Comparing calculated free testosterone with total testosterone for screening and diagnosing late‐onset hypogonadism in aged males: A cross‐sectional study

Abstract

Objective

The aim of this study is to compare calculated free testosterone (cFT) and total testosterone (T) in predicting late‐onset hypogonadism (LOH) in middle‐aged and elderly males.

Methods

We surveyed a random sample of 608 males between the ages of 45 and 87 years from Shanghai, China. The Aging Male Symptoms (AMS) questionnaire and the Androgen Deficiency in Aging Male (ADAM) questionnaire were completed by the subjects. Testosterone (T), sex hormone‐binding globulin (SHBG), albumin, and other blood biochemical indexes were measured in 332 males. The corresponding cFT was obtained using the Vermeulen formula and the correlations between T and cFT were analyzed by SPSS statistical software.

Results

Among the 332 males who underwent biochemical evaluation, 289 males (87.0%) was positively screened by the ADAM questionnaire and 232 males (69.9%) by the AMS questionnaire. As suggested by linear regression, cFT exhibited a negative correlation with age in both ADAM+ and AMS+ group, whereas T did not appear to have significant correlation with age. Besides, there were statistically significant differences in cFT (P<.001) in the AMS questionnaire.

Conclusions

Calculated free testosterone levels are more reliable than T levels for diagnosing LOH in middle‐aged and elderly males.

1. Introduction

Senile diseases have become essential health concerns arising from substantially growing aging population, particularly in China which has the largest aging population. Late‐onset hypogonadism (LOH) is a typical senile disease which is characterized by symptoms such as deficiency in testosterone levels,1 leading to adverse impact on multiple organ systems.2, 3, 4, 5 However, the etiology of LOH remains controversial due to the lack of evidence whether testosterone deficiency in elderly males is a potential cause.2, 6, 7 The corresponding symptoms of hypogonadism in aging males are highly non‐specific and the prevalence of LOH varies with different diagnostic criteria.8 For instance, Wu et al.9 reported the results of the European male aging study (EMAS) and concluded that the presence of three sexual symptoms (poor morning erection, erectile dysfunction, or low sexual desire) and a morning T level <11 nmol/L can be considered as an evidence___based criteria for the diagnosis of LOH. Liu et al.10 suggested that it requires the presence of at least three sexual symptoms with a T level <13.21 nmol/L and a cFT level <268.89 pmol/L to support the recommendations of the latest practice guidelines.

Our previous study suggested that the corresponding levels of T showed a decreasing trend between the ages of 20 and 35, and an increasing trend appeared between the ages of 50 and 70.11 In this study, we investigated the total testosterone (T), sex hormone‐binding globulin (SHBG), luteinizing hormone (LH), albumin, and obtained calculated free testosterone (cFT) and bioavailable testosterone (cBT) by Vermeulen formula in males over 45 years old.12 Meanwhile, we used the Aging Male Symptoms (AMS) questionnaire and the Androgen Deficiency in Aging Male (ADAM) questionnaire to investigate the subjective symptoms of LOH in elderly males.13, 14, 15 Then, we explored the relationship between clinical manifestations and laboratory results in order to verify whether these indicators mentioned above can be used to diagnose LOH in clinical practice.

2. Materials and Methods

2.1. Patients

We conducted a cross‐sectional and observational study which is based on a population of middle‐aged and elderly males. A total of 793 males over 45 years were interviewed and 608 males were eventually included. The corresponding inclusion criteria were as follows: (1) males over 45 years old; (2) males with no current prescription which is able to affect the hypothalamic–pituitary–gonadal axis (including T, antiandrogen, glucocorticoid, opiate, antiepileptic, antipsychotics, 5‐α reductase inhibitors, and aromatase inhibitors); (3) no history of genital system surgeries; (4) no pituitary, testicular, or adrenal diseases. All participants underwent physical examinations including height, weight, blood pressure, ultrasound scans of the liver, and completed the AMS and ADAM questionnaire. Three hundred thirty‐two participants underwent blood tests for biochemical and hormone measurements and others refused for personal reasons.

2.2. Consent to publish

The Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People's Hospital approved this study and all participants received written and oral information with respect to study consent. The design and implementation of this study was complied with the Helsinki II declaration.

2.3. Sample collection

Blood samples were collected by venipuncture between 8 and 10 am after an overnight fast of at least 8 hours. After centrifugation, serum was harvested and stored at −80°C until being processed.

2.4. Questionnaires and assessment criteria

There were 10 questions involved in the ADAM questionnaires with the answers of “Yes” and “No”. Each patient participated completed the ADAM questionnaires according to their actual physical situations. If question 3 or question 7 or any three or more of the other questions was answered by “Yes”, it was defined as positive, otherwise, it was defined as negative.

The AMS scale was made up of 17 questions to evaluate psychological, somatic, and sexual symptoms in aged males. Each question was to be answered with a score from 1 to 5. A total score of 26 or less was considered to be negative; 27‐36 was considered to be mild; 37 or more was considered to be moderate to severe.

2.5. Hormone measurements

Chemiluminescence was used to measure serum T and LH. Serum SHBG levels were determined by the two‐step enzyme immuno assay (Beckman Coulter Inc, Miami, FL, USA) in which detailed parameters were described previously.11 Testosterone secretion index (TSI) was calculated as T divided by LH. Free testosterone index (FTI) was calculated as T divided by SHBG. Total cholesterol, low‐density lipoprotein (LDL) cholesterol, high‐density lipoprotein (HDL) cholesterol, triglycerides, albumin, hemoglobin, and glucose were measured using an automatic biochemical analyzer (Beckman Coulter Inc). We obtained cFT and cBT using the free and bioavailable testosterone calculator on the website: http://www.issam.ch/freetesto.htm. This calculator was invented according to the Vermeulen formula.12

2.6. Statistical analysis

All data were input into an SPSS database after examination and verification. SPSS 21.0 statistical package (SPSS, Inc., Chicago, IL, USA) was used to analyze the data. The measurement data were all depicted as mean value±standard deviation (χ±δ). The chi‐square (χ²) test was used to determine the association between two categorical variables. Linear regression was carried out to analyze the correlation of T, cFT, and cBT with the factor of age. The two‐sided t test and the one‐way analysis of variance (ANOVA) were used to compare continuous variables complied with normal distribution, whereas Kruskal‐Wallis and Mann‐Whitney U tests were used to analyze those that are skewed distribution. The Pearson correlation analysis was used to determine the linear relationship among several continuous variables. Moreover, the corresponding Receiver Operating Characteristics (ROC) curve was plotted in order to visualize and evaluate the predictive power of cFT for LOH. A P value <.05 was considered to be statistically significant.

3. Results

3.1. Study population

Among 608 participants, 512 males (84.2%) with a mean age of 60.4 were defined as positive by the ADAM score, whereas 96 males (15.8%) with a mean age of 51.7 years were defined as negative. Likewise, 437 males (71.9%) with a mean age of 61.3 years were positively screened by the AMS questionnaire, whereas 171 males (28.1%) with a mean age of 53.0 years were negative. Meanwhile, 332 out of 608 males underwent blood tests for assessing their corresponding biochemical and hormone measurements. The detailed characteristics of all study participants are shown in Table 1.

Table 1.

Characteristics of the subjects at baseline

Variable	Number	Mean±SD, M (P25‐P75) or frequency (%)
Age (years)	332	58.83±9.86
BMI (kg/m2)	317	24.32±2.74
Cholesterol (mmol/L)	327	4.89±0.80
HDL (mmol/L)	298	1.21±0.29
LDL (mmol/L)	298	3.10±0.72
Triglycerides (mmol/L)	327	1.40 (0.95‐2.02)
Glucose (mmol/L)	327	5.30 (4.92‐5.81)
PSA (ng/mL)	286	1.04 (0.53‐1.78)
LH (IU/L)	332	4.65 (3.26‐6.43)
SHBG (nmol/L)	332	37.65 (28.48‐49.90)
T (ng/mL)	332	4.20±1.35
cFT (ng/dL)	332	7.72±1.92
cBT (ng/dL)	332	180.99±45.10
FTI (%)	332	40.27±14.16
TSI	332	0.89 (0.64‐1.26)
Blood pressure	317
Normal		184 (58.0)
Hypertension		133 (42.0)
Fatty liver	301
Yes		106 (35.2)
No		195 (64.8)
ADAM	332
Negative		43 (13.0)
Positive		289 (87.0)
AMS	332
Negative		100 (30.1)
Mild		186 (56.0)
Moderate to severe		46 (13.9)

BMI, Body mass index; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; PSA, prostate‐specific antigen; LH, luteinizing hormone; SHBG, sex hormone‐binding globulin; T, total testosterone; cFT, calculated free testosterone; cBT, calculated bioavailable testosterone; FTI, free testosterone index; TSI, testosterone secretion index; ADAM, Aging Male Questionnaire; AMS, Aging Male Symptoms Scale.

3.2. The correlation between T, cFT, cBT, and age

Linear regression analysis showed that both the average cFT and cBT decreased with age (R ²=.153, P<.001) (Figure 1A,B), and that there is no significant correlation between T and age (R ²=.08, P=.11; Figure 1C).

Figure 1.

(A) The trend of calculated free testosterone with age. (B) The trend of calculated bioavailable testosterone with age. (C) The trend of T with age

3.3. Associations between questionnaire scores and testosterone levels

There was no significant difference in T, cFT, and cBT between males (all P>.05). Study participants were also divided into three groups based on their corresponding AMS questionnaire scores: negative, positive, and moderate to severe. There was no significant difference in the mean value of T among the above three groups (P>.05). However, there was significant difference in the mean values of cFT and cBT among the three groups (P<.05). Males with higher AMS scores appeared to have lower cFT and cBT values on average (Table 2).

Table 2.

Associations between questionnaires and testosterone levels

Variable		Number	T (ng/mL)	P value	cFT (ng/dL)	P value	cBT (ng/dL)	P value
ADAM	Negative	43	3.96±1.22	.204a	7.93±2.98	.434a	185.93±49.14	.442a
	Positive	289	4.24±1.36		7.69±1.90		180.26±44.51
AMS	Negative	100	4.05±1.17	.399a	8.07±1.90	.006a	189.14±44.45	.006a
	Mild	186	4.28±1.34		7.72±1.84		180.93±43.13
	Moderate to severe	46	4.24±1.35		6.98±2.14		163.55±50.06

T, total testosterone; cFT, calculated free testosterone; cBT, calculated bioavailable testosterone; ADAM, Aging Male Questionnaire; AMS, Aging Male Symptoms Scale.

^aKruskal‐Wallis test.

3.4. Determine significant factors that affect T, cFT, and cBT

Males with hypertension exhibited significantly lower average T, cFT, and cBT values than those without hypertension (all P<.05). The average level of T in males with fatty liver was significantly lower than that in males without fatty liver (P<.01). However, the average level of cFT and cBT was not significantly different between males with or without fatty liver (Table 3). Through Pearson correlation analysis, we found that T levels were negatively correlated with BMI, triglycerides, and positively correlated with HDL, LH, SHBG, cFT, and cBT (Table 4). Meanwhile, cFT was negatively correlated with age, LH, SHBG, and positively correlated with T and cBT. There was no significant correlation between cFT and BMI (Table 5).

Table 3.

Associations between hypertension, fatty liver, and testosterone levels

Variable		Number	T (ng/mL)	P value	cFT (ng/dL)	P value	cBT (ng/dL)	P value
Hypertension	No	184	4.42±1.41	.001a	8.00±2.00	.011a	187.59±46.93	.011a
	Yes	133	3.92±1.21		7.44±1.79		174.59±41.96
Fatty liver	No	195	4.57±1.45	<.001a	7.75±2.06	.827a	181.69±48.36	.832a
	Yes	106	3.70±0.94		7.80±1.73		182.86±40.61

T, total testosterone; cFT, calculated free testosterone; cBT, calculated bioavailable testosterone.

^aKruskal‐Wallis test.

Table 4.

Correlation analysis of T and other continuous variables

Variable	Number	Pearson correlation coefficient	P value
Age (years)	332	.088	.11a
BMI (kg/m2)	317	−.375	<.001a
Triglycerides (mmol/L)	327	−.278	<.001a
Cholesterol (mmol/L)	327	−.043	.436a
HDL (mmol/L)	298	.29	<.001a
LDL (mmol/L)	298	−.068	.243a
Glucose (mmol/L)	327	−.102	.066a
PSA (ng/mL)	286	.003	.955a
LH (IU/L)	332	.123	.025a
SHBG (nmol/L)	332	.661	<.001a
cFT (ng/dL)	332	.576	<.001a
cBT (ng/dL)	332	.576	<.001a

BMI, Body mass index; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; PSA, prostate‐specific antigen; LH, luteinizing hormone; SHBG, sex hormone‐binding globulin; cFT, calculated free testosterone; cBT, calculated bioavailable testosterone.

^aKruskal‐Wallis test.

Table 5.

Correlation analysis of cFT and other continuous variables

Variable	Number	Pearson correlation coefficient	P value
Age (years)	332	−.391	<.001a
BMI (kg/m2)	317	−.081	.148a
Triglycerides (mmol/L)	327	.041	.46a
Cholesterol (mmol/L)	327	.091	.102a
HDL (mmol/L)	298	−.014	.816a
LDL (mmol/L)	298	.086	.14a
Glucose (mmol/L)	327	−.087	.116a
PSA (ng/mL)	286	−.017	.78a
LH (IU/L)	332	−.224	<.001a
SHBG (nmol/L)	332	−.192	<.001a
cBT (ng/dL)	332	1	<.001a
T (ng/mL)	332	.576	<.001a

BMI, Body mass index; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; PSA, prostate‐specific antigen; LH, luteinizing hormone; SHBG, sex hormone‐binding globulin; cBT, calculated bioavailable testosterone; T, total testosterone.

^aKruskal‐Wallis test.

3.5. ROC curve analysis of T and cFT

We analyzed the ROC curve of T and cFT. When the state variable was positive in ADAM, the optimal cutoff point of T and cFT was 4.42 and 7.62 ng/dL, respectively, but there was no statistically significant difference in their predictive power for LOH. When the state variable was positive in AMS, the optimal cutoff point of T and cFT was 4.25 and 8.2 ng/dL, respectively, and the result of cFT was statistically significant while result of T was not. When the result of AMS was moderate to severe, the optimal cutoff point of T and cFT was 5.18 and 7.61 ng/dL, respectively, and the result of cFT was statistically significant while result of T was not (Table 6). There were significant differences in the predictive power of LOH between T and cFT when the state variable was positive in AMS and the result of AMS was moderate to severe. The corresponding results are shown in Figure 2.

Table 6.

ROC curves of T and cFT

Variable	State variable	AUC	Cutoff point	YI (%)	P value
cFT	ADAM = yes	.570	7.62	19.3	.135*
T	ADAM = yes	.534	4.42	9.2	.456
cFT	AMS = yes	.640	8.2	25.5	.001
T	AMS = yes	.551	4.25	13.4	.239
cFT	AMS = moderate‐severe	.680	7.61	35.2	.027
T	AMS = moderate‐severe	.582	5.18	19.0	.378

T, total testosterone; cFT, calculated free testosterone; ADAM, Aging Male Questionnaire; AMS, Aging Male Symptoms Scale; ROC, Receiver Operating Characteristic; YI, Youden's index.

Figure 2.

(A) Receiver Operating Characteristics (ROC) of T and calculated free testosterone (cFT) when Aging Male Symptoms (AMS) was positive. (B) ROC of T and cFT when AMS was moderate to severe

4. Discussion

The etiology of LOH remains controversial.16 Although a lot of evidence suggesting that testosterone levels gradually decrease with age which contributes to the LOH progress,17, 18, 19 doubts were raised about the theory recently.20, 21, 22, 23 In our previous study, we found that T levels did not decline with age significantly. On the contrary, T levels showed an increased trend over 50 years old.11 Kelsey et al.24 arrived at a similar conclusion through analyzing a dataset obtained from 13 studies (n=10,097; age range: 3‐101 years), and found no progressive decline in testosterone among middle‐aged and older men.

Obesity rather than age is the leading risk factor of decrease in testosterone among aging men. Serum T levels in males at any age with BMI >30 kg/m² is, on average, 30% lower than those of males with BMI <25 kg/m².25 In the EMAS study, 73% of men meeting the criteria of LOH were obese or overweight,9 and the corresponding results confirmed a strong association between low T and obesity.26 This relationship is complex and highly confounded and likely to be bidirectional.27 Obesity can result in low levels of T, but hypogonadism can also promote fat accumulation, insulin resistance, and metabolic syndromes.28

About 60% of testosterone existed in the blood is bound to sex hormone‐binding globulin, 38% is bound to albumin, and the remaining 2% exists as free testosterone. Free testosterone and albumin‐binding testosterone function biologically as androgens and comprise the so‐called bioavailable testosterone.29 Equilibrium dialysis is the gold standard for free testosterone measurement. Free testosterone assays based on analog displacement immunoassays are widely adopted, but in lack of accurate measurement results, they should not be used in our study.30, 31 Alternately, cFT that is calculated from measuring serum SHBG and T levels exhibits reliable correlation with free testosterone as indicated by equilibrium dialysis.12, 30

Various questionnaires have been developed for the objective assessment of the symptoms of LOH: the Androgen Deficiency in Aging Male Questionnaire (ADAM),32 the Aging Male Symptoms Scale (AMS) 13, 14, and the Massachusetts Male Aging Survey Questionnaire (MMAS).15 All the questionnaires are of high sensitivity but poor specificity, hence not recommended for the diagnosis of LOH alone.1, 33, 34 The ADAM questionnaire is considered one of the most efficient clinical tools to assess the males with testosterone deficiency.35 However, we found that the ADAM scores had no significant correlation with the levels of T, cFT, and cBT. Likewise, the corresponding results of AMS were not significantly associated with the levels of T either, whereas they were related to cFT and cBT, the lower the value of cFT and cBT, the higher the score of AMS on average.

Symptoms caused by the lack of testosterone are non‐specific, including sexual, physical, and mental symptoms. Sexual dysfunction, like a poor morning erection and erectile dysfunction are not only caused by testosterone deficiency but also resulted from nerves, blood vessels, and psychological factors. Therefore, sexual dysfunction cannot be simply explained by testosterone deficiency. Using the AMS questionnaire to screen for testosterone deficiency is more complicated, but it is feasible in real situations.

We believe that the cFT level is a better indicator than the T with respect to LOH diagnosis for the following reasons: (1) the average T levels do not decline with age in the middle‐aged and elderly males; (2) obesity is the leading cause for testosterone decline among aging males making serum T concentrations not reliable for diagnosing hypogonadism particularly for those obese males; (3) T levels have no significant correlation with the ADAM and AMS scores; (4) the average cFT declines with age and meanwhile it is not affected by BMI; (5) there is a significant correlation between the calculated free testosterone and the free testosterone as revealed by the equilibrium dialysis; (6) the cFT levels are associated with the symptom questionnaire of androgen deficiency like AMS: the lower the average value of cFT, the higher the average score of AMS.

The first 25 percentile of T levels in 20‐40‐year‐old males are generally considered as the standard for diagnosing testosterone deficiency.33, 36 At present, there is no uniform, free serum testosterone level which can be set as a standard for the diagnosis of LOH. Vermeulen et al. suggested that the free testosterone level below 6.5 ng/mL could be considered as signs of LOH.12, 37 The EMAS proposed that the diagnosis of LOH required FT <6.4 ng/mL.26 Bhasin et al.38 concluded that testosterone deficiency was presented when cFT was <7.0 ng/mL. In our study, the optimal cutoff point of cFT was 7.98 ng/mL when the AMS scores were positive, and 5.89 ng/mL when the AMS scores were mild and moderate, respectively.

5. Conclusion

In conclusion, the corresponding cFT levels in middle‐aged and elderly males can be combined with the AMS questionnaire for screening LOH. However, we recommend our peers to conduct large‐scale studies in order to verify our conclusions.