This study’s protocol was prospectively registered with the PROSPERO (International Prospective Register of Systematic Reviews; CRD42024528132; dated April 3, 2024). The reporting process adhered to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines throughout the review development. Two deviations from the PROSPERO-registered protocol should be noted: (1) expanded synthesis framework: the prespecified quantitative meta-analysis was supplemented with a thematic qualitative analysis to enhance interpretation of frailty determinants, and (2) updated search timeframe: the literature search cutoff was extended from April 2024 to May 2025 to capture emerging evidence.

This systematic review was structured according to the PICo (Population, Interest, and Context) framework: Population (P): adult patients (≥18 years) diagnosed with malignant tumors through pathological examination, for whom chemotherapy constitutes a primary treatment modality; Interest (I): the occurrence of frailty and its predictive factors within this population; and Context (Co): clinical settings where chemotherapy is administered to patients with cancer.

Studies were selected according to the following criteria: (1) study design: observational studies (cohort, cross-sectional, or case-control studies) published in English or Chinese; (2) population: adults (≥18 years) with histologically confirmed malignant neoplasms at any stage and receiving or having received chemotherapeutic regimens; (3) outcome measures: quantified frailty status using validated instruments, including but not limited to the Fried frailty phenotype, Comprehensive Geriatric Assessment, Vulnerable Elder Survey-13 (VES-13), or Fatigue, Resistance, Ambulation, Illnesses, & Loss of Weight (FRAIL) scale [13]; and (4) data requirements: reporting multivariate-adjusted odds ratios with 95% CIs for frailty-associated factors or providing sufficient raw data for effect size calculation. Conversely, the exclusion criteria were (1) nonoriginal research (reviews or editorials), studies with ambiguous case definitions, or inaccessible full-text papers, and (2) publications with overlapping datasets without novel analyses.

Two independent reviewers (MZ and RZ) conducted the methodological evaluations of the included studies using design-specific tools. For cohort and case-control studies, the Newcastle-Ottawa Scale (NOS) [18] was applied to assess 3 critical domains: selection (representativeness of cohorts and exposure ascertainment), comparability (control for confounding variables), and outcome (validity of assessment methods and follow-up completeness). Studies were classified as low (0‐3 points), moderate (4-6), or high (7-9) quality [18]. Cross-sectional studies were evaluated using the Agency for Healthcare Research and Quality (AHRQ) checklist [19], which assigns scores (0‐11) across 11 criteria, including case definition clarity and consecutive participant enrollment. Each criterion was scored 1 (“yes”) or 0 (“no” or “unclear”), with final classifications as low (0-3), moderate (4-7), or high (8-11) quality [19]. Discrepancies in ratings were resolved through iterative team discussions, and unresolved cases were arbitrated by a third reviewer (XD), an evidence synthesis specialist.

Data extraction was conducted in a standardized double-blind manner by 2 investigators (TW and CY) who had completed evidence-based research training. The recorded variables encompassed bibliographic information (first author, publication year, and geographic region), study design characteristics (prospective or retrospective type, sample size, and tumor type), participant demographics (mean age with SD and diagnostic criteria), analytical parameters (frailty assessment tools and covariates adjusted in multivariate models), and outcome metrics represented by multivariate-adjusted odds ratios with 95% CIs. Frailty-associated factors were synthesized through cross-study comparative analysis, focusing on concordance patterns observed between independent odds ratio estimates.

Review Manager (version 5.4; Cochrane Collaboration) and R (version 4.4.3) were used to quantify associations between candidate variables and chemotherapy-induced frailty through the combined pooled odds ratios with 95% CIs. Heterogeneity was assessed through the Cochrane Q-test (significance threshold P<.10) and I^² statistics, with I^² ≥50% indicating substantial heterogeneity. Effect sizes were synthesized using fixed-effect models for homogeneous datasets (I^²<50%) or random-effect models when significant heterogeneity existed. To investigate sources of heterogeneity, prespecified subgroup analyses were conducted by stratifying studies based on tumor type and frailty assessment tools. Sensitivity analyses were conducted by sequentially excluding studies with low quality ratings to assess the robustness of the pooled results. Publication bias was assessed using funnel plots for visual inspection when at least 10 studies were available. Publication bias was examined via Egger regression test, with P<.05 denoting potential bias. For predictors unsuitable for a meta-analysis (I^²>50% or incomplete data reporting), a structured qualitative synthesis was performed. Factors reported in 2 or more studies were systematically categorized, with their associations classified by directionality (risk or protective) and strength of evidence.

This systematic analysis incorporated 14 studies (8 cross-sectional [11,21-27], 3 cohort [28-30], 2 repeated cross-sectional [20,31], and 1 mixed-design [32]) involving 3879 patients with cancer. Most studies (9/14, 64.3%) were published between 2020 and 2025. Asian populations predominated (12/14, 85.7%), including 7 (50%) studies from mainland China, 2 (14.3%) from Taiwan, China, and 1 (7.1% each) each from Korea, Thailand, and Turkey. Two (14.3%) cohort studies were from North America.

Lung cancer was the most studied malignancy (5/14, 35.7%). Three cross-sectional studies (3/14, 21.4%) examined mixed malignancies, followed by single investigations of breast, pancreatic, colorectal, gastric, and biliary tract cancers. Sample sizes ranged from 75 to 1020 participants (median 185, IQR 124‐234). The weighted mean age across studies was 64.16 years (SE 9.07), with individual study means ranging from 53.4 to 71.4 years. Heterogeneous tools were used; Fried phenotype (4/14, 28.6%) and FRAIL scale variants (3/14, 21.4%) were the most common types. Frailty prevalence, reported in 11 (78.6%) studies, varied substantially by cancer type and assessment method: lowest in lung cancer cohorts using the Frailty Index based on Laboratory test data (4.9%) and highest in a mixed malignancy study involving the clinical frailty scale (67.3%).

Through the systematic screening of 14 studies, we identified 23 frailty-associated factors appearing in 2 or more studies. These were classified into 4 domains through independent dual-reviewer consensus (with third-party consultation resolving discrepancies). For demographic parameters, the most foundational domain included age, sex, education level, monthly family income, and marital status, with age being the most universally reported factor across studies. The socioeconomic-psychological factor domain integrated 3 psychosocial determinants, such as depression, anxiety, and social support, reflecting the interplay between mental health, resource accessibility, and care logistics. Clinical-treatment characteristics encompassed 13 factors directly related to disease progression and therapeutic interventions, including cancer type, cancer stage, transfer, chemotherapy frequency, surgical history, fall history, comorbidities, hypoproteinemia, Eastern Cooperative Oncology Group Performance Status (ECOG-PS), nutritional status, BMI, somatic symptom clusters, and emotional symptom clusters. Biological markers, comprising 2 objective laboratory parameters, highlighted physiological vulnerabilities through hemoglobin levels and white blood cell count. Notably, age (reported in 6 studies), depression (5 studies), and cancer stage (4 studies) emerged as the most frequently cited cross-cutting factors. The basic features of the included studies have been presented (Table 1).

Methodological quality was rigorously evaluated using established tools: the 11-item AHRQ checklist for cross-sectional studies (Table 2) and the 8-item NOS for cohort studies (Table 3). The 11 cross-sectional studies achieved a mean AHRQ score of 8.8 (range: 7‐11), with 72.7% (8/11) scoring ≥8 points (high quality). Common methodological limitations included the inadequate description of missing data (7/11, 63.6%) and the absence of follow-up–related data (7/11, 63.6%). Three (27.3%) studies demonstrated a moderate risk of bias owing to the insufficient control of confounding variables. The 3 cohort studies attained high quality on NOS (mean score 8.0, range 7‐9) and demonstrated adequate outcome ascertainment and follow-up completeness. No studies were excluded based on quality thresholds, as per PRISMA guidelines for observational research synthesis.

We performed sensitivity analyses using a two-step approach: (1) the sequential exclusion of individual studies when subgroup analyses included 3 or more studies, and (2) the targeted removal of studies contributing to substantial heterogeneity (I^²>50%). In assessing the robustness of our findings, sensitivity analyses revealed critical insights into heterogeneity patterns (Multimedia Appendix 2). The exclusion of the study by Wang et al [29], which reported the lowest frailty prevalence (4.9%, 95% CI 3.7%‐6.4%), preserved excessive residual heterogeneity (I^²=96%, P<.001), indicating alternative variability sources such as methodological discrepancies in frailty assessment tools (5 distinct instruments used), clinical diversity across cancer types, and geocultural variations in frailty conceptualization contrasting Asian and Western cohorts.

Subsequent domain-specific analyses demonstrated differential heterogeneity profiles. The removal of the study by Limpawattana et al [26] from age-related factor analyses (5 studies) substantially reduced heterogeneity to moderate thresholds (I^²=44%, P=.15) while preserving effect stability (pre-exclusion odds ratio 1.75, 95% CI 1.24‐2.47 vs postexclusion odds ratio 1.95, 95% CI 1.52‐2.49). For depressive symptom studies (4 articles), sequential exclusion procedures yielded minimal heterogeneity fluctuations (ΔI^²<8%) with consistent effect magnitudes. Physical symptom cluster analyses (3 studies) maintained stable effect estimates despite persistent high heterogeneity (I^²>65%). Notably, the exclusion of the study by Xia [21] in emotional symptom cluster investigations (3 studies) effectively resolved heterogeneity (I^²=21%, P=.26) without directional change in effects (pre-exclusion odds ratio 1.27, 95% CI 1.04‐1.55 vs postexclusion odds ratio 1.37, 95% CI 1.20‐1.57).

Potential publication bias was evaluated using funnel plots for outcomes with 10 or more studies. In the meta-analysis of frailty prevalence, which included 11 studies (Figure 3), no significant publication bias was detected using the Egger test (t₉=−0.96,, P=.36). However, the presence of substantial heterogeneity necessitates a cautious interpretation of the pooled results.

Owing to the few studies included in the meta-analysis (applied to outcomes with 3 or more studies, including cancer stage, age, and depression), Egger test was used to assess potential publication bias. While no significant publication bias was detected for the cancer stage (P>.05), statistically significant asymmetry was observed in the analyses of age (P=.04) and depression (P=.02), suggesting publication bias. Notably, the Egger test has reduced statistical power when fewer than 10 studies are included and may overestimate effect sizes or be confounded by heterogeneity. Therefore, these results were interpreted with caution, supplemented by the visual inspection of funnel plots and comprehensive sensitivity analyses.

Our systematic review revealed 7 factors associated with frailty in chemotherapy-treated patients with cancers who were consistently reported in at least 2 studies: cancer type, surgical history, BMI, social support, marital status, anxiety, and history of falls. However, quantitative synthesis through odds ratio pooling was precluded owing to insufficient data reporting across these studies. For 9 additional factors demonstrating substantial heterogeneity (I²>50%), a descriptive synthesis was conducted.

Sociodemographic factors included (1) advanced age: 5 studies [21,22,26,27,32] consistently associated old age with increased frailty risk; (2) female sex: 2 studies [25,32] reported higher frailty prevalence in females than in males, potentially linked to lean mass reduction and postmenopausal endocrine shifts; (3) marital status: 2 studies [11,32] suggested unmarried status as a risk factor; however, the effect estimates were inconsistent; (4) education level: low education was associated with frailty in 2 studies [21,25]; and (5) monthly household income: 2 studies [21,25] indicated low income as a risk factor; however, the pooled results were nonsignificant.

Disease or treatment-related factors included (1) cancer type: frailty prevalence varied markedly by malignancy: 37.2% in breast cancer [28], 45.4% in lung cancer [24], and 25.4% in pancreatic cancer [30]; (2) surgical history: 2 studies [21,29] linked prior surgery to frailty progression; however, the mechanisms remain unclear; and (3) history of falls: 2 studies [25,27] implicated falls as a precursor and consequence of frailty.

Psychosocial and socioeconomic factors included (1) social support: 2 studies [21,31] highlighted inadequate social support as a contributor to frailty; (2) depression: 5 studies [11,21,23,27,32] indicated that patients with high depression severity exhibited an increased susceptibility to frailty, demonstrating a bidirectional association between frailty and depression; (3) anxiety: 2 studies [21,32] associated anxiety with frailty severity, particularly in advanced disease stages; and (4) emotional symptom cluster: 3 studies [20,21,27] demonstrated emotional distress as a frailty accelerator.

Symptom-functional domains included (1) ECOG-PS: 2 studies [22,30] reported that patients with ECOG-PS scores ≥2 exhibited heightened susceptibility to frailty, demonstrating that impaired mobility contributes to frailty progression through deconditioning cycles; (2) nutritional status: 3 investigations [24,27,31] identified that individuals at risk of malnutrition had a high likelihood of developing frailty; (3) BMI: 2 studies [26,31] recognized abnormal BMI values (underweight and overweight statuses) as independent risk factors for frailty; however, one study failed to report the data on the significant association between underweight or overweight status and the progression of frailty; and (4) somatic symptom cluster: 3 studies [20,21,27] indicated that multisystem symptom burden is a critical correlate of frailty.

High-confidence evidence (3 or more consistent studies) supported age, depression, nutritional status, emotional symptom cluster, and somatic symptom cluster as frailty predictors. Other factors, such as marital status and anxiety, required further investigation owing to limited or conflicting data.

This meta-analysis revealed a 35% (95% CI 22%‐50%) pooled prevalence of frailty among patients with cancer undergoing chemotherapy. Notably, substantial heterogeneity was observed (I^²=98.2%, P<.001) in frailty prevalence among chemotherapy patients, which persisted even after subgrouping by cancer type and assessment tools. Fixed-effects modeling identified 7 robust clinical predictors: advanced cancer stage, frequent chemotherapy cycles, care transitions, anemia, white blood cell count, comorbidity burden, and hypoproteinemia, all demonstrating low heterogeneity (I^²=0%‐46%). Qualitative synthesis demonstrated strong evidentiary support for age, depression, nutritional status, emotional symptom clusters, and somatic symptom clusters as frailty determinants, with multiple studies consistently reinforcing these associations across heterogeneous populations. Notably, while quantitative synthesis confirmed biological and treatment-related predictors, qualitative findings highlighted understudied socioeconomic (marital status, education, and income) and symptom cluster determinants. The persistent residual heterogeneity suggests unmeasured contextual factors, potentially including treatment intensity variations and health care access disparities, may mediate frailty trajectories in this population.

This study has several limitations that should be acknowledged. First, regarding the included studies, (1) the majority were observational investigations with small sample sizes, and multicenter large-scale cohort studies remain scarce; and (2) significant methodological diversity existed across studies, including heterogeneous case sources, inconsistent application of frailty assessment tools, and the absence of standardized diagnostic criteria for frailty, which introduced clinical and statistical heterogeneity.

Second, limitations inherent to our systematic review methods include: (1) language restrictions in our search strategy, which excluded studies not published in Chinese or English, may have omitted relevant evidence from regions where these languages are not primary; (2) there are inherent constraints in the use of Egger test to detect publication bias with limited study numbers (n<10), where false-positive rates increase significantly; (3) despite comprehensive database searches, the possible omission of gray literature (eg, conference abstracts and unpublished datasets) might have skewed effect estimates toward statistically significant findings; and (4) our meta-analytic approach could not fully disentangle confounding factors due to insufficient primary data, limiting causal inference.

These constraints highlight the need for future research using standardized frailty criteria, prospective multinational cohorts, and individual-patient-data meta-analyses to address these gaps.

The pooled prevalence of frailty observed in chemotherapy-treated patients with cancers exceeded the 34% prevalence reported in general cancer populations [7] and the 14.2% prevalence documented in noncancer patients admitted to intensive care after surgery [33]. Systematic evidence synthesis revealed substantial heterogeneity in frailty prevalence among older patients with cancer, with a review of 20 studies (n=2916) reporting a median prevalence of 43% (range 4%‐86%) [34]. This discrepancy may reflect our study’s inclusion of chemotherapy patients without age restrictions. This accelerated frailty progression aligns with evidence showing the heightened vulnerability of older patients with cancers to adverse health outcomes compared with that of younger counterparts [35]. Measurement variability across patient populations and assessment tools increased this heterogeneity [35].

Notably, longitudinal data from Gilmore et al [36] demonstrated the profound impact of chemotherapy on frailty trajectories in patients with breast cancer aged ≥50 years. Using modified Fried criteria, the proportion of participants meeting frailty thresholds (Fried score ≥3) increased from 13% prechemotherapy to 46% posttreatment. The pathophysiology of chemotherapy-associated frailty involves complex interactions between aging mechanisms and cytotoxic stress. Cellular senescence and stem cell depletion during typical aging processes reduce physiological reserve capacity [1,35], whereas chemotherapy-induced damage exacerbates these deficits through impaired tissue repair mechanisms [37,38]. This dual burden manifests clinically as accelerated functional decline, weight loss, fatigue, and impaired activities of daily living during treatment [28].

From a clinical outcome perspective, frailty status is an independent predictor for multiple adverse treatment effects. It is associated with reduced chemotherapy response rates [10] and increased risk of treatment failure even after successful tumor burden reduction [9]. Furthermore, frail patients exhibit elevated mortality risks following surgical interventions and chemotherapy regimens [7,29,39]. Notably, frailty correlates with multidimensional quality of life impairments, manifesting as physical functional decline, psychological distress, and social participation limitations [7,12]. These interconnected outcomes underscore the role of frailty as a critical prognostic indicator across the cancer care continuum.

Moreover, the dynamic nature of frailty permits early intervention. Timely identification through validated assessment tools enables the implementation of targeted prehabilitation programs. Current evidence suggests that these interventions prevent, reverse, or delay frailty progression, reducing adverse outcomes and the associated socioeconomic burdens in populations with cancers [7]. This synthesis highlights the critical need for standardized frailty assessments in chemotherapy decision-making and underscores the modifiable nature of frailty when detected in preclinical stages.

Our findings align with Fried frailty phenotype [3] and the biopsychosocial model of frailty and emphasize frailty as a multidimensional syndrome requiring integrated assessment frameworks arising from interconnected biological, clinical, and socioeconomic mechanisms. The synergy between tumor biology (eg, metastatic inflammation) and treatment toxicity (eg, platinum-induced myelosuppression) creates a “double-hit” mechanism in older patients, where chemotherapy amplifies mitochondrial dysfunction and epigenetic aging, aligning with models of accelerated aging in cancer survivors. This biological-clinical interplay is compounded by socioeconomic mediation of risk. Severe depression elevates frailty risk by 10.7-fold through neuroendocrine-immune crosstalk; however, intervention effectiveness may be limited by disparities in care access, underscoring the need for policy reforms to address therapeutic inequities. Similarly, functional-nutritional vicious cycles driven by Eastern Cooperative Oncology Group ≥2 and hypoalbuminemia perpetuate frailty progression—mobility restriction exacerbates sarcopenia, while protein catabolism impairs recovery, highlighting the urgency for prehabilitation protocols integrating physical therapy and nutritional optimization. These findings highlight the significance of stage-adapted frailty screening tools that incorporate biomarkers (albumin or hemoglobin) and psychosocial profiles in clinical settings. For high-risk patients, comorbidity-centered dose modifications and early palliative care referrals may mitigate compounding morbidity, bridging the gap between mechanistic understanding and personalized geriatric oncology care.

With the aging global population, cancer and frailty have become critical health challenges for older adults. Frailty in older patients with cancer is associated with numerous adverse clinical outcomes, including functional decline, falls, impaired mobility, disability, increased hospitalization rates, and mortality [22]. This structured taxonomy enables clinicians to prioritize multidimensional assessments while guiding researchers in developing targeted frailty intervention frameworks. Early identification of individuals at high frailty risk, particularly through age-based screening, and the implementation of tailored interventions are crucial for reducing adverse outcomes. The interaction between biological markers and psychosocial factors particularly warrants further investigation, given their co-occurrence in multiple studies.

Prevention and management strategies for frailty during chemotherapy should also consider individual sex differences, socioeconomic status, and overall health. The disparity in frailty prevalence observed among patients with different malignancies during chemotherapy possibly reflects underlying tumor-specific pathophysiological mechanisms and regimen-related toxicities. For example, the elevated frailty burden in lung cancer may correlate with frequent late-stage diagnoses and high chemotherapy-induced immunosuppression, whereas metabolic derangements and rapid functional decline in pancreatic cancer may accelerate frailty trajectories. Such cancer-specific patterns emphasize the need for tailored screening protocols aligned with tumor biology and treatment regimens.

This systematic review and meta-analysis revealed a substantial prevalence of frailty among chemotherapy-treated populations with cancers. Therefore, we suggest the systematic implementation of validated frailty screening protocols in routine clinical assessments before chemotherapy initiation. Preliminary evidence of frailty risk factors for patients with cancer during the chemotherapy stage was obtained through systematic evaluation. Based on these risk factors, early prevention and management strategies should be implemented to mitigate the impacts of frailty on chemotherapy and improve treatment outcomes. Future in-depth multicenter, large-sample studies are required to explore the role of depression in frailty during chemotherapy and comprehensively evaluate the various factors influencing frailty in patients with cancers undergoing chemotherapy, sufficiently guiding clinical practice.