Calf Circumference and All-Cause Mortality: A Systematic Review and Meta-Analysis Based on Trend Estimation Approaches

Abstract

Objectives

To perform a systematic review and meta-analysis and quantify the associations of total mortality with calf circumference (CC) in adults 18 years and older via combining various analyses based on empirical dichotomic CC, continuous CC, and dose-response CC.

Methods

We conducted a systematic search of relevant studies in PubMed, EMBASE, Cochrane Library, and Web of Science published through April 12, 2022. This systematic review includes longitudinal observational studies reporting the relationships of total mortality with CC. We calculated the pooled relative risk (RR) and 95% confidence interval (CI) of total mortality with CC per 1 cm for each study and combined the values using standard meta-analysis approaches. Newcastle-Ottawa scale (NOS), Grading of Recommendations, Assessment, Development and Evaluations approach (GRADE), and the Instrument for assessing the Credibility of Effect Modification Analyses (ICEMAN) were assessed for meta-analyses.

Results

Our analysis included a total of 37 cohort studies involving 62,736 participants, across which moderate heterogeneity was observed (I2=75.7%, P<0.001), but no publication bias was found. Study quality scores ranged from 6 to 9 (mean 7.7), with only three studies awarded a score of 6 (fair quality). We observed an inverse trend between total death risk and CC per 1 cm increase (RR, 0.95, 95% CI, 0.94–0.96; P<0.001; GRADE quality=high). Only a very slight difference was found among residents of nursing homes (6.9% mortality risk reduction per one cm CC increase), community-dwellers (5.4%), and those living in hospitals (4.8%), respectively (P for meta-regression=0.617). Low credible subgroup difference was found based on the ICEMAN tool.

Conclusions

Calf circumference is a valid anthropometric measure for mortality risk prediction in a community, nursing home, or hospital.

Introduction

As the global population continues to see a rise in its elderly demographic, more and more countries have begun to face a series of challenges inflicted by global aging, including the increased risk of malnutrition, decrease in bone and muscle mass, subsequent impairments in physical, cognitive and immune responses, as well as premature mortality (1, 2). Anthropometry is commonly used to assess nutritional conditions and health statuses in various epidemiological studies and clinical practices due to its convenience and non-invasiveness (3, 4, 5, 6). Such measurements include body mass index (BMI), waist circumstance (WC), waist-to-hip ratio (WHR), and waist-to-height ratio (WHtR), all of which can be used to assess risk of overall obesity or central obesity (7, 8, 9, 10). Two meta-analytic reviews of over 200 individual studies covering more than 30 million community-based populations have reported increased risks of total mortality in both underweight and overweight participants, with the nadir of a U-shaped dose-response curve in the BMI range set at 25 kg/m² (11, 12). Another meta-analytic review of 72 observational studies showed that indices of central fatness (i.e. WC, WHR, and WHtR) were associated with all-cause mortality after BMI was accounted for (13). However, both overall and central obesity may present benefits to reducing the risk of death in the elderly (13, 14, 15).

Excluding indices of overall and central obesity, some extremity circumferences have been used to explain death risk, especially for the elderly and those suffering from chronic diseases, such as mid-upper arm circumference (MUAC), thigh circumference (TC), and calf circumference (CC) (16, 17, 18, 19, 20, 21, 22, 23, 24, 25). In a recent meta-analytic review, larger TC values were associated with a lower death risk, albeit only three individual studies were included in the meta-analysis for TC (13). As an indicator of reduced muscle mass and possible protein energy malnutrition (28), CC was recommended as a better indicator of health status, because the legs contain over half of the body's muscle mass (MM), which is more strongly associated with health than fat mass (FM) (26, 27).

A recent systematic meta-analytic review including 23 studies reported those with low CC having 2.42 times higher mortality risk than those with normal CC, but the meta-analysis only considered studies using varying empirical cut-offs of CC, such as 31, 33 or 34 cm (29). Furthermore, the analysis excluded several studies due to “inappropriate” rulings based on continuous CC and/or respective dose-response CC (18, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44). Hence, we aimed to perform an updated systematic meta-analysis of longitudinal observational studies to evaluate the associations of all-cause mortality with CC in adults aged ≥18 years old by using comprehensive methods of trend estimation from summarized dichotomic and dose-response data (45, 46, 47, 48).

Methods

Literature Searching Strategies

The reporting of this systematic review complied with the PRISMA statement guidelines (Appendix Table S1) (49). Our researchers conducted a literature search to identify longitudinal observational studies assessing the association between calf circumference and total mortality in four electronic databases, including PubMed (National Library of Medicine, National Center for Biotechnology Information, Bethesda, USA), EMBASE (Elsevier, Amsterdam, the Netherlands), Cochrane Library (John Wiley & Sons, Inc., London, United Kingdom), and Web of Science (Clarivate Analytics, Boston, USA) through April 12, 2022 (the complete search strategies for the four databases can be found in Appendix Table S2). Relevant articles utilized various combinations of relevant terms, like ([thigh circumference] OR [calf circumference] OR [extremity circumference] OR [leg circumference] OR [mid-thigh circumference] OR [knee circumference]) AND ([death] OR [mortality] OR [survival]). Duplicate publications were deleted. Each title and/or abstract was checked for relevance. The full text of an article was reviewed only if the abstract indicated that the article reported these associations. Only original studies conducted among human participants were considered for the review. Cross-referencing was employed to complement the study identification process of the systematic review.

Inclusive Criteria

Studies were included in this review if they met the following criteria: 1) adults must be 18 years or older and possess self-care abilities, such as feeding, dressing, and walking independently; 2) longitudinal observational studies, including both retrospective and prospective cohorts; 3) calf circumference had to have been measured as the maximum diameter of the calf spanning the head of the fibula and the lateral malleolus (50); thigh circumference was not considered for this review; 4) total mortality was evaluated during follow-ups on the outcome; 5) hazard ratios (HR) / relative risk (RR) / odds ratio (OR) and their corresponding confidence intervals [CI] were used to evaluate the association of total mortality with CC; 6) studies among patients suffering from debilitating diseases were not excluded, although corresponding sensitivity analyses and subgroup analyses may have been used to determine the credibility of how large an effect debilitating diseases may incur.

Data Extraction

Two researchers (Li X and Lang X) extracted the following data independently from each study in a standardized manner: 1) study characteristics: authors' names, year of publication, study duration, country of origin, number of participants, and deaths observed during the respective follow-up duration; 2) baseline demographics: mean age of participants and male proportions, mean CC and CC categorization, comorbidities, total mortality; 3) effect sizes: HR/RR/OR and their 95% CI for each CC category and/or per unit increase of CC; 4) covariates adjusted for in the analyses. Any initial disagreements were resolved by consensus after additional review of the articles or team discussions.

Quality Assessment

The two researchers, Li X and Lang X, each independently performed the quality assessment of the included studies using the Newcastle-Ottawa scale (NOS). Accounting for eight items across three domains (selection, comparability, and exposure), a study was awarded one star for each item, with the exception of the item related to comparability, which allowed the assignment of two stars. Hence, NOS ranges from 0 to 9, and a score of 7 or greater was considered good quality (51). The grading of recommendations, assessment, development and evaluations (GRADE) approach was used to rate the confidence by which increasing CC correlated with a reduced mortality risk in the broader population. After making the necessary considerations for the study design, risk of bias, imprecision, inconsistency, indirectness, and publication bias, certainty of evidence may be rated as high, moderate, low, or very low (52). Any discrepancies were discussed and resolved in our research team.

Statistical Analyses

All statistical analyses were performed with Stata 15.0. The main effect size of interest consisted the measures of the association between calf circumference and total mortality, including HR, RR, and OR. The reported ORs and HRs were first transferred to RRs before being used for final meta-analyses (47, 48). Depending on the data available, we used median, midpoints or means of various categories for transformation that we saw fit. In two studies, mortality was assessed and stratified via dose-response CC (25, 53). Due to the different categorization of CC across the studies, all results were recalculated for an increase of one centimeter CC, both within studies (taking into account any possible correlations resulting from a common reference category (45, 46)), as well as across studies.

Heterogeneity was evaluated using Cochran's Q test, and significant heterogeneity was defined as P<0.10. The heterogeneity across studies was considered to be moderate or high if the I² statistic was greater than 50%. However, using I² statistics to assess heterogeneity for the pooled estimate can be extremely high even in the presence of modest inconsistencies across the studies, wherein observational studies with extremely large sample sizes may result in very narrow confidence intervals. Additionally, the GRADE approach can also be used to assess heterogeneity (52). The results from both random effects models, using the DerSimonian and Laird method, and fixed effects models based on the Mantel-Haenszel methods were provided. Publication bias was evaluated using Begg and Mazumdar rank correlation tests, with significant publication bias defined as P<0.05.

We also conducted various subgroup to explore potential sources of heterogeneity and the influence of moderator variables on overall effect sizes, such as participant type (community, nursing home, hospital), publication year (≥ 2017 vs. < 2017), follow-up duration (≥ 5 years vs. <5 years), sample size (≥ 500 vs. <500), continents (Asia, America, and Europe), study quality (good vs. fair), age (≥ 65 years vs. <65 years), gender (male, female, both), average CC of targeted participants (<31 cm, ≥ 31 cm, and not available), and body position during CC measurement (standing vs. sitting). Meta-regressions were used to examine the impact of the moderator variables mentioned above on study effect sizes using regression-based techniques. The credibility of any subgroup analysis was also addressed using instrument to assess the credibility of effect modification analyses (ICEMAN) on the eight criteria (54). In order to test the robustness of overall effect sizes, one study was omitted at a time, and sensitivity analyses were performed to evaluate the influence of each individual study on the overall effect sizes. One single study was considered to have influence on the overall effect size if the point estimate of its “omitted” analysis fell outside of the 95% CI of the “combined” analysis.

Results

Study Screening and Selection

A flow diagram of the study search and selection process is presented in Figure 1. We found 827 publications via a systematic literature search, of which 352 records were deleted as duplicates. The remaining 475 titles/abstracts were screened for eligibility and 378 publications were excluded, including 239 irrelevant literature articles, 97 conference abstracts, 24 studies with participants aged <18 years old, 13 unoriginal articles, and 5 animal experiments. 97 studies appeared to be potentially relevant for inclusion, but 60 were excluded during the full-text evaluation due to the following reasons: 35 contained no CC measurements (23, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88), 10 did not identify all-cause mortality as the outcome (89, 90, 91, 92, 93, 94, 95, 96, 97, 98), 8 only reported TC (99, 100, 101, 102, 103, 104, 105, 106), 5 did not report HR/RR/OR for the association between mortality and CC (107, 108, 109, 110, 111), one for including immobile patients (112), and one for being a repeated study (113), which was the same trial with one included study (30). Ultimately, thirty-seven studies met the inclusion criteria (17, 18, 19, 21, 22, 25, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 53, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128).

Figure 1.

Flowchart of literature searching and study selection

Characteristics of Included Studies

Of the thirty-seven longitudinal observational studies included in this review, there were a total of 62,736 participants from three continents: sixteen in America (17, 21, 32, 33, 34, 40, 41, 42, 43, 44, 53, 121, 122, 124, 125, 127), fourteen studies in Asia (18, 19, 22, 25, 30, 31, 37, 38, 39, 115, 116, 118, 119, 120), and seven in Europe (35, 36, 114, 117, 123, 126, 128). Details on the study populations and the study results are shown in Table 1 and Appendix Table S3. Only three studies reported their effect sizes based on both continuous CC and categorical CC patterns (30, 44, 128). Six studies evaluated total death risk only with CC in the continuous pattern (17, 18, 31, 39, 114, 119), and twenty-eight studies provided results only with CC in the categorical patterns (19, 21, 22, 25, 32, 33, 34, 35, 36, 37, 38, 40, 41, 42, 43, 53, 115, 116, 117, 118, 120, 121, 122, 123, 124, 125, 126, 127), two of which were dose-response CC (25, 53). Eleven studies selected adults with no age boundaries (17, 30, 32, 34, 38, 39, 41, 44, 53, 122, 127), while other studies focused only on the elderly. Thirty-two studies provided adjusted effect sizes, but the remaining five studies did not (22, 114, 116, 117, 120) Eight studies reported long-term death risk evaluation (≥ 5 years) (17, 21, 25, 36, 40, 114, 121, 126).

Table 1.

Characteristic of included studies

Authors[ref.], Year	Country (study duration)	male %	No. of deaths/Subjects (Mortality %)	Age (year) Mean±SD	CC (cm) Mean±SD	comorbidities at baseline	Reported effect size	Effect sizes (95% CI)	Covariates in adjusted model	Follow-up (years)
Mason C et al.(17), 2008	Canada (1981–1993)	47.1	571/10638(5.4)	Male: 38.0±13.2 Female: 38.5±13.3	Male: 36.6±2.9 Female: 34.4±3.0	NA	HR	per 1 SD increase Male: 0.78 (0.66–0.92) Female: 0.86 (0.73–1.02)	Adjusted for age, WC, BMI, smoking status, alcohol consumption, leisure-time physical activity	13.0
Wijnhoven HAH et al. (114), 2010	Netherlands (1992–2007)	49.6	1078/1667 (male: 72.5, female: 57.0)	Male: 74.7±5.7 Female: 74.1±5.9	Male: 35.6±2.7 Female: 35.4±3.1	NA	HR	per 1 SD decrease Male: 1.45 (1.22–1.71)a Female: 1.30 (1.15–1.48)	None	15.0
Tsai AC et al. (115), 2011	China Taiwan (1999–2003)	53.2	566/4191 (13.5)	68.8±9.1	Male: 34.7±3.7 Female: 32.9±3.9	NA	HR	Age at 53–64 years CC<30/27: 2.64 (0.82–8.46) CC≥30/27: Ref. Age at 65–74 years CC<30/27: 2.59 (1.66–4.06) CC≥30/27: Ref. Age ≥ 75 years CC<30/27: 1.98 (1.52–2.58) CC≥30/27: Ref.	Adjusted for age, sex, exercise, smoking status	4.0
Lin SJ et al. (116), 2012	China Taiwan (2008–2009)	56.9	23/174(13.2)	82.5±7.3	27.1±4.2	CVD, dementia, DM, COPD, tumor, Parkinson's disease, peptic ulcer	RR	CC<31: Ref. CC≥31: 0b	None	1.0
Sharifi F et al. (18), 2012	Iran (2006–2009)	41.3	74/247 (30.0)	76.7±8.4	32.7±4.2	DM, HTN	HR	per 1 cm increase 0.92 (0.87–0.98)	Adjusted for age and sex	3.3
Tsai AC et al. (19), 2012	China Taiwan (NA)	100.0	19/160(11.9)	81.1±NA	28.3±4.3	HTN, CVD, DM, dyslipidemia, respiratory disease, stroke, kidney disease, cancer, mental disorder	HR	CC<28: 3.79 (1.36–10.52) CC≥28: Ref.	Adjusted for age	1.0
Bourdel MI et al. (117), 2016	France (2007–2012)	52.6	266/606 (43.9)	>70	NA	cancer, lymphoma	OR	CC<31:3.08 (1.69–5.61) CC≥31: Ref.	None	2.0
Ho SC et al. (118), 2016	China Taiwan (2011–2014)	94.2	22/104(21.2)	74.2±6.9	32.9±3.2	COPD	HR	CC<30:4.40 (1.82–10.63) CC≥30: Ref.	Adjusted for age, smoking status and FEV1%.	3.0
Kamiya K et al. (119), 2017	Japan (2008–2014)	65.4	72/599(12.0)	74.8±6.3	31.6±3.3	HTN, DM, dyslipidemia, prior HF	HR	per 1 SD increase 0.91 (0.56–1.47)	Adjusted for age, sex, BMI, grip strength, gait speed, eGFT, logBNP, prior HF	Median: 1.63 IQR: 2.1
Wu CJ et al. (53), 2017c	USA (1999–2006)	48.0	NA/7629 (NA)	49.3±18.7	Median: 37.8 IQR: 35.3–40.6	congestive HF, coronary heart disease, angina/angina pectoris, heart attack stroke, cancer or malignancy	HR	CC<35.3: Ref. CC35.3–37.7: 0.60 (0.47–0.77) CC37.8–40.6: 0.50 (0.37–0.67) CC>40.6: 0.43 (0.28–0.65)	Adjusted for age, sex, race, BMI, SBP, SFG, SC, SA, ALT, SUA, C-reactive protein, history of several diseases*, smoking, moderate to vigorous recreational activity	NA
Yajima Y et al. (120), 2017	Japan (2014–2015)	38.4	15/151 (10.7)	82.0±9.6	NA	NA	RR	CC<31: Ref. CC≥31: 0.59 (0.20–1.77)	None	1.0
Easton JF et al. (121), 2018	Mexico (2001–2015)	36.9	422/1298 (32.5)	62.1±9.5	NA	HTN, DM, cancer, lung disease, heart attack, stroke, articular disease and tuberculosis	HR	CC≤31: 1.31 (1.02–1.69) CC>31: Ref.	Adjusted for age, sex, marital status, living environment, education, smoking, physical activity, number of comorbidities	14.0
Weng CH et al. (25), 2018	China Taiwan (2002–2009)	42.2	255/329 (77.5)	78.6±7.8	27.5±4.3	HTN, DM	HR	Male/female CC<26.0/25.0: Ref. CC26.0–29.9/25.0–28.3: 0.62 (0.45–0.86) CC≥30.0/28.4: 0.51 (0.35–0.74)	Adjusted for age, sex, smoking, alcohol, betel nut, KPS, SA, HTN, DM	7.0
da Silva JR et al.(122), 2019	Brazil (2016–2017)	45.2	127/334 (38.2)	Median 63 IQR55-72	NA	Cancer	HR	Male/female: CC≤34/33: 2.00(1.45–2.76) CC>34/33: Ref.	Adjusted for age, sex, gastrointestinal tract tumor, KPS, C-reactive protein, and Patient-Generated Subjective Global Assessment Short Form	0.2
de Almeida Roediger, M et al. (21), 2019	Brazil (2000–2010)	42.3	769/1504 (40.2)	67.2±0.2	35.4±0.15	HTN, DM, CVD, lung diseases, stroke, cancer	RR	Overall: CC<31: 1.33 (1.08–1.64) CC≥31: Ref. Male: CC<31: 1.13 (0.85–1.50) CC≥31: Ref. Female: CC<31: 1.45 (1.07–1.96) CC≥31: Ref.	Adjusted for age, sex, marital status, schooling, working status, income, weekly alcohol intake, sedentary lifestyle, smoking, HTN, DM, CVD, lung disease, stroke, cancer, number of diseases, Mini Mental State exam, and Geriatric Depression Scale	10.0
Monereo MM et al. (123), 2019	Spain(2014-2018)	56.7	16/187 (8.6)	76.1±0.5	Male/Female Median: 35.0/33.0 IQR: 5.0/5.0	Acute illness or decompensation	HR	100-day mortality for male/female: CC≤31/29: 1.88 (1.20–2.98) CC>31/29: Ref.	Adjusted for sex, and comorbidity	0.3
Rodrigues J et al. (124), 2019	Brazil (2010–2014)	65.0	61/173 (35.2)	Median: 69.4 Range: 64.5–75.4	34.1±3.8	Patients undergoing maintenance hemodialysis	HR	CC<31: 2.27 (1.27–4.05) CC≥31: Ref.	Adjusted for sex, age, dialysis vintage, and presence of DM	2.0
Yang M et al. (31), 2019	China (2016)	31.9	73/329 (22.2)	Median: 85.0 IQR: 3.0	Male: 33.0±4.0 Female: 32.0±3.0	HTN, ischemic heart disease, chronic HF COPD, DM, stroke, cancer, osteoarthritis, Parkinson disease, cognitive impairment, depression	HR	per 1 cm increase 1.60 (0.96–2.66)	Adjusted for age, gender, ischemic heart disease, Parkinson disease, malnutrition risk or malnutrition and BMI	1.0
Aliberti MJR et al. (125), 2020	Brazil (2014–2017)	36.7	90/665 (13.5)	78.7±8.3	NA	infectious diseases, acute anemia, uncontrolled HTN, decompensated DM, symptomatic congestive HF, refractory pain	HR	Male/female: CC≤34/33: 1.22 (0.74–1.99) CC>34/33: Ref.	Adjusted for age, sex, race, income, Charlson comorbidity index, depressive symptoms, cognitive impairment, and unintentional weight loss.	1.0
de Sousa OV et al. (126), 2020	Portugal (2012–2017)	40.0	22/79 (27.8)	Median: 79 IQR: 28	NA	Alzheimer's disease	HR	CC≤31: 4.27 (1.63–11.16) CC>31: Ref.	Adjusted for sex, age, marital status, education, Mini Mental State Examination	5.0
Moon S et al. (22), 2020	Korea (2011–2013)	26.4	94/276 (34.1)	80.5±7.1	27.6±3.8	HTN, DM, heart disease, chronic respiratory disease, urinary incontinence	OR	CC<28: 2.01 (1.21,3.34) CC≥28: Ref.	None	2.0
Sousa IM et al. (34), 2020	Brazil (2017–2019)	47.2	40/250 (16.0)	Median: 63 Range: 55–73	NA	cancer	HR	Male/female: CC≤34/33: 3.01 (1.52–5.98) CC>34/33: Ref.	Adjusted for age, sex, and stage of disease	1.0
Tarnowski M et al. (127), 2020	Brazil (2013–2015)	45.4	20/528 (3.9)	52.8 ± 16.2	Male: 36.7±12.9 Female: 35.7±5.3	gastroenterological disorders, cancer and related complications, cardiac diseases, and neurologic diseases	HR	Male/female: CC<34/33: 2.02 (0.76–5.36) CC≥34/33: Ref.	Adjusted for age and stress of diseases	Median: 9.0 days IQR: 3.0–19.0 days
Valmorbida E et al. (128), 2020	Italy (2015–2017)	28.5	52/144 (36.1)	Median: 86 IQR: 82–91	Median: 30.8 IQR: 28.0–33.0	DM, cardiovascular disease, HTN, stroke, pressure ulcers, dementia, cancer	HR	per 1 cm decrease 1.16 (1.07–1.26) CC<31: 3.39 (1.80–6.39) CC≥31: Ref.	Adjusted for age, sex, DM, CVD, stroke, cancer, recent hip fracture, number of drugs used, Short Portable Mental Status Questionnaire	1.5
Abreo AP et al. (44), 2021	US (1999–2004)	50.2	748/11871 (6.3)	50.5±18.5	38.2±4.4	coronary artery disease, HF, stroke, DM, history of malignancy, HTN, emphysema	HR	per 1 cm increase Male: 0.92 (0.88–0.96) Female: 0.88 (0.84–0.92) Male: CC<34: 3.15 (2.54–3.89) CC≥34: Ref Female: CC<33: 3.52 (2.76–4.50) CC≥33: Ref.	Adjusted for age, race, coronary artery disease, HF, stroke, DM, history of malignancy, HTN, emphysema, kidney disease, BMI, WC, SAe	4.7
Dos Santos RCO et al. (41), 2021	Brazil (2018–2019)	51.4	47/601 (7.8)	55.7±14.8	36.5±4.3	Cancer, CVD and pulmonary disease	OR	Male/female: CC≤34/33: 2.49 (1.27–4.85) CC>34/33: Ref.	Adjusted for White ethnicity and Charlson comorbidity index	0.5
Fernandes DPS et al. (40), 2021	Brazil (2009–2018)	46.6	197/736 (26.8)	Median: 70 IQR: 64–77	Median: 35.6 IQR: 33.1–38.2	NA	HR	CC<34.5: 2.36 (1.51–3.68) CC≥34.5: Ref. CC<31: 2.56 (1.47–4.49) CC≥31: Ref.	Adjusted for sex, age, education, quality of diet, physical activity, smoking and BMI	9.0
Ge YZ et al. (39), 2021	China (2013–2020)	NA	NA/524 (NA)	>18	31.4±3.6	cancer cachexia	OR	per 1 cm decrease 0.95 (0.90–1.01)	Adjusted for EORTC QLQ-C30, albumin, neutrophils, erythrocyte, chronic disease and family history of cancer	1.0
Nishikawa H et al (38), 2021	Japan (2013–2020)	49.0	64/631 (10.1)	Median: 65 IQR: 52–71	Male/Female Median: 37.3/34.7 IQR:4.6/4.5	Liver diseases	HR	Male/female: CC≤34/33: 5.03 (2.62–9.84) CC>34/33: Ref.	Adjusted for age, sex, presence of liver cirrhosis, presence of hepatocellular carcinoma, grip strength, skeletal muscle index, SA, platelet count, prothrombin time, aspartate aminotransferase, eGFR, cause	3.5
Sanchez RD et al (36), 2021	Belgium (2013–2019)	44.0	52/373 (13.9)	73.1±6.0	NA	NA	HR	CC<31: 3.35 (1.85–6.07) CC≥31: Ref.	Adjusted for age, sex, number of concomitant diseases, number of drugs, physical activity level, and cognitive status	5.0
Sobestiansky S et al. (35), 2021	Sweden (2009–2010)	32.1	26/56 (46.4)	84.1±7.3	33.8±4.4	HF, MI, arrhythmia, peripheral vascular disease, DM, COPD, rheumatic disease, CVD, cognitive disease, depression, cancer, osteoporosis, fracture	HR	CC<31: 1.99 (0.82–4.82) CC≥31: Ref.	Adjusted for age and Charlson comorbidity index	2.0
Teixeira PP et al.(33), 2021	Brazil (2019–2020)	43.8	16/176 (9.1)	68.2±10.4	31.7±3.7	COPD	HR	Male/female: CC≤34/33: 1.20 (0.25–5.67) CC>34/33: Ref.	Adjusted for modified Medical Research Council Scale and age	NA
Yin L et al. (30), 2021	China (2012–2019)	52.7	2751/12138 (22.7)	56.9±11.6	33.1±3.8	Cancer, HTN, DM, coronary heart disease, anemia	HR	per 1 sd increase 0.94 (0.90–0.98) Male/female: CC<32.8/30.0: 1.13 (1.03–1.23) CC≥32.8/30.0: Ref	Adjusted for age, sex, drinking, tumor stage, cancer site, differentiation grade, radical surgery, chemotherapy, hospitalization duration, prealbumin, NRS2002 score, global quality of life score, KPS score, TSF	2.8
Bernardes S et al. (43), 2022	Brazil (2002–2009)	65.3	18/144 (12.5)	64.6±8.5	34.4±3.7	COPD	OR	Male/female: CC≤34/33: 3.69 (1.06–12.87) CC>34/33: Ref.	Adjusted for age, smoking status, and forced expiratory volume in 1 s	2.0
de Sousa IM et al. (42), 2022	Brazil (2017–2019)	50.6	46/178 (25.8)	60.8±12.6	NA	gastric and colorectal cancer	HR	Male/female: CC<34/33: 2.40 (1.10–5.20) CC≥34/33: Ref	adjusted for age, TNM stage, type of treatment, and site of cancer	1.0
Ren C et al. (37), 2022	China (2017–2018)	71.6	105/514 (20.4)	86.5±6.0	NA	Cerebral infraction, coronary heart disease, DM, HTN	HR	CC<28.8: 2.78 (1.27–5.98) CC≥28.8: Ref	Adjusted for fasting plasma glucose, hemoglobin and nutritional risk	2.0
Wu SE et al.(32), 2022	US (1999–2002)	51.7	NA/2532 (NA)	Range: 50–85	34.4±3.7	Arthritis, congestive HF, coronary heart disease, angina pectoris, heart attack, stroke, emphysema	HR	Male/female: CC<34/33: 1.55 (0.70–3.42) CC≥34/33: Ref	Adjusted for age, sex, race, smoking, comorbidity, education level, BMI	NA

Abbreviation: SD, standard deviance; CC, calf circumference; HR, hazard ratio; CI, confidence interval; NA, not available; WC, waist circumference; BMI, body mass index; SBP, systolic blood pressure; eGFT, estimated glomerular filtration rate; logBNP, logarithm for B-type natriuretic peptide; IQR, interquartile range; HF, heart failure; SFG, serum fasting glucose; SC, serum cholesterol; SA, serum albumin; ALT, alanine aminotransferase; SUA, serum uric acid; HTN, hypertension; DM, diabetes mellitus; CVD, cardiovascular diseases; EORTCQLQ-C30, The European Organization for Research and Treatment of Cancer QLQ-C30; MI, myocardial infarction; TSF, triceps skinfold thickness; KPS, Karnofsky Performance Status score; a. Only males with CC<37.6 cm were evaluated; b.0 died in those with CC≥31 cm; c. Detailed information about mortality dates were obtained from the National Center for Health Statistics of the Centers for Disease Control and Prevention (CDC) in USA, or censored at December 31, 2006; d. History of several diseases included congestive heart failure, coronary heart diseases, angina/angina pectoris, heart attack, stroke, cancer; e. Adjusted covariates for continuous CC.

Study Quality and Certainty Assessment

Study quality scores ranged from 6 to 9 (mean 7.7), with nine studies having a full score of 9 (21, 30, 31, 36, 39, 40, 44, 115, 121). Only three studies were awarded the status of fair quality (19, 33, 114), whereas the other studies were rated as good quality (Table 2). The certainty for estimate of mortality was high, and more detailed GRADE assessments can be found in Appendix Table S4.

Table 2.

Assessment of risk of bias of the studies included

Authors (ref.), Year	Selection			Comparability			Outcome		Overall Score*
	Representativeness of the exposed cohort	Selection of the non-exposed cohort	Assessment of exposure	Demonstration that outcome of interest was not present at start of study	Comparability of cohorts on the basis of the design or analysis	Ascertainment of outcome	Long enough follow-up for outcomes to occur	Adequacy of follow-up of cohorts
Mason C et al. (17), 2008	*	*	*	*	**	*	*	-	8
Wijnhoven HAH et al. (114), 2010	*	*	*	*	-	*	*	-	6
Tsai AC et al. (115), 2011	*	*	*	*	**	*	*	*	9
Lin SJ et al. (116), 2012	-	*	*	*	*	*	*	*	7
Sharifi F et al. (18), 2012	*	*	*	*	**	*	-	*	8
Tsai AC et al. (19), 2012	-	*	*	*	*	*	*	-	6
Bourdel MI et al. (117), 2016	*	*	*	*	-	*	*	*	7
Ho SC et al. (118), 2016	-	*	*	*	*	*	*	*	7
Kamiya K et al. (119), 2017	-	*	*	*	**	*	*	-	7
Wu CJ et al.(53), 2017c	*	*	*	*	**	*	*	-	8
Yajima Y et al.(120), 2017	-	*	*	*	*	*	*	*	7
Easton JF et al. (121), 2018	*	*	*	*	**	*	*	*	9
Weng CH et al. (25), 2018	-	*	*	*	**	*	*	*	8
da Silva JR et al. (122), 2019	-	*	*	*	**	*	-	*	7
de Almeida Roediger, M et al. (21), 2019	*	*	*	*	**	*	*	*	9
Monereo MM et al. (123), 2019	-	*	*	*	**	*	-	*	7
Rodrigues J et al. (124), 2019	-	*	*	*	**	*	*	*	8
Yang M et al. (31), 2019	*	*	*	*	**	*	*	*	9
Aliberti MJR et al. (125), 2020	-	*	*	*	**	*	*	*	8
de Sousa OV et al. (126), 2020	-	*	*	*	**	*	*	*	8
Moon S et al. (22), 2020	*	*	*	*	*	*	*	*	8
Sousa IM et al. (34), 2020	-	*	*	*	**	*	*	*	8
Tarnowski M et al. (127), 2020	-	*	*	*	**	*	-	*	7
Valmorbida E et al. (128), 2020	-	*	*	*	**	*	*	-	7
Abreo AP et al.(44), 2021	*	*	*	*	**	*	*	*	9
Dos Santos RCO et al.(41), 2021	*	*	*	*	*	*	-	*	7
Fernandes DPS et al. (40), 2021	*	*	*	*	**	*	*	*	9
Ge YZ et al. (39), 2021	*	*	*	*	**	*	*	*	9
Nishikawa H et al. (38), 2021	-	*	*	*	**	*	*	*	8
Sanchez RD et al.(36), 2021	*	*	*	*	**	*	*	*	9
Sobestiansky S et al. (35), 2021	-	*	*	*	*	*	*	*	7
Teixeira PP et al. (33), 2021	-	*	*	*	*	*	-	*	6
Yin L et al. (30), 2021	*	*	*	*	**	*	*	*	9
Bernardes S et al. (43), 2022	-	*	*	*	*	*	*	*	7
de Sousa IM et al. (42), 2022	-	*	*	*	*	*	*	*	7
Ren C et al. (37), 2022	-	*	*	*	*	*	*	*	7
Wu SE et al. (32), 2022	*	*	*	*	**	*	-	*	8

Abbreviation: “-” =criteria not met; “*”=one point for meeting criteria; “**”=two points for meeting criteria; * Overall score: 0–3 points = “poor quality”; 4–6 points = “fair quality”; 7–9 points = “good quality”

Meta-analyses for Effect Sizes

Figure 2 illustrates the forest plot of the pooled RR values for all-cause mortality per one cm CC, and further subgroup analyses results are also shown for the various participant types, including community, nursing home, and hospital. Among the thirty-seven studies that met the inclusion criteria (17, 18, 19, 21, 22, 25, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 53, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128), two were not included in the meta-analysis for the following reasons: one study among 174 elderly reported that nobody died in those with CC≥ 31 cm (116), while the other study didn't report the mortality so the OR could not be converted to RR (39). A significant association was observed in the pooled analyses using either the fixed effects model (RR, 0.967; 95% CI, 0.963–0.971; P<0.01) or the random effects model (RR, 0.947; 95% CI, 0.937–0.958; P<0.01). No significant difference was detected for those living in communities, nursing homes, or hospitals (P=0.617). Significant heterogeneity was observed across thirty-five individual studies (I²=75.7%, P<0.01), but no publication bias was found (P=0.20). We also performed sensitivity analyses to evaluate the stability of overall effect size and detected no outliers (Appendix Figure S1).

Figure 2.

Forest plots for pooled associations between total death and CC (A) Fixed-effects model; (B) Random-effects model

Subgroup analyses

In subgroup analyses, we found statistical significance in most subgroups, except for lying during CC measurements (RR, 1.10; 95% CI, 0.77–1.58; P=0.607). However, only two studies reported this measurement when lying (31, 35) (Table 3). There was a low credible subgroup difference based on the eight criteria introduced by the ICEMAN tool (Appendix Table S5).

Table 3.

The results of subgroup analyses and meta-regression for the associations of all-cause mortality per one cm increase

Moderators	No. of studies	Subgroup analyses				Heterogeneity		Meta-regression	Publication bias (Begg's test)
		Fixed effect models		Random effect models
		RR (95% CI)	P value	RR (95% CI)	P value	I2	P value	P value	P value
All study combined	35	0.967 (0.963–0.971)	<0.001	0.947 (0.937–0.958)	<0.001	75.7%	<0.001		0.200
Publication year								0.632
< 2017	8	0.950 (0.941–0.959)	<0.001	0.947 (0.933–0.960)	<0.001	31.3%	0.158		0.074
≥ 2017	27	0.971 (0.966–0.976)	<0.001	0.949 (0.936–0.961)	<0.001	78.1%	<0.001		0.253
Follow up duration (years)								0.806
< 5	24	0.969 (0.964–0.974)	<0.001	0.947 (0.933–0.960)	<0.001	77.1%	<0.001		0.414
≥ 5	8	0.963 (0.955–0.971)	<0.001	0.952 (0.936–0.968)	<0.001	72.7%	<0.001		0.043
NA	3	0.915 (0.882–0.948)	<0.001	0.928 (0.874–0.986)	0.015	47.5%	0.149		1.000
Sample size								0.249
< 500	18	0.949 (0.940–0.958)	<0.001	0.942 (0.930–0.955)	<0.001	32.2%	0.093		0.405
≥ 500	17	0.972 (0.967–0.976)	<0.001	0.952 (0.939–0.965)	<0.001	82.1%	<0.001		0.024
Continents								0.782
Asia	12	0.974 (0.968–0.980)	<0.001	0.941 (0.918–0.965)	<0.001	83.2%	<0.001		0.150
America	17	0.965 (0.959–0.971)	<0.001	0.953 (0.940–0.966)	<0.001	70.1%	<0.001		0.010
Europe	6	0.941 (0.928–0.954)	<0.001	0.940 (0.926–0.955)	<0.001	15.6%	0.311		0.368
Study quality-NOS								0.364
Fair (4–6 stars)	3	0.934 (0.913–0.955)	<0.001	0.934 (0.913–0.955)	<0.001	0.0%	0.408		1.000
Good (7–9 stars)	32	0.968 (0.964–0.972)	<0.001	0.949 (0.930–0.960)	<0.001	76.4%	<0.001		0.112
Age (years)								0.869
< 65	10	0.977 (0.972–0.983)	<0.001	0.951 (0.933–0.969)	<0.001	83.3%	<0.001		0.064
≥ 65	23	0.952 (0.945–0.958)	<0.001	0.944 (0.932–0.956)	<0.001	58.1%	0.079		0.272
NA	2	0.966 (0.951–0.981)	<0.001	0.966 (0.951–0.981)	<0.001	0.0%	0.860		1.000
Gender								0.975
Male	4	0.956 (0.940–0.971)	<0.001	0.948 (0.915–0.982)	0.003	76.4%	0.005		0.308
Female	4	0.947 (0.933–0.962)	<0.001	0.947 (0.924–0.971)	<0.001	60.1%	0.057		1.000
Both	31	0.969 (0.965–0.974)	<0.001	0.947 (0.936–0.959)	<0.001	76.1%	<0.001		0.518
Average CC of participants (cm)								0.104
< 31	4	0.938 (0.919–0.958)	<0.001	0.938 (0.919–0.958)	<0.001	0.0%	0.438		0.734
≥ 31	21	0.967 (0.962–0.972)	<0.001	0.942 (0.928–0.957)	<0.001	80.6%	<0.001		0.657
NA	10	0.970 (0.962–0.977)	<0.001	0.962 (0.949–0.976)	<0.001	57.7%	0.011		0.409
Position during CC measurements								0.088
Standing	5	0.943 (0.930–0.956)	<0.001	0.943 (0.930–0.956)	<0.001	0.0%	0.565		0.072
Sitting	18	0.970 (0.964–0.975)	<0.001	0.936 (0.918–0.954)	<0.001	83.8%	<0.001		0.327
Lying	2	0.961 (0.899–1.027)	0.240	1.100 (0.765–1.581)	0.607	77.0%	0.037		1.000
NA	10	0.969 (0.962–0.976)	<0.001	0.966 (0.954–0.978)	<0.001	53.2%	0.019		0.119
Adjusted effect sizes used								0.627
Yes	31	0.968 (0.964–0.973)	<0.001	0.948 (0.938–0.959)	<0.001	77.1%	<0.001		0.236
No	4	0.944 (0.928–0.961)	<0.001	0.943 (0.924–0.963)	<0.001	18.0%	0.300		0.462
BMI adjusted effect sizes used								0.292
Yes	7	0.935 (0.921–0.948)	<0.001	0.937 (0.916–0.957)	<0.001	49.0%	0.047		0.348
No	24	0.972 (0.967–0.976)	<0.001	0.953 (0.941–0.965)	<0.001	76.5%	<0.001		0.027

Note: CC, calf circumference; RR, relative risk; CI, confidence interval; NOS, Newcastle-Ottawa scale; NA, not available

Discussions

To our knowledge, this is the most comprehensive systematic review and meta-analysis so far that summarizes the results of thirty-seven longitudinal observational studies adding up to a total of 62,736 subjects to evaluate the association of all-cause mortality with calf circumference by combining various original analyses based on empirical dichotomic CC, continuous CC, and dose-response CC. Due to heterogeneous CC patterns from various studies, effect sizes in categorical CC were transcribed into ones in the continuous pattern. 5% death risk reduction was found per one cm CC increase from the included studies, albeit 75.7% heterogeneity was found. Thus, because using I² statistics to assess heterogeneity in observational studies is problematic in the context of precise estimates and large sample sizes (17, 30, 44, 53), we assess the certainty of our meta-analysis using the GRADE approach as a high quality and critically important feature from the study design, risk of bias, imprecision, inconsistency, indirectness, and publication bias.

This updated review confirms and extends the results from the included studies by a comprehensive trend analysis of adjusted risk ratio over the entire range of CC levels, both within and across studies. A systematic meta-analytic review including 23 studies was published very recently and provided the overall hazard ratio (HR) and 95% confidence interval for dichotomous variables, reporting that those with low CC had a 2.42 times higher mortality risk than those with normal CC (29). However, only studies using specific empirical cut-offs of CC value were considered in their meta-analyses, and some studies were left out due to “inappropriate” analyses based on continuous CC and/or respective dose-response CC (18, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44). A 1 cm increase in CC results in a 5% reduction in mortality rate. When one CC unit is increased to 3 cm, 4 cm, and 5 cm, the corresponding reductions in mortality rate are 15%, 19%, and 24%, respectively. Early warning signs of dangerous calf circumferences should be explored for death risks in the context of additional longitudinal studies and in-depth analyses, which could help further our understanding of the role that calf circumference plays on predicting death in various ethnic groups.

Sarcopenia is a condition related to aging and is associated with decreased activity owing to reduced skeletal muscle mass (129, 130, 131, 132, 133). Although diagnostic imaging approaches have been recommended to measure muscle mass, including computed tomography (CT), magnetic resonance imaging (MRI), and dual energy X-ray absorptiometry (DXA), their large-scale application into communities or epidemiological research studies is restricted due to radiation exposure risks, the need for skilled technicians, time-consuming procedures, and expensive costs (131, 132, 134, 135, 136, 137). Therefore, a simpler method is needed for assessing muscle mass than diagnostic imaging. Compared to other anthropometric parameters, CC was observed to be most associated with muscle mass and malnutrition, because of the greater lean mass in the calves (17, 126, 138, 139). The systematic reviews and meta-analyses conducted here provide further evidence that CC is inversely related to total mortality, especially in elderly populations, though only few available studies focused on those with ages of 18–65 years (17, 30, 32, 41, 44, 53, 127).

As proposed by a recent study using a diverse population sample, using BMI-adjusted CC could help to remove the confounding effects of obesity and thereby improve CC as a useful practical estimate of skeletal muscle mass (140). As we know, CC values are much higher in overweight/obese populations, while smaller in those with extremely low BMI, regardless of age, ethnicity, or race. In other words, obesity may be considered as one moderating effect on CC diagnostic performance for muscle mass (141, 142). Among the thirty-seven included studies, however, only seven studies provided BMI as one covariate in their multivariate analyses (17, 31, 32, 40, 44, 53, 119). Further subgroup analyses and meta-regression failed to find significant variations between BMI-adjusted effect sizes and unadjusted ones. Nevertheless, more studies should be conducted to consider the influence of obesity may have when developing early warning values of CC for death risk predictions.

The first study regarding the association of mortality with CC reported that CC had a significantly protective effect in men, but not for women (17), but this result contrasted with the results of another American study (21). The remaining studies in our review did not present their respective results in women and men (18, 19, 22, 25, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 53, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128), though sex-specific cut-off points of CC were applied in the analyses of fourteen studies (25, 30, 32, 33, 34, 38, 41, 42, 43, 115, 122, 123, 125, 127). We obtained marginally significant results in our subgroup analyses in continuous CC (per one cm increase of CC) after merging sex-separated effect sizes in these three studies (17, 21, 44, 114). Therefore, no sufficient evidence can be used to resolve the inconsistent results in women and men thus far.

A major strength of our study is the application of advanced techniques of statistical analyses that allowed us to summarize adjusted associations across studies and over the entire range of CC values, despite the very heterogeneous categorization of CC in the individual studies. Nonetheless, some important limitations must be noted when interpreting these results. Firstly, as the original data in individuals were not available from eligible articles, the risk estimation may be less accurate based on reported effect sizes, especially when the effect sizes were transferred into RR per one cm CC increase from odds ratio or hazard ratio. Secondly, various potentially influential covariates were used for adjustments in the original analyses, which may influence the overall effect sizes in the metaanalysis. Moreover, some studies included in our meta-analysis did not provide adjusted effect sizes (22, 114, 117, 120), though no significant variation was detected in subgroup analyses and meta-regression between adjusted and unadjusted effect sizes. Thirdly, since subgroup data was obtained from the original articles, some studies were not included in our subgroup analyses if no follow-up duration was reported (32, 33, 53) and no age subgroup of the participants was provided (32, 120, 121, 122, 126). Finally, although our systematic review searched four databases and extensive checks for completeness by cross-referencing was employed, we could not exclude missed studies of relevance.

Conclusions

Despite these limitations, we provided the most comprehensive empirical evidence to date that adequate CC is inversely related to all-cause mortality risk and the certainty for the result possesses high quality based on the GRADE approach. However, in-depth analyses of the assessed associations in the context of additional longitudinal studies are highly desirable to enable more precise estimates and a better understanding of the role of calf circumference in mortality prediction abilities, especially for adults of 18–65 years of age. Further international studies are also needed for the optimal cutoff points of calf circumference.

Conflict of interest

No.

Electronic Supplementary Material

Supplementary material is available for this article at https://doi.org/10.1007/s12603-022-1838-0 and is accessible for authorized users.

Supplementary material, approximately 341 KB.