Frailty measurement, prevalence, incidence, and clinical implications in people with diabetes: a systematic review and study-level meta-analysis

Abstract

Background: Frailty, a state of increased vulnerability to adverse health outcomes, is important in diabetes management. We aimed to quantify the prevalence of frailty in people with diabetes, and to summarise the association between frailty and generic outcomes (eg, mortality) and diabetes-specific outcomes (eg, hypoglycaemia).

Methods: In this systematic review and study-level meta-analysis, we searched MEDLINE, Embase, and Web of Science for observational studies published between Jan 1, 2001 (the year of the original publication of the Fried frailty phenotype), to Nov 26, 2019. We included studies that assessed and quantified frailty in adults with diabetes, aged 18 years and older; and excluded conference abstracts, grey literature, and studies not published in English. Data from eligible studies were extracted using a piloted data extraction form. Our primary outcome was the prevalence of frailty in people with diabetes. Secondary outcomes were incidence of frailty and generic and diabetes-specific outcomes. Data were assessed by random-effects meta-analysis where possible and by narrative synthesis where populations were too heterogeneous to allow meta-analysis. This study is registered with PROSPERO, CRD42020163109.

Findings: Of the 3038 studies we identified, 118 studies using 20 different frailty measures were eligible for inclusion (n=1 375 373). The most commonly used measures of frailty were the frailty phenotype (69 [58%] of 118 studies), frailty (16 [14%]), and FRAIL scale (10 [8%]). Studies were heterogenous in setting (88 studies were community-based, 18 were outpatient-based, ten were inpatient-based, and two were based in residential care facilities), demographics, and inclusion criteria; therefore, we could not do a meta-analysis for the primary outcome and instead summarised prevalence data using a narrative synthesis. Median community frailty prevalence using frailty phenotype was 13% (IQR 9-21). Frailty was consistently associated with mortality in 13 (93%) of 14 studies assessing this outcome (pooled hazard ratio 1·51 [95% CI 1·30-1·76]), with hospital admission in seven (100%) of seven, and with disability in five (100%) of five studies. Frailty was associated with hypoglycaemia events in one study (<1%), microvascular and macrovascular complications in nine (82%) of 11 studies assessing complications, lower quality of life in three (100%) of three studies assessing quality of life, and cognitive impairment in three (100%) of three studies assessing cognitive impairment. 13 (11%) of 118 studies assessed glycated haemoglobin finding no consistent relationship with frailty.

Interpretation: The identification and assessment of frailty should become a routine aspect of diabetes care. The relationship between frailty and glycaemia, and the effect of frailty in specific groups (eg, middle-aged [aged <65 years] people and people in low-income and lower-middle-income countries) needs to be better understood to enable diabetes guidelines to be tailored to individuals with frailty.

Funding: Medical Research Council.

Figure 1

Study selection

Figure 2

Prevalence of frailty by setting and frailty definition, ordered by mean age of study population Full list of references of all the studies mentioned is included in the appendix (p 46)). eFI=electronic frailty index. ACG=adjusted clinical groups. CGA=comprehensive geriatric assessment. VES-13=vulnerable elders survey RAND-36=research and Development Corporation. Kihon=kihon checklist. mPPT-modified physical performance test. Gronigen=Gronigen frailty indicator.

Figure 3

Random-effects meta-analysis of odds of incident frailty associated with diabetes

Figure 4

Random effects meta-analysis of association between frailty and mortality

Figure 5

Harvest plot of association between frailty and generic (A) and diabetes-specific (B) clinical outcomes HbA_1c= glycated haemoglobin.