Endogenous Testosterone Concentrations and Muscle Mass, Strength and Performance in Women, a Systematic Review of Observational Studies

Sasha Taylor; Rakibul M. Islam; Robin J. Bell; Chandima Hemachandra; Susan R. Davis

Clin Endocrinol. 2023;98(4):587-602.

Abstract and Introduction

Abstract

Objective: To explore the associations between endogenous testosterone blood concentrations and muscle mass, strength and performance in community dwelling women.

Design, Patients and Measurements: Online databases, including Ovid MEDLINE, EMBASE and Web of Science, were searched for observational studies, with at least 100 female participants, reporting associations between endogenous testosterone blood concentrations and muscle mass, strength and performance. The findings were synthesized in a narrative review. Heterogeneity in study design and analysis precluded a meta-analysis.

Results: Of the 36 articles retrieved for full-text review, 10 met the inclusion criteria. Eight studies were cross-sectional, 1 longitudinal and 1 provided both cross-sectional and longitudinal data. Testosterone was measured by liquid chromatography-tandem mass spectrometry in two studies and by immunoassay in 8. An association between total testosterone and muscle mass, strength or performance in women was not found. The studies of calculated free or bioavailable testosterone and lean muscle mass reported a positive association, but no association was reported for muscle strength or performance. Each included study was limited by a high risk of bias in at least one assessed domain.

Conclusions: This review does not support an association between testosterone and muscle mass, strength or performance in women. This, together with the reported associations between free or bioavailable testosterone and muscle mass should be interpreted cautiously due to the predominant use of immunoassay and the inaccuracy of calculated variables. Additionally, biological significance of nonprotein bound testosterone has not been established. Further studies examining the relationship between precisely measured testosterone and muscle mass and function in women are required.

Introduction

The maintenance of muscle strength in women as they age is important, especially after menopause. Loss of muscle mass and strength are well-described components of sarcopenia,^[1,2] and contribute to poor physical functioning, and increase the risk of falls and fractures.^[1,3,4] Muscle mass also influences metabolic health, with decreased skeletal muscle mass predisposing to insulin resistance and development of type 2 diabetes mellitus.^[5] These consequences of diminished muscle mass and strength have broad physical and psychological health implications, and may ultimately result in loss of independence and the need for formalized care.^[1,6]

While exercise and weight training are known to help maintain muscle health in women as they age,^[7] sex hormones may also play an important role. Testosterone has anabolic actions in muscle, directly through androgen receptors and indirectly via conversion to estradiol.^[8] However, the association between endogenous testosterone blood concentrations, including total testosterone, estimated nonprotein bound (hereafter referred to as 'free') testosterone or non-sex hormone binding globulin (SHBG) bound ('bioavailable') testosterone and skeletal muscle health in women is less well characterized.^[9] Evidence for a contribution of testosterone to skeletal muscle health in women could potentially justify further investigation into the use of exogenous testosterone as muscle conserving therapy.

The objective of this systematic review was to examine the published literature to determine whether there is evidence for an association between endogenous testosterone blood concentrations and muscle mass, as well as measures of muscle strength and performance, in women.

1 2 3 4

Next Section

Abstract and Introduction
Methods
Results
Discussion
References

Table 1. Association between total testosterone and muscle mass, strength and performance (cross-sectional studies)
Author, year, country	Study population^a	Sample size	Age range or mean age (sd)	Exposure measurement	Outcome	Statistical analyses	Variables adjusted	Findings	Association with testosterone
Muscle mass^b
Alexander et al. (2021)¹⁷ US	Community based sample drawn from NHANES (US health survey) Not taking anabolic steroids	716	30 (6) (18−40)	LCMS/MS	Lean mass index - Total - Upper body - Lower body	Linear regression	Age, ethnicity, past female hormone use, oestrogen and SHBG levels, body fat percentage, physical activity, alcohol, T examination timing (morning, afternoon or evening), dietary protein, vitamin C, D, magnesium	Lean mass index Total: β: 0.05 (95% CI: −0.04 to 0.15); p = .237 Upper body: β: 0.06 (95% CI: −0.03 to 0.16); p = .171 Lower body: β: 0.05 (95% CI: −0.04 to 0.15); p = .228	No association
Bann et al. (2015)¹⁶ UK	Community based sample drawn from the UK MRC National Survey of Health and Development (NSHD)	436	60−64	LCMS/MS	Appendicular lean mass	Linear regression	Fat-free mass index	Mean percentage differences in appendicular lean mass (95% CI) at 60–64 years per 1 standard deviation increase in serum T concentration at 60–64 years: 0.27 (95% CI: −0.68 to 1.22); p = .58	No association
Haren et al. (2007)¹⁸ US	Community based sample drawn from the African American health project	244	56.9 (4.4)	Immunoassay	Total lean mass (corrected for height) Appendicular skeletal mass: total and percent lean	Pearson's correlation coefficients (r) and associated p values from linear regression analysis	No	Total lean mass: r: .317, p = .0007 Appendicular skeletal mass - total: r: .316, p = .0007 percent lean: r: −.19, p = .041	Positive association with total lean mass and appendicular skeletal mass Inverse association with percent lean mass in appendicular skeleton
Rariy et al. (2011)¹⁹ US	Community based sample drawn from cardiovascular health study not taking corticosteroids	232	75.6 (4.6)	Immunoassay	Total LBM	Linear regression β coefficients are for T term in the model	Age, race, activity levels	All women: β: −0.001; p = .92 Excluding oestrogen users: β: −0.006; p = .51 Excluding oestrogen users and adjusting for estradiol use: β: −0.012; p = .22	No association
Sipila et al. (2006)¹³ Finland	Community based sample drawn from evergreen project study	187	75	Immunoassay	LBM (by Bioelectrical impedance)	One-way analysis of variance	No	Low (<0.867 nmol/L): 44 ± 4 kg Middle (0.867–2.045 nmol/L): 45 ± 5 kg High (>2.045 nmol/L): 46 ± 4 kg p Value not reported as not statistically significant	No association
Van Geel et al. (2009)¹⁴ Netherlands	Community based sample drawn from region of Netherlands	329	66.9 (6.51)	Immunoassay	LBM	Partial correlation coefficients	Age	Partial r: .016 p Value not reported as not statistically significant	No association
Muscle strength
Alexander et al. (2021)¹⁷ US	Community based sample drawn from NHANES (US health survey) Not taking anabolic steroids	716	30 (6) (18−40)	LCMS/MS	Combined HGS	Linear regression	Age, height, ethnicity, past female hormone use, oestrogen and SHBG levels, body fat percentage, physical activity, alcohol, T examination timing (morning, afternoon or evening), dietary protein, vitamin C, D, magnesium intake	Combined HGS: β: 0.01 (95% CI:−0.11 to 0.12); p = .926	No association
Carcaillon et al. (2012)¹⁵ Spain	Community based sample drawn from the Toledo study for healthy aging not taking hormone therapy	735	≥65	Immunoassay	HGS (adjusted for sex and BMI)	Logistic regression for association between HGS and T Levels of T were reported by geometric mean calculated by linear regression	Age	Geometric mean 0.35 (IQR: 0.22−0.58); p = .059	No association
Du et al. (2021)²⁰ China	Community based sample drawn from community health services in Shanghai, China Not taking hormone therapy or glucocorticosteroids	478	66.77 (sd not reported)	Immunoassay	HGS	Linear regression	Age	β: −1.224 (95% CI: −4.6854 to −2.236), p = .484	No association
Haren et al. (2007)¹⁸ US	Community based sample drawn from the African American Health project	244	56.9 (4.4)	Immunoassay	Maximum HGS Lower limb strength (peak torque corrected for body weight) Ankle flexion/ extension Knee flexion/extension Hip flexion/extension	Pearson's correlation coefficients (r) and associated p values from linear regression analysis	No	Maximum HGS: r: −.002, p = .979 Lower limb strength: Peak torque/body weight (120°) Ankle extensors: r: −.18, p = .062 Ankle flexors: r: −.069, p = .484 Knee extensors: r: −.21, p = .031 Knee flexors: r: −.17, p = .079 Hip extensors: r: −.10, p = .335 Hip flexors: r: −.147, p = .155	Inverse association with knee extension peak torque No association with other lower limb strength measures or maximum HGS
Sipila et al. (2006)¹³ Finland	Community based sample drawn from evergreen project study	187	75	Immunoassay	Isometric muscle strength for elbow flexion and knee extension, combined into one score for muscle strength using principal component analysis	One-way analysis of variance	Height	N/A Note: Figure presented with arbitrary units; p value not reported as not statistically significant	No association
Van Geel et al. (2009)¹⁴ Netherlands	Community based sample drawn from region of Netherlands	329	66.9 (6.51)	Immunoassay	Maximum quadriceps extension strength Maximum HGS	Partial correlation coefficients	Age	Partial r Maximum quadriceps extension strength 0.104 Maximum HGS −0.045 p Value not reported as not statistically significant	No association
Muscle performance
Carcaillon et al. (2012)¹⁵ Spain	Community based sample; drawn from the Toledo study for healthy aging not taking hormone therapy	735	≥65	Immunoassay	3 m gait speed (adjusted for sex and height)	Logistic regression for association between 3 m gait speed and T Serum T reported by geometric mean calculated by linear regression	Age	Geometric mean 0.36 (IQR: 0.20−0.61); p = .076	No association
Du et al. (2021)²⁰ China	Community based sample drawn from community health services in Shanghai, China Not taking hormone therapy or glucocorticosteroids	478	66.77 (sd not reported)	Immunoassay	6 m gait speed	Linear regression	Age	β: −1.198 (95% CI: −4.389 to −1.993), p = .458	No association
Haren et al. (2007)¹⁸ US	Community based sample drawn from the African American health project	244	56.9 (4.4)	Immunoassay	Sit-stand test 6 m walk time 6 min walk distance	Pearson's correlation coefficients (r) and associated p values from linear regression analysis	No	Sit-stand test: r: .08, p = .451 6 m walk time r: −.235, p = .027 6 min walk distance r: −.256, p = .016	No association with sit-stand test Positive association with 6 m walk time Inverse association with distance walked in 6 min
Van Geel et al. (2009)¹⁴ Netherlands	Community based sample from region of Netherlands	329	66.9 (6.51)	Immunoassay	Timed up and go test	Partial correlation coefficients	Age	Partial r: −.051 p Value not reported because was not statistically significant	No association

Abbreviations: BMI, body mass index; HGS, hand grip strength; LBM, lean body mass; LCMS/MS, liquid chromatography with tandem mass spectrometry; SHBG, sex hormone binding globulin; T, testosterone.
^aFor this analysis, medications of interest in relation to the study population include: oral contraceptive pill, antiandrogens, glucocorticoids, androgen therapy, aromatase inhibitors, menopausal hormone therapy. Where the authors have identified and addressed any of these medications, this has been reported in the table.
^bMuscle mass assessed by duel-energy X-ray absorptiometry (DXA) (or bioelectrical impedance if specified in table).

Table 2. Association between total testosterone and muscle mass ^a (longitudinal studies)
Author, year, country	Study population^b	Sample size	Age range or mean age (sd)	Exposure measurement	Outcome	Statistical analyses	Variables adjusted	Findings	Association with testosterone
Bann et al. (2015)¹⁶ UK	Community based sample, drawn from the UK MRC National Survey of Health and Development (NSHD)	436	60−64	LCMS/MS Measured at age 53, and age 60−64	Appendicular lean mass Measured at age 60−64	Linear regression Associations between T at age 53, and change (Δ) in T from age 53 to 60−64, in relation to lean mass at age 60−64 were examined	All models were adjusted for fat-free mass index Δ T: adjusted for T concentration at 53 years Δ T, adjusted: highest household occupational class, smoking status, self-rated health, menopausal status at 53 years (pre, peri or post)	Mean percentage differences in appendicular lean mass (95% CI) at 60−64 years per 1 standard deviation increase in serum T concentration at 53 (or change in T): T, age 53: −0.18 (95% CI: −1.20 to 0.84); p = .73 Δ T: 0.43 (95% CI: −0.81 to 1.67); p = .49 Δ T, adjusted: 0.57 (95% CI: −0.71 to 1.85); p = .38	No association
Yuki et al. (2015)²¹ Japan	Community based sample drawn from National Institute for Longevity Sciences- Longitudinal cohort study of aging Not taking hormone replacement therapy	430	At baseline: 55.3 (9.8)	Immunoassay Measured at baseline only	Skeletal muscle mass Measured at 8 year follow-up	Cochran–Mantel–Haenszel statistics for the baseline T levels (trend) and the prevalence of low skeletal muscle mass at follow-up after 8 years Logistic regression for associations of baseline T with risk of low skeletal muscle mass at 8 year follow-up, comparing low level of T with high level of T	Baseline age group, physical activity, nutritional intake, medical history, menopause and smoking history	Prevalence of low skeletal muscle mass (%): Low T: 11.9% Med T: 13.4% High T: 6.9% p-trend = .1695 Logistic regression: OR: 1.32 (0.53−3.27) p-trend = .7816	No association

Abbreviations: LCMS/MS, liquid chromatography with tandem mass spectrometry; T, testosterone.
^aMuscle mass assessed by duel-energy X-ray absorptiometry (DXA).
^bFor this analysis, medications of interest in relation to the study population include: oral contraceptive pill, antiandrogens, glucocorticoids, androgen therapy, aromatase inhibitors, menopausal hormone therapy. Where the authors have identified and addressed any of these medications, this has been reported in the table.

Table 3. Association between free and bioavailable testosterone (calculated) and muscle mass, strength and performance (cross-sectional and longitudinal studies)
Author, year, country	Study population^a	Sample size	Age range or Mean age (sd)	Exposure measurement	Outcome	Statistical analyses	Variables adjusted	Findings	Association with bioavailable/free testosterone
Muscle mass^b
Bann et al. (2015)¹⁶ UK	Community based sample, drawn from the UK MRC National Survey of Health and Development (NSHD)	436	60−64	Free T (total T from LCMS/MS)	Appendicular lean mass	Linear regression	Fat-free mass index	Mean percentage differences in appendicular lean mass (95% CI) at 60−64 years per 1 standard deviation increase in serum free T concentration at 60−64 years: 1.16 (0.18−2.13); p = .02	Positive association
Hwang et al. (2013)²² Taiwan	Community based sample from one town in Taiwan	250	63.3 (8.5)	Free T Bioavailable T (total T from immunoassay)	Skeletal muscle mass	One-way analysis of variance Skeletal muscle mass results divided into quartiles	No	Free T Q1 0.28 ± 0.17 Q2 0.42 ± 0.28 Q3 0.47 ± 0.33 Q4 0.50 ± 0.29 p < .001 Bioavailable T Q1 6.9 ± 4.2 Q2 10.2 ± 6.8 Q3 11.7 ± 8.2 Q4 12.2 ± 6.8 p < .001	Positive association
Van Geel et al. (2009)¹⁴ Netherlands	Community based sample drawn from region of Netherlands	329	66.9 (6.51)	Free T (total T from immunoassay)	Total LBM	Partial correlation coefficients	Age	Partial r: .167, p < .01	Positive association
Longitudinal
Bann et al. (2015)¹⁶ UK	Community based sample, drawn from the UK MRC National Survey of Health and Development (NSHD)	436	60−64	Free T (total T from LCMS/MS) Measured at age 53, and age 60−64	Appendicular lean mass Measured at age 60−64	Linear regression Associations between T at age 53, and change (Δ) in T from age 53 to 60−64, in relation to lean mass at age 60−64 were examined	All models were adjusted for fat-free mass index Δ free T: adjusted for free T concentration at 53 years Δ Free T, adjusted: highest household occupational class, smoking status, self-rated health, menopausal status at 53 years (pre, peri or post)	Mean percentage differences in appendicular lean mass (95% CI) at 60−64 years per 1 standard deviation increase in serum free T concentration at 53 (or change in free T): Free T, age 53: 0.97 (95% CI: −0.06 to 2.00); p = .07 Δ Free T: 0.62 (95% CI: −0.55 to 1.79); p = .30 Δ Free T, adjusted: 0.71 (95% CI: −0.49 to 1.91); p = .24	Positive association between free T at age 53 and appendicular lean mass at age 60−64 years No association between change in free T (from 53 to 60−64 years) and appendicular lean mass at 60−64 years, before and after adjustment
Muscle strength
Hwang et al. (2013)²² Taiwan	Community based sample from one town in Taiwan	250	63.3 (8.5)	Free T Bioavailable T (total T from immunoassay)	HGS	One-way analysis of variance HGS results divided into quartiles	No	Free T Q1 0.40 ± 0.33 Q2 0.38 ± 0.23 Q3 0.43 ± 0.28 Q4 0.45 ± 0.26 p = .574 Bioavailable T Q1 9.5 ± 7.7 Q2 9.3 ± 5.8 Q3 10.6 ± 6.9 Q4 11.1 ± 6.6 p = .479	No association
Van Geel et al. (2009)¹⁴ Netherlands	Community based sample drawn from region of Netherlands	329	66.9 (6.51)	Free T (total T from immunoassay)	Maximum quadriceps extension strength Maximum HGS	Partial correlation coefficients	Age	Maximum quadriceps extension strength Partial r: .178, p < .01; Maximum HGS partial r: −.017 p Value not reported because was not statistically significant	Positive association with maximum quadriceps extension strength No association with maximum HGS
Muscle performance
Hwang et al. (2013)²² Taiwan	Community based sample from one town in Taiwan	250	63.3 (8.5)	Free T Bioavailable T (total T from immunoassay)	6 m walk time	One-way analysis of variance Gait speed results divided into quartiles	No	Free T Q1 0.39 ± 0.26 Q2 0.45 ± 0.28 Q3 0.38 ± 0.24 Q4 0.44 ± 0.38 p = .344 Bioavailable T Q1 9.4 ± 6.3 Q2 10.9 ± 6.6 Q3 9.3 ± 5.9 Q4 11.0 ± 9.4 p = .363	No association
Van Geel et al. (2009)¹⁴ Netherlands	Community based sample drawn from region of Netherlands	329	66.9 (6.51)	Free T (total T from immunoassay)	Timed up and go test	Partial correlation coefficients	Age	Partial r: −.013 p Value not reported because was not statistically significant	No association

Abbreviations: HGS, hand grip strength; LBM, lean body mass; LCMS/MS, liquid chromatography with tandem mass spectrometry; T, testosterone.
^aFor this analysis, medications of interest in relation to the study population include: oral contraceptive pill, antiandrogens, glucocorticoids, androgen therapy, aromatase inhibitors, menopausal hormone therapy. Where the authors have identified and addressed any of these medications, this has been reported in the table.
^bMuscle mass assessed by duel-energy X-ray absorptiometry (DXA).

Table 4. Risk of bias assessment of included studies
Cross-sectional studies—modified Hoy tool^a
Study	Item 1	Item 2	Item 3	Item 4	Item 5	Item 6	Item 7	Item 8	Item 9	Item 10
Total testosterone
Alexander et al.¹⁷	Y	N	Y	Y	Y	N/A	Y	Y	Y	Y
Bann et al.¹⁶	Y	Y	Y	Y	Y	N/A	Y	Y	Y	N
Carcaillon et al.¹⁵	N	Y	Y	N	Y	Y	Y	Y	N	Y
Du et al.²⁰	N	N	N	N	Y	Y	Y	Y	N	Y
Haren et al.¹⁸	N	N	Y	Y	Y	N/A	Y	Y	N	N
Rariy et al.¹⁹	N	Y	Y	Y	Y	N/A	Y	Y	N	Y
Sipila et al.¹³	N	Y	Y	N	Y	N/A	Y	Y	N	N
Van Geel et al.¹⁴	N	Y	Y	Y	Y	N/A	Y	Y	N	Y
Free or bioavailable testosterone
Bann et al.¹⁶	Y	Y	Y	Y	Y	N/A	Y	Y	Y	N
Hwang et al.²²	N	Y	Y	N	Y	N/A	Y	Y	N	N
Van Geel et al.¹⁴	N	Y	Y	Y	Y	N/A	Y	Y	N	Y
Longitudinal studies—modified Joanna Briggs Institute critical appraisal checklist for cohort studies ^b
Study	Item 1	Item 2	Item 3	Item 4	Item 5	Item 6	Item 7	Item 8	Item 9	Item 10	Item 11
Total testosterone
Bann et al.¹⁶	Y	Y	Y	Y	Y	N/A	Y	Y	N	N	Y
Yuki et al.²¹	Y	Y	N	Y	Y	Y	Y	Y	N	N	Y
Free testosterone
Bann et al.¹⁶	Y	Y	Y	Y	Y	N/A	Y	Y	N	N	Y

Abbreviations: N, no; N/A, not applicable; Y, yes.
^aModified Hoy tool (items 1−4 indicate external validity, 5−10 internal validity). Item 1: national representativeness; item 2: sampling frame; item 3: sampling technique; item 4: minimal nonresponse bias; item 5: data collected from subjects; item 6: acceptable case definition used; item 7: valid and reliable study instrument used; item 8: same mode of data collection for all subjects; item 9: exposure measured in valid and reliable way; item 10: relevant confounders controlled/adjusted for.
^bModified Joanna Briggs Institute critical appraisal checklist for cohort studies. Item 1: groups similar and from same population; item 2: exposures measured similarly across groups; item 3: exposure measured validly and reliably; item 4: confounding factors identified; item 6: participants free of outcome at start; item 7: outcomes measured in valid and reliable way; item 8: follow-up appropriate; item 9: follow-up complete, or explained if not; item 10: strategies used to address incomplete follow-up; item 11: appropriate statistical analysis used.