Anemia and testosterone deficiency risk: insights from NHANES data analysis and a Mendelian randomization analysis

Figures & data

Figure 1. General flow chart of the study.

A schematic diagram illustrating the general flow of this study. In brief, data from cross-sectional surveys in the NHANES database revealed anemia as a risk factor for testosterone deficiency, which was subsequently validated through Mendelian randomization.

Table 1. Demographic and clinical characteristics of participants included in the NHANES.

Characteristics	Total Median (IQR) or N (%)	Anemia status Median (IQR) or N (%)		p Value
	n = 21,786	Yes (n = 2318, 10.6%)	No (n = 19,468, 89.4%)
Testosterone (ng/dL)	39.1 (16.2, 362.0)	27.4 (14.4, 223.7)	42.3 (16.5, 375.0)	<0.001
Hb (g/L)	141 (131, 151)	115.0 (108.0, 119.0)	143.0 (134.0, 153.0)	<0.001
Age	47 (31, 62)	53 (37, 71)	46 (31, 61)	<0.001
Sex				<0.001
Male	11,440 (52.5%)	829 (35.8%)	10,611 (54.5%)
Female	10,346 (47.5%)	1489 (64.2%)	8857 (45.5%)
Ethnic				<0.001
Mexican American	3198 (14.7%)	281 (12.1%)	2917 (14.9%)
Other Hispanic	2167 (9.9%)	206 (8.9%)	1961 (10.1%)
Non-Hispanic White	8550 (39.2%)	531 (22.9%)	8019 (41.2%)
Non-Hispanic Black	4696 (21.6%)	978 (42.2%)	3718 (19.1%)
Other	3175 (14.6%)	322 (13.9%)	2853 (14.7%)
BMI (kg/m^2)	27.6 (24.1, 32.2)	28.0 (24.3, 33.3)	27.6 (24.0, 32.1)	<0.001
Waist (cm)	96.2 (87.2, 107.6)	96.1 (89.2, 107.8)	96.2 (87.0, 107.6)	0.015
Marital status				<0.001
Married	12,160 (55.8%)	1345 (60.7%)	10,815 (58.2%)
Divorced	2141 (9.8%)	240 (10.8%)	1901 (10.2%)
Single	6496 (29.8%)	632 (28.5%)	5864 (31.6%)
Missing	989 (4.6%)
Education				<0.001
Less than high school	2018 (9.3%)	273 (11.8%)	1745 (8.9%)
High school	7655 (35.1%)	870 (37.5%)	6785 (34.9%)
More than high school	12,113 (55.6%)	1175 (50.7%)	10,938 (56.2%)
Hypertension				<0.001
Yes	7326 (33.6%)	1072 (47.5%)	6254 (32.8%)
No	14,009 (64.3%)	1186 (52.5%)	12,823 (67.2%)
Missing	451 (2.1%)
Diabetes				<0.001
Yes	3709 (17.0%)	638 (29.8%)	3071 (17.5%)
No	15,946 (71.1%)	1501 (70.2%)	14,445 (82.5%)
Missing	2581 (11.9%)
LDL (mg/dL)	97 (53, 127)	86.0 (52.0, 117.0)	98.0 (54.0, 128.0)	<0.001
Missing	12,931 (59.3%)
TC (mg/dL)	185 (157, 213)	172.0 (144.0, 200.0)	186.0 (159.0, 214.0)	<0.001
Missing	5016 (23.0%)
HDL (mg/dL)	49 (41, 60)	51.0 (42.0, 62.0)	49.0 (41.0, 60.0)	<0.001
Missing	9922 (45.5%)
MCV (fL)	89.6 (86.3, 92.8)	85.3 (78.4, 90.8)	89.9 (86.8, 92.9)	<0.001
MCH (pg)	30.4 (29, 31.6)	28.2 (25.4, 30.5)	30.5 (29.2, 31.6)	<0.001
MCHC (g/L)	33.8 (33.1, 34.4)	32.8 (32.0, 33.7)	33.8 (33.2, 34.5)	<0.001

Hb: hemoglobin; BMI: body mass index; LDL: low-density lipoprotein; TC: total cholesterol; HDL: high-density lipoprotein; MCV: mean corpuscular volume; MCH: mean corpuscular hemoglobin; MCHC: mean corpuscular hemoglobin concentration; IQR: interquartile range.

Table 2. Causal relationships between anemia and testosterone deficiency: logistic regression.

Exposure variable	Non-adjusted	Adjust model I	Adjust model II
Anemia	Testosterone, β (95% CI), p value	Testosterone, β (95% CI), p value	Testosterone, β (95% CI), p value
Yes (n = 2318, 10.6%)	Reference	Reference	Reference
No (n = 19468, 89.4%)	93.723 (81.996, 105.449), p < 0.01	17.936 (10.359, 25.513), p < 0.01	22.616 (3.873, 41.359) p = 0.018

Notes: Unadjusted: not adjusted for any variables.

Adjust Model I: adjusted for age, sex, and ethnicity.

Adjust Model II: adjusted for age, sex, ethnicity, BMI, waist, education, marital status, TC, HDL, LDL, diabetes, hypertension, MCV, MCH, and MCHC.

Table 3. Results of regression analysis for anemia and testosterone levels after 5 imputations.

Model		Testosterone level (β (95% CI) p value)
		Non-adjusted	Adjust Model I	Adjust Model II
Model 1
	Anemia
	YES	Reference	Reference	Reference
	NO	93.723 (81.996, 105.449) p < 0.01	17.936 (10.359, 25.513) p < 0.01	17.632 (10.239, 25.025) p < 0.01
Model 2
	Anemia
	YES	Reference	Reference	Reference
	NO	93.723 (81.996, 105.449) p < 0.01	17.936 (10.359, 25.513) p < 0.01	17.723 (10.329, 25.116) p < 0.01
Model 3
	Anemia
	YES	Reference	Reference	Reference
	NO	93.723 (81.996, 105.449) p < 0.01	17.936 (10.359, 25.513) p < 0.01	17.932 (10.540, 25.324) p < 0.01
Model 4
	Anemia
	YES	Reference	Reference	Reference
	NO	93.723 (81.996, 105.449) p < 0.01	17.936 (10.359, 25.513) p < 0.01	18.186 (10.794, 25.579) p < 0.01
Model 5
	Anemia
	YES	Reference	Reference	Reference
	NO	93.723 (81.996, 105.449) p < 0.01	17.936 (10.359, 25.513) p < 0.01	17.362 (9.974, 24.751) p < 0.00001
Overall	Pool estimates from 5 multiple imputed data			17.767 (10.358, 25.176) p < 0.01

Note: Exposure variable: Anemia.

Outcome variable: Testosterone.

Unadjusted: not adjusted for any variables.

Adjust Model I: adjusted for age; sex; ethnic.

Adjust Model II: adjusted for age, sex, ethnicity, BMI, waist circumference, education, marital status, TC, HDL, LDL, diabetes, hypertension, MCV, MCH, and MCHC.

Consistent results from multiple models after multiple imputations of missing covariates show that testosterone levels are higher in non-anemic patients compared to anemic patients.

Table 4. Stratified analysis of testosterone levels in anemic and nonanemic patients.

Characteristic	N (%) total = 21,786	β (95% CI), p value
Sex
Male	11,440 (52.5)	56.6 (38.6, 74.6), p < 0.01
Female	10,346 (47.5)	−1.4 (−3.0, 0.1), p = 0.08
Age
18–35	6902 (31.7)	152.7 (127.3, 178.2), p < 0.01
36–56	7461 (34.2)	114.0 (94.6, 133.4), p < 0.01
57–85	7423 (34.1)	30.8 (13.3, 48.3), p < 0.01
Ethnic
Mexican American	3198 (14.7)	159.7 (131.1, 188.3), p < 0.01
Other Hispanic	2167 (9.9)	101.6 (67.4, 135.7), p < 0.01
Non-Hispanic White	8550 (39.2)	74.1 (51.2, 96.9), p < 0.01
Non-Hispanic Black	4696 (21.6)	111.9 (94.5, 129.4), p < 0.01
Other	3175 (14.6)	95.9 (68.4, 123.4), p < 0.01
BMI (kg/m^2)
13.4 − 25.3	7213 (33.1)	77.6 (53.6, 101.7), p < 0.01
25.4 − 30.3	7200 (33.0)	108.5 (87.7, 129.3), p < 0.01
30.4 − 82.1	7373 (33.9)	89.3 (73.9, 104.8), p < 0.01
Waist (cm)
55.5 − 91.1	7224 (33.2)	91.7 (67.2, 116.1), p < 0.01
91.2 − 103.4	7282 (33.4)	91.1 (71.8, 110.5), p < 0.01
103.5 − 177.9	7280 (33.4)	99.3 (82.0, 116.5), p < 0.01
Education
Less than high school	2018 (9.3)	111.2 (78.8, 143.7), p < 0.01
High school	7655 (35.1)	90.4 (71.0, 109.7), p < 0.01
More than High school	12,113 (55.6)	95.9 (79.7, 112.2), p < 0.01
Marital status
Married	12,160 (55.8)	75.9 (61.1, 90.7), p < 0.01
Divorced	2141 (9.8)	41.7 (6.5, 77.0), p = 0.02
Single	6496 (29.8)	138.6 (115.2, 162.0), p < 0.01
Missing	989 (4.6)
TC (mg/dL)
0.776 − 166	5555 (25.5)	99.9 (79.3, 120.6), p < 0.01
167–202	5557 (25.5)	109.7 (86.0, 133.4), p < 0.01
203–813	5658 (26.0)	99.9 (73.9, 126.0), p < 0.01
Missing	5016 (23.0)
HDL (mg/dL)
10–43	3771 (17.3)	111.8 (85.8, 137.8), p < 0.01
44–55	4060 (18.6)	118.9 (91.3, 146.5), p < 0.01
56–175	4033 (18.5)	58.0 (33.0, 83.0), p < 0.01
Missing	9922 (45.5)
LDL (mg/dL)
0.181 − 74	2915 (13.4)	77.0 (42.0, 111.9), p < 0.01
75–116	2985 (13.7)	109.9 (73.9, 145.8), p < 0.01
117–629	2955 (31.2)	125.7 (85.5, 165.9), p < 0.01
Missing	12927 (59.3)
Diabetes
Yes	3709 (17.0)	60.1 (39.7, 80.5), p < 0.01
No	15,946 (71.1)	103.1 (88.5, 117.7), p < 0.01
Missing	2581 (11.9)
Hypertension
Yes	7326 (33.6)	66.3 (50.0, 82.7), p < 0.01
No	14,009 (64.3)	109.9 (93.4, 126.5), p < 0.01
Missing	451 (2.1)

95% CI: 95% confidence interval; BMI: body mass index; LDL: low-density lipoprotein; TC: total cholesterol; HDL: high-density lipoprotein

Analysis was stratified by various indicators including age, gender, race, BMI, waist circumference, education, marital status, blood lipids, hypertension, and diabetes. The results show that only the gender-stratified analysis was inconsistent, with no statistically significant difference in testosterone levels between female anemic and non-anemic patients. However, all other stratifications consistently demonstrated lower testosterone levels in anemic patients compared to non-anemic patients.

Figure 2. Mendelian randomization reveals a causal relationship between anemia and testosterone deficiency. Hemoglobin as the exposure variable and total testosterone was used as the outcome variable. The results revealed greater testosterone levels in nonanemic individuals than in anemic individuals (p < 0.05 for MR–Egger, weighted median, inverse–variance weighting, and weighted mode).

Figure 3. Kaplan–Meier survival analysis of testosterone levels differences between anemic and nonanemic patients. The red line represents individuals with testosterone deficiency, while the green line represents individuals with normal testosterone levels. (a) Kaplan–Meier survival plot comparing anemic patients with testosterone deficiency to anemic patients with normal testosterone levels. (b) Kaplan–Meier survival plot comparing nonanemic patients with testosterone deficiency to nonanemic patients with normal testosterone levels.

Data availability statement

The data used in this study are available at the official websites of the NHANES (https://www.cdc.gov/nchs/nhanes/index.htm) and GWAS (https://gwas.mrcieu.ac.uk/) projects.