METAB O LI S M C LINI C AL A ND E X PERIMENT A L 65 ( 2 0 1 6) 883 - 892
`
`Available online at www . sciencedirect . com
`
`Metabolism
`
`www. metabolismjou rnal. com
`
`ELSEVIER
`
`Muscle grip strength predicts incident type 2
`diabetes: Population-based cohort study
`
`Clin ica l and f-,,:pe.rirmmtal
`
`Metabolism
`.,_..__,.,._,,_ __
`·-•---•'(cid:173)
`··--· .. -· ... -·-·
`.,_ ...... _ ... ___ _
`
`CrossMark
`
`Joule]. u a,b,*, Gary A. Wittertb, Andrew Vincent b, Evan Atlantis C, Zumin Shi d,
`Sarah L. Appleton a,b, Catherine L. Hill a, Alicia]. Jenkins e,
`Andrzej S. Januszewski e, Robert]. Adams a
`
`a The Health Observatory, School of Medicine, University of Adelaide
`b Freemasons Foundation Centre for Men's Health, School of Medicine, University of Adelaide
`c School of Nursing and Midwifery, Western Sydney University
`d Population Research and Outcome Studies, School of Medicine, University of Adelaide
`e NHMRC Clinical Trials Centre, Sydney Medical School, University of Sydney
`
`ARTICLE INFO
`
`ABSTRACT
`
`Article history:
`Received 11 February 2016
`Accepted 22 March 2016
`
`Keywords:
`Skeletal muscle
`Lean mass
`Dual energy X-ray absorptiometry
`Muscle strength
`Grip strength
`
`Objectives. To determine the longitudinal relationship of muscle mass and strength
`with incident type 2 diabetes, and previously unstudied mediating effects of testosterone
`and inflammation.
`Methods. Community-dwelling male participants (aged :>35 years) of the Men Androgen
`Inflammation Lifestyle Environment and Stress (MAILES) Study underwent biomedical
`assessment in 2002-2006 and 2007-2010, including hand grip strength (dynamometer),
`testosterone and inflammatory markers. Body composition (dual-energy X-ray absorptiometry)
`was assessed at baseline only. Incident type 2 diabetes was defined as a self-reported doctor
`diagnosis, diabetes medication use, fasting plasma glucose :>7.0 mmoVL, or glycated
`haemoglobin :>6.5% (48 mmoVmol) at follow-up, that was not present at baseline.
`Results. Ofn = 1632 men, incident type 2 diabetes occurred in 146 (8.9%). Muscle mass was
`not associated with incident type 2 diabetes. Grip strength was inversely associated with
`incident type 2 diabetes [unadjusted odds ratio (OR) per 5 kg: 0.87, 95% confidence interval (CI):
`0.80--0.95; adjusted OR, 95% CI: 0.87, 0.78--0.97]. Arm muscle quality (grip strength divided by
`arm lean mass) was similarly associated with incident type 2 diabetes. Testosterone, IL-6 and
`TNF-a did not significantly mediate the associations. The population attributable fraction of
`type 2 diabetes from low grip strength was 27% (13-40%), assuming intervention could
`increase strength by 25%.
`Conclusions. Reduced muscle strength, but not reduced muscle mass, is a risk factor for
`incident type 2 diabetes in men. This is not mediated by testosterone or inflammation.
`Intervention could prevent a substantial proportion of disease.
`© 2016 Elsevier Inc. All rights reserved.
`
`Abbreviations: ASM, appendicular skeletal muscle mass; ASMI, appendicular skeletal muscle mass index; DXA, dual-energy xray
`absorptiometry; FAMAS, Florey Adelaide Male Ageing Study; FM!, fat mass index; MAILES, Men Androgen Inflammation Lifestyle
`Environment and Stress; NWAHS, North West Adelaide Health Study; MET, metabolic equivalent; PAF, population attributable fraction.
`* Corresponding author at: The Health Observatory Basil Hetzel Institute, Discipline of Medicine, University of Adelaide, The Queen
`Elizabeth Hospital Campus, Woodville Rd, Woodville, South Australia, 5011, Australia. Tel.: +61404 787 603; fax: +61 8 8222 6042.
`E-mail address: joule.li@adelaide.edu.au /J.J. Li).
`
`http://dx.doi.org/10.1016/j .metabol.2016.03.011
`0026-0495/© 2016 Elsevier Inc. All rights reserved.
`
`Sanofi Exhibit 2179.001
`Mylan v. Sanofi
`IPR2018-01676
`
`

`

`884
`
`MET A B O LI S M C LINIC A L A ND E X PE R IME N T A L 65 (2016) 883-892
`
`1.
`
`Introduction
`
`The prevalence of type 2 diabetes has been increasing
`worldwide [1], in association with rising obesity. However,
`increased adiposity contributes to only 37-77% of incident
`type 2 diabetes, depending on the population studied [2,3] .
`Thus, identification of new and potentially modifiable risk
`factors is needed to inform strategies for prevention.
`Multiple cross-sectional studies have shown type 2 diabetes
`is inversely associated with skeletal muscle mass [4,5] and
`strength [6,7] . Muscle strength is also inversely associated
`with the development of insulin resistance [8] . However, there
`are few longitudinal studies that have investigated the role of
`muscle mass and strength in the development of type 2
`diabetes. Only three longitudinal studies have investigated
`muscle strength and incident type 2 diabetes, and they showed
`conflicting results which may due to methodological limita(cid:173)
`tions from an inability to identify undiagnosed cases [9--11] .
`Furthermore, only one previous study has investigated muscle
`mass and incident type 2 diabetes, which found that changes in
`muscle mass did not predict incident type 2 diabetes [12].
`However, that study relied on bioelectrical impedance analysis
`instead of gold standard measures such as dual-energy X-ray
`absorptiometry (DXA) [12] . The association between reduced
`skeletal muscle mass or strength and incident type 2 diabetes is
`therefore unclear.
`Testosterone is a determinant of muscle mass and strength
`[13], and low testosterone has also been associated with type 2
`diabetes in men [14] . Therefore low testosterone may mediate
`the association between skeletal muscle dysfunction and type 2
`diabetes in men. Similarly, inflammation predicts decline in
`skeletal muscle mass and strength [15], while also predicting
`incident type 2 diabetes [16] . Hence inflammation may
`also mediate any association between skeletal muscle and
`incident type 2 diabetes. These mechanisms have not been
`previously explored.
`We therefore aimed to investigate the association between
`measures of skeletal muscle mass and strength with incident
`type 2 diabetes in a prospective community-dwelling cohort
`of men, and whether testosterone or inflammation mediates
`that association.
`
`2.
`
`Methods
`
`2.1.
`
`Cohort participants
`
`The Men Androgen Inflammation Lifestyle Environment and
`Stress (MAILES) study is a longitudinal cohort of community(cid:173)
`dwelling men, and has been described previously [17] . In brief,
`MAILES consists of two concurrent prospective cohorts:
`the Florey Adelaide Male Ageing Study (FAMAS) [18] and the
`age-matched men from the North West Adelaide Health
`Study (NWAHS) [19] . The two cohorts are largely representa(cid:173)
`tive of the male population of South Australia, and used the
`same methodology for random population sampling.
`Detailed demographic, comorbidity (doctor diagnosed
`diabetes, cardiovascular disease), hand dominance, and risk
`factor data (smoking, physical activity) were collected by self-
`
`completed questionnaire. Biomedical assessments at baseline
`and follow-up were conducted in 2 hospital-based clinics using
`standardised and reproducible study protocols, including grip
`strength and blood pressure measurement, anthropometry, and
`fasting blood samples (lipids, glucose, glycated hemoglobin,
`testosterone and inflammatory markers). Hypertension was
`defined as any of: self-report, a clinically measured systolic
`blood pressure of 2:140 mmHg (mean of two or three readings),
`diastolic blood pressure of 2:90 mmHg (mean of two or three
`readings), or use of anti-hypertensive medication. Metabolic
`syndrome was defined by International Diabetes Federation
`criteria. Mild (walking), moderate, and vigorous physical activity
`levels were determined by the Australian National Health Survey
`Physical Activity instrument and converted into metabolic
`equivalents (METs). In the NWAHS, depression was defined as a
`score of:C:21 on the Center for Epidemiological Studies Depression
`Scale, whereas in FAMAS, depression was defined as a score of
`2:12 on the Beck Depression Inventory. Cardiovascular disease
`was defined as a self-report of doctor-diagnosed myocardial
`infarction, angina, stroke or transient ischemic attack.
`Approval for MAILES was obtained from the Human
`Research Ethics Committees of the North West Adelaide Health
`Service and the Royal Adelaide Hospital. All participants gave
`written informed consent. Baseline data were obtained in 2002-
`2006 and follow-up data in 2007-2010. Fig. 1 shows the
`participant flowchart. Almost all (96%) participants were born
`in Australia or Western Europe (including the UK/Ireland).
`
`2.2.
`
`Skeletal muscle measures
`
`Grip strength was measured with a Jamar analog hand dyna(cid:173)
`mometer in the NW AHS cohort (Lafayette Instrument Company,
`Lafayette, IN) or a Smedley analog hand dynamometer (Stoel ting
`Corporation, Wood Dale, IL) in the F AMAS cohort [17] . To account
`for any systematic differences in the type of hand dynamometer
`used, we included cohort as a covariate in adjusted statistical
`analyses. Main analyses used the mean of three measurements
`in the dominant hand. Sensitivity analyses used the maximum
`(i.e. peak) measurement recorded in either hand. We also
`undertook sensitivity analyses to account for: 1) bodily (e.g.
`hand) pain, as determined by the bodily pain scale of the SF-36; 2)
`self-reported current smoking at baseline; and 3) use of systemic
`corticosteroids at follow-up.
`Body composition was measured at baseline in a sub-set of
`the cohort (n = 1181): NWAHS participants aged 2:50 years [20],
`and all FAMAS participants [18] . Whole-body, arms, and legs
`lean mass and fat mass were measured by DXA using default
`settings on either a pencil-beam (DPX +, Lunar software v4• 7 e) or
`fan-beam (Prodigy DF+ 14759, Encore software v9·15) densi(cid:173)
`tometer. Both machines were from GE Lunar (Madison, WI) and
`provided similar results [21] . Appendicular skeletal muscle
`mass (ASM) was calculated by summing the lean mass of
`arms and legs. ASM and fat mass were normalised for height by
`dividing by height squared to generate ASM index (ASMI), and
`fat mass index (FM!), respectively [20,22] .
`Arm muscle quality (grip strength corrected for arm lean
`mass) was calculated by dividing the sum of the grip strength
`in both hands by arms lean mass [23,24] . Sensitivity analyses
`were also undertaken using arm muscle quality calculated
`with peak grip strength.
`
`Sanofi Exhibit 2179.002
`Mylan v. Sanofi
`IPR2018-01676
`
`

`

`MET ABO LI S M C LINI C AL A N D E X PERIMENT A L 65 (2016) 883-892
`
`885
`
`MAILES participants at baseline
`Total n=2563; FAMAS n=1195 plus NWAHS n=1368)
`
`MAILES participants with baseline diabetes data
`(n=2490; 97.2%)
`
`,__ __ Exclude baseline diabetes
`(n=379)
`
`MAILES participants without diabetes at baseline
`(n=2111)
`
`MAILES participants at 5-year follow-up
`(n=1680; 79.6%)
`
`Complete grip strength data
`(n=1632; 77.3%)
`
`Fig. 1 - Flow of MAILES participants throughout the study. MAILES, Men Androgen Inflammation Lifestyle Environment and
`Stress study; FAMAS, Florey Adelaide Male Aging Study; NWAHS, North West Adelaide Health Study.
`
`2.3.
`
`Type 2 diabetes
`
`Type 2 diabetes was defined as: previous doctor diagnosis,
`diabetes medication use, fasting plasma glucose (FPG)
`2:7.0 mmol/L (2:126 mg/di), or glycated haemoglobin (HbAlc)
`2:6.5% (48 mmol/mol). Diabetes medication use was identified
`by prescription of Anatomical Therapeutic Chemical classifi(cid:173)
`cation A10 drug(s) from the Pharmaceutical Benefits Scheme
`of Australia. FPG was quantified using an automated chem(cid:173)
`istry analyzer system (Olympus AU5400; Olympus, Tokyo,
`Japan). HbAlc was measured by high-performance liquid
`chromatography using a spherical cation exchange gel (CV,
`2% at 6% of total haemoglobin).
`
`2.4.
`
`Testosterone and inflammatory markers
`
`Serum total testosterone was measured by validated stable(cid:173)
`isotope dilution liquid chromatography-tandem mass spec(cid:173)
`trometry (API-5000, Applied Biosystems/MDS SCIEX, Ontario,
`Canada) [25]. All measurements were undertaken on morning
`samples after an overnight fast; further detail is available in
`previous publications [17,26] . Serum interleukin-6 (IL-6) and
`serum tumour necrosis factor-alpha (TNF-a) were measured by
`enzyme-linked immunosorbent assay (ELISA) (R&D Systems,
`Minneapolis, MN) [17].
`
`2.5.
`
`Statistical analysis
`
`Statistical analysis was using SPSS version 20 (SPSS, IL) and R
`version 3.2.1 (R Foundation for Statistical Computing, Vienna,
`Austria). Differences in baseline characteristics between groups
`were determined using t-tests (numerical variables) or x2-tests
`(categorical variables). The unadjusted and adjusted associa(cid:173)
`tions between skeletal muscle measures and incident type 2
`diabetes, and any interaction effects, were determined using
`binary logistic regression models. Collinearity was assessed
`using the variance inflation factor (VIF). All VIF values in this
`study were less than 3.
`We used binary logistic regression models further adjusted
`for testosterone and inflammatory markers to determine if
`additional adjustment for these variables attenuated the
`
`association between grip strength (or arm muscle quality)
`and incident type 2 diabetes. To formally quantify any
`mediating effect of testosterone or inflammatory markers,
`we undertook mediation analysis using the R mediation
`package. For testosterone, we also undertook a sensitivity
`analysis that excluded participants taking medications
`known to affect serum testosterone; major health problems
`including prostate cancer, orchidectomy, and primary tes(cid:173)
`ticular disease; and outlier values including total testos(cid:173)
`terone >40 nmol/L, luteinising hormone > 12 U/L, follicular
`stimulating hormone >8 U/L, and sex hormone-binding
`globulin >100 nmol/L [26] .
`To quantify the proportion of incident type 2 diabetes
`cases that could be prevented if grip strength was to
`be increased, we calculated the population attributable
`fraction (PAF) using the R attribrisk package. There is no
`widely accepted cut-off value for low grip strength, and
`resistance training increases muscle strength by 25-30%
`after 3 months [27] . Hence we calculated PAFs based on
`five potential "intervention targets", as follows: 1) all
`participants increased grip strength by 25%, 2) all partici(cid:173)
`pants increased grip strength by 10%, 3) participants
`increased grip strength to our cohort's mean (47.4 kg),
`4) participants increased grip strength to one standard
`deviation above the cohort mean (57.0 kg) and 5) partici(cid:173)
`pants increased grip strength to one standard deviation
`below the cohort mean (37.2 kg) .
`There were <2% missing data for all confounders, except
`physical activity, which had 6.4% missing. Data for missing
`confounders, but not the primary predictor or outcome
`variable, were imputed as either the median (numerical
`variables) or the modal category (categorical variables). Data
`were not imputed for serum total testosterone, IL-6, or TNF-a.
`Sensitivity analyses using only the complete, non-imputed,
`dataset (n = 1493) were also conducted.
`
`2.6.
`
`Role of the funding source
`
`The funders had no role in study design, data collection
`and analysis, decision to publish, or preparation of the
`manuscript.
`
`Sanofi Exhibit 2179.003
`Mylan v. Sanofi
`IPR2018-01676
`
`

`

`886
`
`MET A B O LI S M C LINIC A L A ND E X PE R IME N T A L 65 (2016) 883-892
`
`Results
`3.
`Diabetes data were available inn = 2490 men (Fig. 1). Excluding
`n = 379 men with diabetes at baseline, n = 2111 were included in
`this study, of which follow-up data were available inn= 1680
`(79.6%) men. Grip strength data was available inn= 1632 (77.3%
`of then = 2111 included in this study). Incident type 2 diabetes
`occurred in 146 men (8.9%) over a median follow-up of 4.95 (!QR
`4.35-5.00) years. Baseline DXA measurements were available in
`a sub-sample of 1181 men (414 from NWAHS and 767 from
`FAMAS). Table 1 shows the baseline characteristics of the cohort,
`overall as well as grouped by whether the men developed
`incident type 2 diabetes or not. Compared to those that did not
`develop type 2 diabetes ('never diabetes') during follow-up, the
`participants that developed incident type 2 diabetes were at
`baseline: older, had higher body mass index (BM!), waist
`circumference (WC) and serum IL-6, and had higher prevalence
`of impaired FPG, family history of diabetes, hypertension,
`
`dyslipidaemia, metabolic syndrome and self-reported cardio(cid:173)
`vascular disease. They had lower annual income, education
`level, serum total testosterone and grip strength. In the DXA
`sub-study, compared to the participants that did not develop
`type 2 diabetes, participants that developed type 2 diabetes
`had higher total fat mass and fat mass index, and had lower
`lean mass percentage, and arm muscle quality. The Pearson
`correlation coefficient between grip strength and arm muscle
`quality was 0.68.
`Table 2 shows the associations of grip strength, arm muscle
`quality, and lean mass measures with incident type 2 diabetes.
`In both unadjusted and adjusted logistic regression models, grip
`strength and arm muscle quality were significantly inversely
`associated with incident type 2 diabetes. In unadjusted models,
`lean mass percentage was inversely associated with incident
`type 2 diabetes, but these associations did not persist after
`adjustment. In models that were mutually adjusted for either
`grip strength or whole-body lean mass (Model 3), grip strength
`
`Table 1 - Baseline characteristics of the cohort.
`Overall
`(n = 1632)
`
`Never diabetes
`(n = 1486)
`
`Incident type 2 diabetes
`(n = 146)
`
`Age (years)
`Annual income
`<$40,000
`$40,000-$79,999
`:>$80,000
`Education
`High school or lower
`Trade
`Diploma/Certificate
`Bachelor or higher
`Physical activity :>600 MET min/week
`FPG
`<5.5 mmol/L
`5.5--<>.1 mmol/L
`6.1--<>.9 mmol/L
`Family history of diabetes
`Current smoking
`BM! (kg/m2
`)
`Waist circumference (cm)
`Hypertension
`HDL <1.0 mmol/L
`Triglycerides > 1.7 mmol/L
`Metabolic syndrome
`Depression
`Cardiovascular disease
`Total testosterone(nmol/L)
`Serum IL-6 (pg/mL)
`Serum TNF-cx (pg/mL)
`Grip strength (kg)
`DXA sub-study
`Total lean mass (kg)
`Lean mass percentage (%)
`ASM (kg)
`ASMI (kg/m2
`)
`Total fat mass (kg)
`Fat mass index (kg/m2
`)
`Arm muscle quality (kg/kg)
`
`54.1 ± 11.4
`
`53.7 ± 11.4
`
`57.7 ± 10.9
`
`42.5 (694)
`38.8 (633)
`18.7 (305)
`
`31.4 (513)
`30.5 (498)
`24.5 (400)
`13.5 (221)
`41.3 (674)
`
`88.1 (1437)
`9.7 (158)
`2.3 (37)
`28.2 (461)
`19.4 (316)
`28.0 ± 4.2
`99.5 ± 11.2
`49.5 (808)
`13.4 (218)
`34.6 (564)
`29.5 (481)
`8.2 (134)
`7.2 (117)
`17.3 ± 5.7
`2.0 ± 1.8
`1.9 ± 2.7
`47.1 ± 9.9
`(n = 1181)
`58.1 ± 7.7
`69.0 ± 6.6
`26.3 ± 3.8
`8.6 ± 1.1
`23.5 ± 8.3
`8.0 ± 2.8
`12.4 ± 2.2
`
`41.1 (610)
`39.5 (587)
`19.4 (289)
`
`30.1 (447)
`30.8 (457)
`24.9 (370)
`14.3 (212)
`41.3 (614)
`
`90.8 (1349)
`7.9 (117)
`1.3 (20)
`27.3 (406)
`19.7 (293)
`27.9 ± 4.1
`99.0 ± 11.0
`47.6 (707)
`12.7 (189)
`33.4 (497)
`27.0 (401)
`7.9 (117)
`6.7 (100)
`17.6 ± 5.7
`1.9 ± 1.7
`1.9 ± 2.8
`47.4 ± 9.9
`(n = 1062)
`58.0 ± 7.7
`69.3 ± 6.6
`26.3 ± 3.8
`8.6 ± 1.0
`23.1 ± 8.2
`7.6 ± 2.7
`12.5 ± 2.2
`
`57.5 (84)
`31.5 (46)
`11.0 (16)
`
`45.2 (66)
`28.1 (41)
`20.5 (30)
`6.2 (9)
`41.1 (60)
`
`60.3 (88)
`28.1 (41)
`11.6 (17)
`37.7 (55)
`15.8 (23)
`29.9 ± 4.8
`104.8 ± 12.3
`69.2 (101)
`19.9 (29)
`45.9 (67)
`54.8 (80)
`11.6 (17)
`11.6 (17)
`15.0 ± 5.7
`2.3 ± 2.6
`1.7 ± 0.9
`44.6 ± 9.4
`(n = 119)
`58.7 ± 7.5
`66.6 ± 5.5
`26.2 ± 3.8
`8.7 ± 1.0
`27.0 ± 8.1
`9.0 ± 2.7
`11.5 ± 2.0
`
`p-value
`
`p < 0.001
`p < 0.001
`
`p = 0.001
`
`p = 0.958
`p < 0.001
`
`p = 0.008
`p = 0.247
`p < 0.001
`p < 0.001
`p < 0.001
`p = 0.015
`p = 0.003
`p < 0.001
`p = 0.113
`p = 0.028
`p < 0.001
`p = 0.025
`p = 0.494
`p = 0.001
`
`p = 0.339
`p < 0.001
`p = 0.825
`p = 0.200
`p < 0.001
`p < 0.001
`p < 0.001
`
`Data are presented as % (n), or mean ± standard deviation. P-values (t-test or x2-test) compare the incident diabetes group with the never
`diabetes group. MET, metabolic equivalents; FPG, fasting plasma glucose; BM!, body mass index; HDL, high density lipoprotein; DXA, dual(cid:173)
`energy X-ray absorptiometry; ASM, appendicular skeletal muscle mass; ASMI, appendicular skeletal muscle mass index.
`
`Sanofi Exhibit 2179.004
`Mylan v. Sanofi
`IPR2018-01676
`
`

`

`MET ABO LI S M C LINI C AL A N D E X PERIMENT A L 65 (2016) 883-892
`
`887
`
`Table 2 - Odds ratios (OR) and 95% confidence intervals (CI) for incident type 2 diabetes associated with grip strength, arm
`muscle quality, and lean mass measures.
`n
`
`Model 3
`
`Unadjusted
`
`Model 2
`
`Grip strength
`Arm muscle quality
`Whole-body lean mass
`Lean mass percentage
`ASM
`ASMI
`
`1632
`1172
`1181
`1181
`1181
`1181
`
`OR (95%CI)
`
`0.87 (0.80-0.95)
`0.81 (0.74-0.89)
`1.06 (0.94-1.21)
`0.76 (0.67-0.87)
`0.97 (0.76-1.25)
`1.14 (0.94-1.38)
`
`OR (95%CI)
`
`0.87 (0.78-0.97)
`0.85 (0.75-0.95)
`0.96 (0.82-1.13)
`0.91 (0.77-1.12)
`0.80 (0.57-1.11)
`0.97 (0.76-1.24)
`
`OR (95%CI)
`
`0.85 (0.74-0.98)
`0.84 (0.75-0.94)
`1.03 (0.87-1.24)
`0.96 (0.80-1.20)
`0.93 (0.64-1.34)
`1.08 (0.83-1.39)
`
`Odds ratios (95% confidence intervals) for incident diabetes are per 5 kg unit increases in grip strength, lean mass and ASM, per 1 kg/kg unit
`increases in arm muscle quality, per 5% unit increases for lean mass percentage, and per 1 kg/m2 increases in ASML Significant associations
`highlighted in bold. ASM, appendicular skeletal muscle mass; ASMI, appendicular skeletal muscle mass index. Model 2: adjusted for age,
`income, cohort, waist circumference, fasting plasma glucose, physical activity, hypertension, triglycerides and family history of diabetes.
`Model 3: Model 2 plus mutual adjustment for either whole-body lean mass or grip strength.
`
`and arm muscle quality were significantly inversely associated
`with incident type 2 diabetes independent of whole-body
`lean mass, but the lean mass measures were not signifi(cid:173)
`cantly associated with incident type 2 diabetes independent of
`grip strength.
`Table 3 shows the proportion mediated by testosterone, IL-6,
`and TNF-a of the associations of grip strength and arm muscle
`quality with incident type 2 diabetes. Further adjustment for
`serum total testosterone slightly attenuated the odds ratios for
`grip strength. The association between grip strength and
`incident type 2 diabetes was reduced to borderline significance
`(p = 0.051). The association between arm muscle quality and
`incident type 2 diabetes remained significant after testosterone
`adjustment. Further adjustment for serum IL-6 and TNF-a also
`did not attenuate any of the associations. In mediation analysis,
`testosterone, IL-6, and TNF-a did not significantly mediate the
`association between grip strength or arm muscle quality and
`incident type 2 diabetes.
`Table 4 shows the PAFs of incidenttype 2 diabetes due to low
`grip strength. The PAFs ranged from 3% (95% CI: 0.1-6%) to 29%
`(95% CI: 8-49%), depending on the intervention target chosen.
`Stratified analyses are shown in Figs. 2 and 3. Fig. 2 shows
`associations of grip strength with incident type 2 diabetes,
`stratified by age, FPG, cohort, general adiposity, central adiposity,
`family history of diabetes, and serum total testosterone. The
`interaction effects were not significant for age, baseline FPG,
`cohort, family history of diabetes, or serum total testosterone.
`However, there were significant interaction effects for WC
`(p = 0.002) and BM! (p = 0.001). The inverse association of grip
`strength with type 2 diabetes was stronger in men with lower
`central adiposity (OR for WC < 102 cm, 0.80 [95% CI: 0.68-0.94]
`vs. WC 2:102 cm, 0.88 [95% CI: 0.76-1.03]) and lower general
`adiposity (OR for BM! <30 kg/m 2
`, 0.79 [95% CI: 0.69-0.92] vs.
`BM! 2:30 kg/m 2
`, 0.94 [95% CI: 0.80-1.13]).
`Fig. 3 shows associations of arm muscle quality with
`incident type 2 diabetes, stratified by age, FPG, cohort, general
`adiposity, central adiposity, family history of diabetes, serum
`total testosterone, and lean mass per arm. The interaction
`effects were not significant for age, baseline FPG, cohort,
`family history of diabetes, serum total testosterone, or lean
`mass per arm. However, there were significant interaction
`effects for WC (p = 0.001) and BM! (p < 0.001). The inverse
`association of arm muscle quality with type 2 diabetes was
`
`stronger in men with lower central adiposity (OR for WC
`<102 cm, 0.70 [95% CI: 0.59--0.83] vs. WC 2:102 cm, 0.93 [95% CI:
`0.80-1.09]) and lower general adiposity (OR for BM! <30 kg!m2
`,
`0.71 [95% CI: 0.62--0.83] vs. BM! 2:30 kg!m2
`, 1.03 [95% CI: 0.86-1.23]).
`Sensitivity analyses are presented in the online Supple(cid:173)
`mentary Data (Appendix A). Table Al shows the analyses
`using peak grip strength (instead of mean dominant hand grip
`strength) and peak arm muscle quality (instead of overall arm
`muscle quality), as well as the analyses additionally adjusting
`for SF-36 bodily pain scale, current smoking, and systemic
`corticosteroid use at follow-up. Table A2 shows sensitivity
`analyses limited to the non-imputed complete dataset. The
`results of these sensitivity analyses were similar to the main
`analyses . The sensitivity analysis for testosterone mediation
`also yielded similar results to the main analysis (proportion
`mediated of grip strength 4% [95% CI: -70 to 71%], p = 0.49;
`proportion mediated of arm muscle quality 1 % [95% CI: -30 to
`24%], p = 0.90).
`
`4.
`
`Discussion
`
`In a large prospective community-dwelling sample of men,
`we found that grip strength and arm muscle quality were
`inversely associated with incident type 2 diabetes. Neither
`testosterone nor inflammation significantly mediated these
`associations. These associations, robust to multiple sensitiv(cid:173)
`ity analyses, were consistent across age, impaired fasting
`glucose, cohort, family history, and serum total testosterone
`strata. However, while these associations were especially
`strong in non-obese men, they were not significant in obese
`men. Our findings suggest that measurement of grip strength
`has a particularly important role in identifying non-obese
`men at risk of type 2 diabetes.
`Only three previous studies have investigated the associa(cid:173)
`tion between muscle strength and incident type 2 diabetes
`[9-11] . The largest study of those found inconclusive results:
`the adjusted association between grip strength and incident
`diabetes was not significant in the main analyses, but was
`significant in the sensitivity analyses [11] . However, that study
`relied on self-report to exclude baseline diabetes, with incident
`diabetes defined via self-report, identification on death case
`report forms, or identification on hospitalisation case report
`
`Sanofi Exhibit 2179.005
`Mylan v. Sanofi
`IPR2018-01676
`
`

`

`888
`
`MET A B O LI S M C LINIC A L A ND E X PE R IME N T A L 65 (2016) 883-892
`
`Table 3 - Mediation analyses of associations [Odds ratios (OR) and 95% confidence intervals (CI)] between grip strength and
`arm muscle quality with incident type 2 diabetes.
`
`Exposure
`
`Mediator
`
`Reference model*
`
`Reference model + mediator
`
`Proportion mediated
`(95% CI)
`
`Grip strength
`Grip strength
`Grip strength
`Arm muscle quality
`Arm muscle quality
`Arm muscle quality
`
`Testosterone
`IL-6
`TNF-cx
`Testosterone
`IL-6
`TNF-cx
`
`OR (95%CI)
`
`0.88 (0.78-0.99)
`0.88 (0.78-0.99)
`0.88 (0.78-0.99)
`0.86 (0.76-0.97)
`0.85 (0.76-0.96)
`0.87 (0.77-0.98)
`
`OR (95%CI)
`
`0.89 (0.79-1.001)
`0.88 (0.78-0.99)
`0.88 (0.77-0.98)
`0.87 (0.77-0.98)
`0.86 (0.76-0.96)
`0.86 (0.77-0.98)
`
`2% (-15 to 25%)
`1% (-6 to 9%)
`1% (-6 to 9%)
`0% (-12 to 13%)
`0% (-3 to 5%)
`0% (-10 to 13%)
`
`Odds ratios (95% confidence intervals) for incident diabetes are per 5 kg unit increases in grip strength, and per 1 kg/kg- 1 unit increases in arm
`muscle quality. Significant results are highlighted in bold. IL-6, interleukin-6; TNF-cx, tumour necrosis factor-alpha. 'Reference models for grip
`strength: adjusted for age, income, cohort, waist circumference, fasting plasma glucose, physical activity, hypertension, triglycerides and
`family history of diabetes. Reference models for arm muscle quality: adjusted for age, income, cohort, waist circumference, fasting plasma
`glucose, and family history of diabetes. The odds ratios for the Reference models differ slightly between testosterone, IL-6, and TNF-cx because
`of missing data.
`
`forms. Undiagnosed cases, representing46% of diabetes world(cid:173)
`wide [1], were not identified. Similarly, an older study defined
`type 2 diabetes by self-report and found no association between
`grip strength and incident type 2 diabetes [10] . The lack of
`identification of undiagnosed cases would have biased those
`studies towards the null. In contrast, using our comprehensive
`definition of type 2 diabetes, which included FPG and HbA1c
`criteria to identify undiagnosed cases, we found an inverse
`association that persisted after adjustment. When we replicat(cid:173)
`ed the self-reported methodology of previous studies in our
`own cohort, we too found no significant association after
`adjustment for confounders (data not shown).
`In a small (n = 394) study of Japanese Americans, where
`75 g oral glucose tolerance testing and use of diabetes
`medication defined type 2 diabetes, lower hand grip strength
`predicted incident type 2 diabetes in lean, but not in obese,
`participants [9] . We extend their findings by demonstrating
`in a larger cohort that the association is consistent across
`
`Table 4- Population attributable fractions (PAFs) of
`incident type 2 diabetes from low grip strength,
`according to intervention target.
`25%
`10% Mean grip Mean + 1 SD Mean - 1 SD
`Target
`Target
`target
`target
`target
`
`12%
`PAF 27%
`(13-40%) (5-18%)
`
`13%
`(5-20%)
`
`29%
`(10-47%)
`
`3%
`(1-5%)
`
`PAFs are shown as% (95% confidence interval). SD, standard deviation.
`25% Target: all participants increase grip strength by 25%.
`10% Target: all participants increase grip strength by 10%.
`Mean grip target: participants with grip strength below the mean
`(47.4 kg) have an intervention target of 47.4 kg; participants with
`grip strength above the mean remain the same.
`Mean + 1 SD target: participants with grip strength below the mean + 1
`standard deviation (57.0 kg) have an intervention target of 57.0 kg;
`participants with grip strength above 57.0 kg remain the same.
`Mean - 1 SD target: participants with grip strength below the mean - 1
`standard deviation (37.2 kg) have an intervention target of 37.2 kg;
`participants with grip strength above 37.2 kg remain the same.
`
`other strata (such as age, baseline FPG, family history, and
`serum testosterone), and is also independent of many
`confounding variables.
`Cross-sectional studies have established an inverse
`association between muscle mass and type 2 diabetes [4,5] .
`In contrast, the only previous longitudinal study found
`changes in lean mass did not predict incident type 2 diabetes
`[12] . However, that previous longitudinal study relied on
`bioelectrical impedance analysis and not gold standard
`methods such as DXA. Using multiple indices of skeletal
`muscle mass derived from DEXA, we have confirmed that
`skeletal muscle mass does not predict incident type 2 diabetes.
`The discrepancy between cross-sectional and longitudinal
`studies may be because low muscle mass is an effect of type 2
`diabetes, instead oflow muscle mass predicting type 2 diabetes
`development [28] .
`We found testosterone does not mediate the association
`between muscle strength and incident type 2 diabetes . This
`contrasts to existing knowledge that testosterone increases
`muscle mass and strength [29] and activates glucose
`metabolism-related signaling within skeletal muscle [30]. Our
`results suggest testosterone's effect on cardiometabolic risk is
`separate from its effect on skeletal muscle. Furthermore,
`despite previous observations that inflammation, particularly
`IL-6 and TNF-a, contributes to both decreased grip strength as
`well as incident type 2 diabetes [15,161, we found that IL-6 and
`TNF-a also do not mediate the association between grip
`strength and incident type 2 diabetes. Hence, much of the
`association between muscle strength and incident type 2
`diabetes remains unexplained, and further studies are required
`to elucidate other factors that may contribute.
`A major strength of our study was the strong methodological
`basis. Measurement of grip strength by hand dynamometer is a
`valid and reliable measure of muscle strength [31], while DXA is
`an accurate measure of muscle mass [32] . Testosterone was
`measured using validated li