interaction = 0.208).
Prevention of Care (POC) Approach Metzelthin 57 compared the POC approach with usual care in community dwelling frail older people and found no treatment effect on the Groningen Activity Restriction Scale, and the effect was not different across frailty subgroups measured by GFI (p-forinteraction > 0.05).

DISCUSSION
We found several RCTs that examined whether the effects of pharmacological, non-pharmacological, and multicomponent interventions varied by frailty levels as measured using a deficit accumulation frailty index, frailty phenotype, or other frailty tools.For most pharmacological interventions including flu vaccines, the effect was consistent across frailty levels with a few exceptions: some interventions (e.g., edoxaban, 21 prasugrel, 42 chemotherapy for multiple myeloma 59,61,62 ) were more effective or safer in robust patients, while the benefit of sacubitril/valsartan 28 was greater in frail patients.Intensive lifestyle changes 11 and exercise interventions 14,35,44,45,47,[49][50][51][52]67 seem to benefit frail older adults as much as or more than non-frail older adults. The ffect of complex, multicomponent interventions depended on the type of the intervention.A multicomponent intervention delivered by an interdisciplinary team of physiotherapists, a dietician, a geriatrician, a rehabilitation physician, and a nurse resulted in a greater improvement in life space and gait speed among patients with frailty.53,54 Similarly, a DMP for HF patients led to a greater reduction in all-cause admissions in frail patients.19 In contrast, frailty screening in routine primary care settings 36,57 or a multicomponent program including n3 PUFA supplementation in patients with memory complaints 20 failed to demonstrate the benefit compared to usual care, regardless of frailty levels.Psychosocial support was more effective in functional recovery after stroke in less frail patients.69 The evaluation of treatment effect heterogeneity by patients' frailty levels offers great potential to allow healthcare providers to optimize medical interventions in older adults.In scenarios where frailty significantly modifies treatment effects, healthcare providers can individualize the delivery of medical interventions that maximizes the benefit and minimizes the risks based on patients' frailty levels.A common misconception in practice is that having frailty is conflated with decreased treatment efficacy, which may lead to under-treatment of frail patients as demonstrated by the delayed uptake of newly approved medications.[72][73][74] By showing that treatment benefits do not vary by frailty status, under-treatment of frail patients can be avoided.Likewise, over-treatment can be minimized for frail patients near the end-of-life who may not gain benefit from treatments.
In applying the findings of our review to fully realize the potential of frailty-guided clinical management, there are important caveats to consider. 8First, the adoption of a standardized, validated frailty assessment tool is essential.Frailty assessment tools used in RCTs were often created post hoc using the items only available in the trial settings or were modified from the original tool, resulting in measurement error (i.e., misclassifying patients' frailty levels).It also limits the applicability of the study findings.Adoption of a brief, standardized, validated frailty screening is imperative to translate the frailty-specific effects from RCTs into clinical practice.The Clinical Frailty Scale is a clinical judgment-based assessment that can be completed within 3 min and has been validated in various clinical settings, including primary care, emergency department, inpatient, and preoperative settings. 75,76 econd, post hoc analysis of RCTs should be considered hypothesis generating until further independent confirmation.Secondary analysis of RCTs that is not pre-specified is subject to confounding, type I error (false positive findings from multiple testing), and type II error (false negative findings due to inadequate power).Future adequately powered RCTs are needed to test for heterogeneity of treatment effects by the spectrum of frailty.Third, enrolling more older adults with frailty in RCTs is needed.Given the challenges in recruiting patients with frailty and assessing frailty in RCTs, pragmatic clinical trial or hybrid effectiveness and implementation study design in routine care settings should be considered.Lastly, the mechanisms by which frailty influences response to treatment are unknown.Further research is warranted to elucidate the biological mechanisms.Moreover, it remains uncertain whether improving frailty prior to treatment can change response to treatments by enhancing the benefits and mitigating the harms associated with treatments.

Limitations
Our review has a few limitations.Despite our effort to identify RCTs that assessed the frailty subgroup-specific treatment effect in multiple electronic databases, we did not include publications in non-English literature and the studies that did not report frailty-specific effect estimates, raising the possibility of publication bias.The heterogeneity of frailty assessment tools, type of interventions, and outcomes precluded meta-analysis.

CONCLUSION
This systematic review of post hoc analysis of RCTs suggests that the effectiveness and safety of certain pharmacological interventions can vary according to patients' frailty levels and that the lifestyle or exercise interventions may benefit patients with frailty as much as or more than those without frailty.Although this preliminary evidence supports the potential utility of frailty assessment before treatment decisions, it also reveals the current challenges in delivering frailty-guided clinical care based on the post hoc analyses of RCTs.We call for further investigation into frailty-specific treatment effects by adopting a validated frailty assessment in RCTs of medical interventions involving older adults.

Figure 1
Figure 1 Evidence search and selection.

Modified 5 )
Fried frailty criteria (range, 0-5) A: frailty phenotype (1 to 2) B: frailty phenotype(3 to 5) SPPB score A: effect size 0.8 (− 0.1, 1.6) B: effect size 2.1 (1.3, 2.9) Physical rehabilitation intervention was associated with greater improvement in SPPB score compared with attention control in patients with 3 or more frailty phenotypes, but not in patients with 1-2 components of the frailty phenotype (p-for-interaction = 0.03) Days at home A: IRR 1.03 (0.96, 1.11) B: IRR 1.04 (0.96, 1.12)There was no evidence that the effect of physical exercise compared with usual care on days at home was different across the frailty spectrum (p-for-interaction = NR)Intervention: resistance-type exercise training vs no exercise training Population: 127 adults aged 65 years or older with pre-frail or frail Follow-up: 24 weeks Reference: Tieland, 2015, Netherland 50 Fried frailty criteria (range, 0-5) A: Frailty phenotype (1 to 2) B: Frailty phenotype (3 to 5) Dominant handgrip strength NR There was no evidence that the effect of resistance-type exercise training compared with no exercise training on dominant handgrip strength was different across the frailty spectrum (p-for-interaction > 0.05) Intervention: moderate-intensity physical activity vs health education Population: 1635 adults aged 70-89 years with functional limitations Follow-up: 2 years Reference: Trombetti, 2018, US 51Study of Osteoporotic FracturesIndex (range, 0-3) A: frailty phenotype (0 to < 2) B: frailty phenotype (2 to 3) MMD ‡ A vs B: HR 0.92 vs HR 0.96 PBD ‡ A vs B: HR 0.92 vs HR 0.94 There was no evidence that the effect of moderate-intensive physical activity compared with health education on MMD (p-for-interaction = 0.91) and PBD (p-forinteraction = 0.64) was different across the frailty spectrumIntervention: walking and nutrition vs walking vs control Population:

Table for
HR, hazard ratio; NR, not reported; MPR, melphalan-prednisone-lenalidomide; CPR, cyclophosphamide-prednisone-lenalidomide; Rd, lenalidomide and dexamethasone; Rd18, lenalidomide and dexamethasone for 18 cycles; MPT, melphalan + prednisone + thalidomide; ECOG, Eastern Cooperate Oncology Group; PFS, progression-free survival; OS, overall survival; RT, radiation therapy; HR-QOL, health-related quality of life; SPPB, Short Physical Performance Battery; INTERMED-E-SA, INTERMED for the Elderly Self-Assessment; GFI, Groningen Frailty Indicator; EQ-5D-3L, EuroQol-5D three-level version; PAIEC, Patient Assessment of Chronic Illness Care; HF, heart failure; ADL, activity of daily living; AMTS, a mental test score; POC, prevention of care; DMP, hospital-based disease management programs Notes: Data were extracted from supplemental files † Data were extracted by GetData Graph Digitizer software a Patients' frailty status was classified as fit, unfit, or frail on the basis of four components: age, the Katz Index of Independence in Activities of Daily Living, the Lawton Instrumental Activities of Daily Living Scale, and the Charlson Comorbidity Index Scoring System b Vision loss, hearing loss, impaired mobility, difficulty toileting, difficulty bathing, difficulty dressing, difficulty grooming, difficulty going out, skin problems, resting tremor, changes in sleep, urinary complaints, gastrointestinal problems, and hypertension * c A summary frailty index combing measures of physical and mental functioning: depressive symptoms, Mini-Mental State Exam, NIH Stroke Scale, rehabilitation days, and number of pre-existing conditions d Embrace (in Dutch: SamenOud [ageing together]