Potential of seven insulin resistance indicators as biomarkers to predict infertility risk in U.S. women of reproductive age: a cross-sectional study.

Table  1 presents the weighted baseline characteristics of the study sample. This study included 1,110 women of reproductive age (20 to 45 years) with a mean age of 32.05 ± 7.57 years, of whom 140 (12.61%) were diagnosed with infertility. Infertile women were significantly older and exhibited elevated insulin resistance indicators (METS-IR, TyG, TyG-WC, TyG-BMI, TyG-WHtR, and HOMA-IR), while the TG/HDL indicator showed no significant difference between groups. Additionally, infertile women were more likely to be married or cohabiting and demonstrated higher prevalence rates of overweight/obesity, hypertension, and diabetes. A greater proportion of infertile women experienced menarche before age 10 and had a history of pelvic inflammatory disease. Table 1 Weighted baseline characteristics of study participants Characteristic Total ( n  = 1110) Non-infertility ( n  = 970) Infertility ( n  = 140) P value Age, mean ± SD (years) 32.05 ± 7.57 31.64 ± 7.51 34.82 ± 7.39 ˂0.0001 METS-IR, mean ± SD 42.48 ± 14.81 41.59 ± 14.42 48.36 ± 16.00 ˂0.0001 TyG, mean ± SD 8.22 ± 0.64 8.20 ± 0.64 8.33 ± 0.65 0.0274 TyG-WC, mean ± SD 798.24 ± 192.80 787.34 ± 189.06 870.70 ± 201.58 < 0.0001 TyG-BMI, mean ± SD 246.28 ± 80.22 241.81 ± 78.68 275.99 ± 84.00 < 0.0001 TyG-WHtR, mean ± SD 4.90 ± 1.17 4.84 ± 1.16 5.31 ± 1.18 < 0.0001 HOMA-IR, mean ± SD 3.01 ± 3.31 2.93 ± 3.15 3.60 ± 4.18 0.0232 TG/HDL, mean ± SD 1.90 ± 3.15 1.85 ± 3.19 2.26 ± 2.80 0.1407 Race (%) 0.7913 Mexican American 182 (11.16) 160 (11.26) 22 (10.50) Other Hispanic 113 (7.46) 102 (7.82) 11 (5.09) Non-Hispanic White 401 (58.23) 345 (57.70) 56 (61.74) Non-Hispanic Black 223 (12.74) 194 (12.76) 29 (12.57) Other Race 191 (10.41) 169 (10.45) 22 (10.10) Education (%) 0.6712 Less than 9th grade 47 (2.86) 43 (2.91) 4 (2.60) 9-11th grade 120 (8.28) 104 (8.00) 16 (10.17) High school graduate/ GED or equivalent 213 (20.15) 186 (20.51) 27 (17.72) Some college or AA degree 404 (35.63) 342 (35.06) 62 (39.39) College graduate or above 326 (33.08) 295 (33.52) 31 (30.12) Marital (%) < 0.0001 Married/Living with partner 635 (59.73) 531 (56.61) 104 (80.47) Widowed/Devoiced/Separate 128 (10.49) 113 (10.65) 15 (9.37) Never married 347 (29.78) 326 (32.74) 21 (10.16) PIR (%) 0.6327 Low 411 (29.92) 359 (29.88) 52 (30.23) Middle 413 (38.40) 357 (37.96) 56 (41.31) High 286 (31.68) 254 (32.16) 32 (28.47) BMI (%) 0.0088 Underweight 30 (2.72) 28 (2.81) 2 (2.12) Normal 357 (32.99) 322 (34.61) 35 (22.21) Overweight and Obesity 723 (64.29) 620 (62.58) 103 (75.67) Smoke (%) 0.2063 Never 761 (64.94) 674 (65.88) 87 (58.67) Former 144 (15.16) 120 (14.57) 24 (19.06) Now 205 (19.90) 176 (19.55) 29 (22.28) Drink (%) 0.9042 Yes 700 (67.87) 610 (67.81) 90 (68.31) No 410 (32.13) 360 (32.19) 50 (31.69) Hypertension (%) 0.0003 Yes 188 (15.87) 150 (14.34) 38 (26.05) No 922 (84.13) 820 (85.66) 102 (73.95) DM (%) 0.0028 Yes 80 (6.00) 65 (5.17) 15 (11.48) No 1030 (94.00) 905 (94.83) 125 (88.52) Menstrual (%) (years) 0.0301 ˂ 10 51 (3.89) 40 (3.32) 11 (7.72) 10–14 924 (84.11) 809 (84.36) 115 (82.47) 15–20 135 (11.99) 121 (12.32) 14 (9.81) History of previous treatment for pelvic infection (%) 0.0315 Yes 53 (4.35) 41 (3.84) 12 (7.74) No 1057 (95.65) 929 (96.16) 128 (92.26) History of contraceptive use (%) 0.8041 Yes 771 (75.76) 669 (75.64) 102 (76.59) No 339 (24.24) 301 (24.36) 38 (23.41) History of female hormone use (%) 0.2095 Yes 40 (4.23) 33 (3.94) 7 (6.19) No 1070 (95.77) 937 (96.06) 133 (93.81) Weighted baseline characteristics of study participants Stratified multivariable weighted logistic regression models were used to systematically evaluate the associations between seven insulin resistance indicators (METS-IR, TyG, TyG-WC, TyG-BMI, TyG-WHtR, HOMA-IR, and TG/HDL) and infertility risk. Detailed results are presented in Table  2 . When analyzed as continuous variables, the insulin resistance indicators METS-IR, TyG-WC, TyG-BMI, and TyG-WHtR were significantly positively associated with infertility across all three adjusted models. In contrast, TyG, HOMA-IR, and TG/HDL showed no significant association with infertility risk ( P  > 0.05). After a stratified analysis by quartiles, we observed in Model 3, which adjusted for all confounding variables, that the risk of infertility in the third quartile of TyG, TyG-WC, TyG-BMI, TyG-WHtR, and TG/HDL was 1.86, 2.83, 2.90, 4.08, and 2.56 times higher than that in the first quartile, respectively. Compared to the lowest quartile, the risk of infertility in the highest quartile of METS-IR, TyG-WC, TyG-BMI, and TyG-WHtR increased by 4.48, 3.36, 4.31, and 4.65 times, respectively. Furthermore, increases in METS-IR, TyG-WC, TyG-BMI, and TyG-WHtR exhibited a significant dose-response relationship with infertility risk (trend P-value < 0.05). Table 2 Association between IR-related indexes and infertility among US reproductive women in NHANES 2013–2018 Exposure Model 1 OR (95% CI) P Model 2 OR (95% CI) P Model 3 OR (95% CI) P METS-IR 1.03 (1.01, 1.04) <0.0001 1.02 (1.01, 1.04) 0.0001 1.03 (1.01, 1.05) 0.0003 TyG 1.28 (0.99, 1.64) 0.0591 1.22 (0.92, 1.61) 0.1718 1.13 (0.82, 1.56) 0.4409 TyG-WC 1.002 (1.001, 1.003) < 0.0001 1.002 (1.001, 1.003) 0.0005 1.002 (1.001, 1.003) 0.0020 TyG-BMI 1.005 (1.003, 1.007) < 0.0001 1.004 (1.002, 1.006) 0.0003 1.005 (1.002, 1.008) 0.0013 TyG-WHtR 1.35 (1.17, 1.56) <0.0001 1.30 (1.11, 1.53) 0.0010 1.38 (1.11, 1.72) 0.0038 HOMA-IR 1.04 (1.00, 1.07) 0.0604 0.03 (0.99, 1.07) 0.1194 1.02 (0.98, 1.07) 0.3897 TG/HDL 1.02 (0.98, 1.06) 0.2690 1.02 (0.98, 1.06) 0.3070 1.02 (0.98, 1.06) 0.4315 METS-IR quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 0.96 (0.53, 1.74) 0.8890 0.82 (0.44, 1.51) 0.5241 1.14 (0.55, 2.38) 0.7167 Q3 1.58 (0.92, 2.73) 0.1000 1.36 (0.76, 2.42) 0.3011 2.71 (0.98, 7.48) 0.0543 Q4 2.73 (1.64, 4.55) 0.0001 2.26 (1.31, 3.93) 0.0036 4.48 (1.63, 12.37) 0.0037 P for trend < 0.0001 0.0001 0.0001 TyG-WC quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 0.91 (0.49, 1.68) 0.7629 0.78 (0.42, 1.48) 0.4527 0.99 (0.49, 1.97) 0.9713 Q3 2.04 (1.19, 3.48) 0.0093 1.68 (0.95, 2.97) 0.0763 2.83 (1.20, 6.66) 0.0173 Q4 2.61 (1.55, 4.40) 0.0003 2.16 (1.23, 3.80) 0.0073 3.86 (1.58, 9.45) 0.0031 P for trend < 0.0001 0.0003 0.0006 TyG-BMI quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 0.96 (0.53, 1.74) 0.8890 0.83 (0.45, 1.53) 0.5427 1.15 (0.55, 2.41) 0.7123 Q3 1.68 (0.98, 2.89) 0.0592 1.44 (0.81, 2.57) 0.2094 2.90 (1.04, 8.10) 0.0418 Q4 2.61 (1.56, 4.36) 0.0002 2.15 (1.24, 3.75) 0.0066 4.31 (1.54, 12.04) 0.0053 P for trend < 0.0001 0.0003 0.0004 TyG-WHtR quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 1.47 (0.80, 2.72) 0.2155 1.29 (0.69, 2.41) 0.4254 1.79 (0.89, 3.58) 0.0999 Q3 2.64 (1.50, 4.65) 0.0007 2.29 (1.26, 4.14) 0.0062 4.08 (1.78, 9.39) 0.0009 Q4 2.92 (1.67, 5.10) 0.0002 2.48 (1.36, 4.51) 0.0030 4.65 (1.92, 11.23) 0.0006 P for trend < 0.0001 0.0008 0.0015 TyG quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 1.64 (0.95, 2.82) 0.0748 1.46 (0.83, 2.54) 0.1860 1.44 (0.82, 2.53) 0.2092 Q3 2.15 (1.28, 3.63) 0.0040 1.96 (1.13, 3.38) 0.0161 1.86 (1.06, 3.24) 0.0297 Q4 1.38 (0.79, 2.40) 0.2611 1.19 (0.66, 2.15) 0.5582 0.98 (0.51, 1.86) 0.9434 P for trend 0.2583 0.5651 0.9955 HOMA-IR quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 0.86 (0.51, 1.48) 0.5949 0.81 (0.46, 1.40) 0.4416 0.75 (0.42, 1.33) 0.3271 Q3 1.22 (0.74, 2.02) 0.4340 1.23 (0.73, 2.07) 0.4376 1.12 (0.63, 1.99) 0.7064 Q4 1.37 (0.84, 2.24) 0.2131 1.21 (0.72, 2.04) 0.4680 1.04 (0.56, 1.93) 0.9010 P for trend 0.0966 0.2451 0.5771 TG/HDL quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 1.93 (1.09, 3.42) 0.0251 1.79 (1.00, 3.22) 0.0504 1.74 (0.96, 3.15) 0.0694 Q3 2.57 (1.48, 4.47) 0.0008 2.56 (1.45, 4.54) 0.0013 2.56 (1.42, 4.63) 0.0019 Q4 2.04 (1.16, 3.61) 0.0139 1.86 (1.02, 3.38) 0.0433 1.65 (0.86, 3.15) 0.1305 P for trend 0.0834 0.2110 0.5269 Notes: OR (95%CI) derived from multibariable logistic regression models. Model 1: Non-adjusted model adjust for: None. Model 2: Adjust for: age, race, education, marital status, and PIR. Model 3: Adjust for: age, race, education, marital status, PIR, BMI, smoke, drink, hypertension, diabetes, age at first menstrual period, previous pregnancy, previous use of female hormones, prior use of birth control pills, and history of treatment for a pelvic infection Association between IR-related indexes and infertility among US reproductive women in NHANES 2013–2018 Notes: OR (95%CI) derived from multibariable logistic regression models. Model 1: Non-adjusted model adjust for: None. Model 2: Adjust for: age, race, education, marital status, and PIR. Model 3: Adjust for: age, race, education, marital status, PIR, BMI, smoke, drink, hypertension, diabetes, age at first menstrual period, previous pregnancy, previous use of female hormones, prior use of birth control pills, and history of treatment for a pelvic infection Receiver operating characteristic (ROC) curve analysis was employed to assess the diagnostic discrimination capabilities of seven insulin resistance-related indices for infertility, as shown in Table  3 . The results indicated that METS-IR had the highest predictive performance for infertility risk, with an AUC of 0.619. TyG-BMI (AUC: 0.617), TyG-WC (AUC: 0.611), and TyG-WHtR (AUC: 0.610) followed closely in predictive accuracy. These four indices demonstrated similar performance in predicting infertility risk (Fig.  2 ). In contrast, TG/HDL, HOMA-IR, and TyG exhibited lower predictive efficacy for infertility risk, with AUC values between 0.5 and 0.6 (Fig.  3 ). Furthermore, we calculated and incorporated the AUC for BMI and WC as independent indicators in the cohort, with respective values of 0.6183 and 0.6181. Table 3 AUC and cut-off values of seven IR surrogates, BMI and WC for prediction of infertility Variables AUC (95% CI) Cut-off Specificity Sensitivity METS-IR 0.6186 (0.5681–0.6692) 48.7883 0.7320 0.4929 BMI 0.6183(0.5678–0.6688) 30.3500 0.6186 0.6214 WC 0.6181(0.5678–0.6684) 95.5500 0.5794 0.6286 TyG-BMI 0.6168 (0.5666–0.6669) 267.9790 0.6814 0.5357 TyG-WC 0.6110 (0.5613–0.6606) 775.7858 0.5433 0.6714 TyG-WHtR 0.6099(0.5611–0.6586) 4.6594 0.5072 0.7000 TG/HDL 0.5662 (0.5193–0.6131) 0.9537 0.3402 0.8143 HOMA-IR 0.5458 (0.4933–0.5984) 2.3427 0.5577 0.5500 TyG 0.5436 (0.4961–0.5911) 7.7099 0.2227 0.8929 AUC and cut-off values of seven IR surrogates, BMI and WC for prediction of infertility Fig. 2 Receiver operating characteristic curves of METS-IR, TyG-BMI, TyG-WC, TyG-WHtR, BMI and WC for identifying infertility Receiver operating characteristic curves of METS-IR, TyG-BMI, TyG-WC, TyG-WHtR, BMI and WC for identifying infertility Fig. 3 Receiver operating characteristic curves of TG/HDL, HOMA-IR and TyG for identifying infertility Receiver operating characteristic curves of TG/HDL, HOMA-IR and TyG for identifying infertility Building upon the previously presented findings, this study further examines the linear and nonlinear relationships, as well as the saturation effects, between METS-IR, TyG-BMI, TyG-WC, TyG-WHtR, and infertility. Smoothing curve fitting analysis revealed that METS-IR (Fig.  4 ) and TyG-BMI (Fig.  5 ) exhibit a nonlinear positive correlation with infertility, while TyG-WC (Fig.  6 ) and TyG-WHtR (Fig.  7 ) show a linear positive correlation. To gain deeper insights into the nonlinear associations and saturation effects of METS-IR and TyG-BMI on infertility, a threshold effect analysis was conducted (see Table  4 ). The results identified a critical threshold for METS-IR at 49.92. Below this threshold, an increase in METS-IR is significantly positively associated with the risk of infertility; however, beyond this value, the correlation stabilizes. Similarly, TyG-BMI is significantly positively correlated with infertility risk when below 290.18, but this correlation weakens and stabilizes once the threshold is surpassed. Fig. 4 The smooth curve between METS-IR and infertility The smooth curve between METS-IR and infertility Fig. 5 The smooth curve between TyG-BMI and infertility The smooth curve between TyG-BMI and infertility Fig. 6 The smooth curve between TyG-WC and infertility The smooth curve between TyG-WC and infertility Fig. 7 The smooth curve between TyG-WHR and infertility. Note: ( A ) Each blue dot represents a sample. ( B ) The solid red line indicates the smooth curve fitting between the variables, and the blue band represents the 95% confidence interval of the fit The smooth curve between TyG-WHR and infertility. Note: ( A ) Each blue dot represents a sample. ( B ) The solid red line indicates the smooth curve fitting between the variables, and the blue band represents the 95% confidence interval of the fit Table 4 Threshold effect analysis of METS-IR, TyG-BMI, and infertility METS-IR OR (95% CI) P -Value < 49.92 1.08 (1.03, 1.13) 0.0008 ≥ 49.92 1.01 (0.99, 1.03) 0.3882 Likelihood ratio 0.021 TyG-BMI OR (95% CI) P-Value < 290.18 1.02 (1.01, 1.02) 0.0004 0.0004 ≥ 290.18 1.00 (1.00, 1.01) 0.8151 0.8151 Likelihood ratio 0.007 Notes: Adjusted for age, race, education, marital status, PIR, BMI, smoke, drink, hypertension, diabetes, age at first menstrual period, previous pregnancy, previous use of female hormones, prior use of birth control pills, and history of treatment for a pelvic infection Threshold effect analysis of METS-IR, TyG-BMI, and infertility Notes: Adjusted for age, race, education, marital status, PIR, BMI, smoke, drink, hypertension, diabetes, age at first menstrual period, previous pregnancy, previous use of female hormones, prior use of birth control pills, and history of treatment for a pelvic infection Considering the potential impact of age on infertility treatment, participants were divided into two age groups (< 35 years and ≥ 35 years), and subgroup analyses were conducted to assess the robustness of the positive correlation between METS-IR, TyG-BMI, TyG-WC, and TyG-WHtR and infertility across these groups. Detailed results are presented in Table  5 . In each subgroup, a multivariable-adjusted model was used, adjusting for all covariates except age. The results indicated that, in participants aged < 35 years, for each one-unit increase in METS-IR, TyG-BMI, TyG-WC, and TyG-WHtR, the risk of infertility increased by 3.74%, 0.62%, 0.28%, and 53.96%, respectively. As shown in Fig.  5 , the interaction test results indicate that the positive correlation between METS-IR, TyG-BMI, TyG-WC, TyG-WHtR, and infertility was not significantly influenced by age group (all P-values > 0.05). Table 5 Subgroup analysis of the effect of age on METS-IR, TyG-BMI, TyG-WC, and TyG-WHtR in relation to infertility Subgroup Infertility OR (95% CI) P P for Interaction METS-IR Age 0.1765 <35 1.0374 (1.0180, 1.0572) 0.0001 ≥35 1.0209 (0.9992, 1.0430) 0.0596 TyG-BMI Age 0.2147 <35 1.0062 (1.0026, 1.0097) 0.0006 ≥35 1.0034 (0.9995, 1.0074) 0.0895 TyG-WC Age 0.0553 <35 1.0028 (1.0013, 1.0044) 0.0002 ≥35 1.0010 (0.9993, 1.0028) 0.2583 TyG-WHtR Age 0.1007 <35 1.5396 (1.1985, 1.9776) 0.0007 ≥35 1.1861 (0.8895, 1.5817) 0.2451 Subgroup analysis of the effect of age on METS-IR, TyG-BMI, TyG-WC, and TyG-WHtR in relation to infertility To investigate the relationship between insulin resistance-related indicators and infertility in individuals with normal body mass index (BMI), we performed a weighted baseline characteristic analysis (Table  6 ) and a weighted multiple logistic regression analysis (Table  7 ) on 357 women of reproductive age with normal BMI. The results showed that, except for the HOMA-IR index, there were no statistically significant differences between infertile and non-infertile participants in terms of METS-IR, TyG, TyG-WC, TyG-BMI, TyG-WHtR, and TG/HDL. In normal BMI participants, continuous variable analysis revealed no significant associations between METS-IR, TyG, TyG-WC, TyG-BMI, TyG-WHtR, HOMA-IR, and TG/HDL and the occurrence of infertility in all three adjusted models ( P  > 0.05). Further stratified analysis by quartiles indicated that only in the unadjusted Model 1 did participants in the fourth quartile of HOMA-IR have an infertility risk 0.22 times that of participants in the first quartile. Table 6 Weighted baseline characteristics of participants with normal BMI Characteristic Total ( n  = 357) Non-infertility ( n  = 322) Infertility ( n  = 35) P value Age, mean ± SD (years) 30.62 ± 7.54 30.00 ± 7.30 37.04 ± 7.01 ˂0.0001 METS-IR, mean ± SD 29.37 ± 3.19 29.39 ± 3.21 29.14 ± 3.00 0.6776 TyG, mean ± SD 7.94 ± 0.50 7.94 ± 0.51 7.91 ± 0.39 0.7609 TyG-WC, mean ± SD 632.93 ± 66.38 631.80 ± 66.61 644.59 ± 62.77 0.3036 TyG-BMI, mean ± SD 175.29 ± 17.70 175.21 ± 17.82 176.07 ± 16.31 0.7971 TyG-WHtR, mean ± SD 3.88 ± 0.42 3.87 ± 0.42 3.96 ± 0.43 0.2389 HOMA-IR, mean ± SD 1.48 ± 0.88 1.51 ± 0.88 1.17 ± 0.82 0.0361 TG/HDL, mean ± SD 1.13 ± 0.83 1.14 ± 0.85 1.02 ± 0.55 0.4416 Race (%) 0.7047 Mexican American 32 (6.72) 30 (7.09) 2 (2.86) Other Hispanic 43(9.23) 39 (9.31) 4 (8.36) Non-Hispanic White 136 (62.35) 123 (62.16) 13 (64.34) Non-Hispanic Black 48 (8.89) 41 (8.36) 7 (14.35) Other Race 98 (12.81) 89 (13.07) 9 (10.09) Education (%) 0.6229 Less than 9th grade 6 (0.81) 6 (0.89) 0 (0.00) 9-11th grade 29 (6.62) 24 (6.02) 5 (12.83) High school graduate/ GED or equivalent 62 (17.32) 55 (17.65) 7 (13.93) Some college or AA degree 113 (30.73) 99 (31.01) 14 (27.79) College graduate or above 147 (44.52) 138 (44.43) 9 (45.44) Marital (%) 0.0427 Married/Living with partner 204 (59.05) 179 (57.64) 25 (73.63) Widowed/Devoiced/Separate 28 (6.33) 25 (5.86) 3 (11.21) Never married 125 (34.62) 118 (36.50) 7 (15.16) PIR (%) 0.3377 Low 104 (24.57) 91 (24.22) 13 (28.20) Middle 119 (31.70) 110 (32.82) 9 (20.04) High 134 (43.73) 121 (42.96) 13 (51.76) Smoke (%) 0.0646 Never 263 (67.99) 242 (69.60) 21 (51.32) Former 43 (14.72) 36 (13.50) 7 (27.36) Now 31 (17.29) 44 (16.90) 7 (21.32) Drink (%) 0.1497 Yes 235 (74.85) 216 (75.88) 19 (64.20) No 122 (25.15) 106 (24.12) 16 (35.80) Hypertension (%) 0.0933 Yes 26 (6.08) 20 (5.42) 6 (12.91) No 331 (93.92) 820 (94.58) 29 (87.09) DM (%) 0.6402 Yes 3 (0.63) 3 (0.69) 0 (11.48) No 354 (99.37) 319 (99.31) 35 (100.00) Menstrual (%) (years) ˂0.0001 ˂ 10 6 (1.96) 3 (0.63) 3 (15.79) 10–14 286 (80.09) 262 (81.86) 24 (61.70) 15–20 65 (17.95) 57 (17.51) 8 (22.51) History of previous treatment for pelvic infection (%) 0.2272 Yes 13 (2.54) 10 (2.23) 3 (5.78) No 344 (97.46) 312 (97.77) 32 (94.22) History of contraceptive use (%) 0.5682 Yes 234 (75.01) 214 (75.42) 20 (70.80) No 123 (24.99) 108 (24.58) 15 (29.20) History of female hormone use (%) 0.4051 Yes 5 (1.97) 5 (2.16) 0 (0.00) No 352 (98.03) 317 (98.84) 35 (10.00) Weighted baseline characteristics of participants with normal BMI Table 7 Association between insulin resistance-related indices and infertility among U.S. Women of reproductive age with normal BMI in NHANES 2013–2018 Exposure Model 1 OR (95% CI) P Model 2 OR (95% CI) P Model 3 OR (95% CI) P METS-IR 1.06 (0.96, 1.18) 0.2297 1.05 (0.94, 1.17) 0.3776 1.06 (0.93, 1.20) 0.3857 TyG 1.11 (0.56, 2.22) 0.7652 0.94 (0.41, 2.11) 0.8740 0.89 (0.36, 2.20) 0.7989 TyG-WC 1.00 (1.00, 1.01) 0.2014 1.00 (1.00, 1.01) 0.7641 1.00 (1.00, 1.01) 0.6455 TyG-BMI 1.01 (1.00, 1.03) 0.1478 1.01 (0.99, 1.03) 0.4015 1.01 (0.99, 1.03) 0.3842 TyG-WHtR 1.92 (0.89, 4.14) 0.0945 1.28 (0.52, 3.15) 0.5848 1.38 (0.51, 3.76) 0.5304 HOMA-IR 0.71 (0.44, 1.15) 0.1665 0.81 (0.49, 1.34) 0.4113 0.82 (0.47, 1.43) 0.4841 TG/HDL 1.02 (0.70, 1.49) 0.9323 0.99 (0.65, 1.52) 0.9673 1.01 (0.63, 1.62) 0.9611 METS-IR quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 0.60 (0.19, 1.92) 0.3915 0.68 (0.20, 2.35) 0.5436 0.41 (0.10, 1.65) 0.2103 Q3 1.73 (0.68, 4.41) 0.2495 2.00 (0.72, 5.62) 0.1861 1.59 (0.54, 4.73) 0.4027 Q4 1.12 (0.41, 3.06) 0.8176 1.07 (0.36, 3.23) 0.9009 0.96 (0.30, 3.06) 0.9439 P for trend 0.4450 0.5500 0.6427 TyG quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 1.32 (0.47, 3.71) 0.6011 1.36 (0.44, 4.18) 0.5869 1.05 (0.31, 3.57) 0.9361 Q3 1.50 (0.54, 4.14) 0.4320 1.40 (0.46, 4.28) 0.5539 1.41 (0.44, 4.44) 0.5626 Q4 1.29 (0.46, 3.62) 0.6338 0.93 (0.28, 3.13) 0.9105 0.82 (0.22, 3.02) 0.7647 P for trend 0.6426 0.8487 0.8404 TyG-WC quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 1.16 (0.40, 3.34) 0.7874 1.02 (0.32, 3.21) 0.9785 0.99 (0.29, 3.40) 0.9862 Q3 1.65 (0.61, 4.48) 0.3237 1.21 (0.40, 3.61) 0.7372 1.14 (0.35, 3.72) 0.8259 Q4 1.30 (0.46, 3.66) 0.6175 0.87 (0.27, 2.81) 0.8201 0.87 (0.25, 2.99) 0.8231 P for trend 0.5059 0.8696 0.8630 TyG-BMI quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 1.56 (0.53, 4.57) 0.4212 1.20 (0.37, 3.89) 0.7581 1.30 (0.37, 4.58) 0.6803 Q3 1.37 (0.45, 4.11) 0.5788 1.27 (0.38, 4.19) 0.6996 1.35 (0.38, 4.77) 0.6416 Q4 2.13 (0.76, 5.95) 0.1497 1.45 (0.46, 4.55) 0.5210 1.47 (0.44, 4.94) 0.5366 P for trend 0.1784 0.5206 0.5618 TyG-WHtR quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 0.48 (0.14, 1.64) 0.2406 0.42 (0.11, 1.56) 0.1927 0.36 (0.08, 1.55) 0.1707 Q3 1.73 (0.68, 4.41) 0.2495 1.54 (0.55, 4.34) 0.4113 1.49 (0.49, 4.58) 0.4822 Q4 1.27 (0.48, 3.37) 0.6374 0.78 (0.25, 2.42) 0.6732 0.90 (0.26, 3.07) 0.8679 P for trend 0.2819 0.9814 0.7931 HOMA-IR quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 0.63 (0.25, 1.63) 0.3454 0.61 (0.22, 1.70) 0.3412 0.63 (0.20, 1.98) 0.4318 Q3 1.00 (0.42, 2.36) 1.0000 1.05 (0.41, 2.73) 0.9127 1.09 (0.38, 3.16) 0.8714 Q4 0.22 (0.06, 0.81) 0.0231 0.29 (0.07, 1.16) 0.0798 0.29 (0.06, 1.35) 0.1151 P for trend 0.0371 0.1366 0.1921 TG/HDL quartiles Q1 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) Q2 1.16 (0.40, 3.34) 0.7874 1.18 (0.38, 3.64) 0.7748 1.03 (0.32, 3.34) 0.9567 Q3 1.16 (0.40, 3.34) 0.7874 1.01 (0.32, 3.25) 0.9809 0.55 (0.14, 2.09) 0.3764 Q4 1.80 (0.67, 4.81) 0.2399 1.71 (0.57, 5.15) 0.3388 1.76 (0.55, 5.68) 0.3436 P for trend 0.1992 0.3028 0.2626 Model 1: Non-adjusted model adjust for: None. Model 2: Adjust for: age, race, education, marital status, and PIR. Model 3: Adjust for: age, race, education, marital status, PIR, smoke, drink, hypertension, diabetes, age at first menstrual period, previous pregnancy, previous use of female hormones, prior use of birth control pills, and history of treatment for a pelvic infection Association between insulin resistance-related indices and infertility among U.S. Women of reproductive age with normal BMI in NHANES 2013–2018 Model 1: Non-adjusted model adjust for: None. Model 2: Adjust for: age, race, education, marital status, and PIR. Model 3: Adjust for: age, race, education, marital status, PIR, smoke, drink, hypertension, diabetes, age at first menstrual period, previous pregnancy, previous use of female hormones, prior use of birth control pills, and history of treatment for a pelvic infection

This cross-sectional study uses data from the National Health and Nutrition Examination Survey (NHANES), conducted by the National Center for Health Statistics (NCHS). NHANES employs a stratified, multistage probability sampling design to assess the nutritional and health status of both adults and children in the United States. The survey covers multiple domains, including demographic information, socioeconomic status, dietary habits, health conditions, and physiological and laboratory data. All data are available on the NHANES website ( https://www.cdc.gov/nchs/nhanes/index.htm ). All protocols were approved by the NCHS Institutional Review Board (IRB), and written informed consent was obtained from all participants. This study analyzed data from the 2013–2018 NHANES cycles, which included a total sample size of 29,400 participants. During data screening, males, females outside the 20–45 age range ( n  = 25,545), pregnant women ( n  = 141), individuals without insulin resistance (IR)-related indicators ( n  = 2,332), participants missing infertility information ( n  = 131), and those with incomplete covariate data ( n  = 141) were excluded. In total, 1,100 participants who met the study’s inclusion criteria were retained for analysis. The selection process is illustrated in Fig.  1 . Fig. 1 Flowchart of participants selection Flowchart of participants selection The exposure variables in this study include seven indicators of insulin resistance: the insulin resistance (METS-IR), triglyceride-glucose index (TyG), triglyceride-glucose waist circumference (TyG-WC), triglyceride-glucose body mass index (TyG-BMI), triglyceride-glucose waist-to-height ratio (TyG-WHtR), homeostatic model assessment of insulin resistance (HOMA-IR), and the triglyceride/high-density lipoprotein ratio (TG/HDL). Calculation methods for each indicator are detailed below. METS-IR = ln [2 × FPG (mg/dL) + fasting serum TG (mg/dL)] × body mass index (BMI, kg/m 2 )/ln [HDL-C (mg/dL)]; TyG = ln [fasting serum TG (mg/dL) × FPG (mg/dL)/2]; TyG-WC = TyG index × waist circumference (cm); TyG-BMI = = TyG index × body mass index (BMI, kg/m2); TyG-WHtR = TyG index × waist(cm)/height(cm); HOMA-IR= [fasting serum insulin (µU/mL) × FPG (mg/dL)/405]; TG/HDL = fasting serum TG (mg/dL)/HDL-C (mg/dL). METS-IR = ln [2 × FPG (mg/dL) + fasting serum TG (mg/dL)] × body mass index (BMI, kg/m 2 )/ln [HDL-C (mg/dL)]; TyG = ln [fasting serum TG (mg/dL) × FPG (mg/dL)/2]; TyG-WC = TyG index × waist circumference (cm); TyG-BMI = = TyG index × body mass index (BMI, kg/m2); TyG-WHtR = TyG index × waist(cm)/height(cm); HOMA-IR= [fasting serum insulin (µU/mL) × FPG (mg/dL)/405]; TG/HDL = fasting serum TG (mg/dL)/HDL-C (mg/dL). The outcome variable in this study is self-reported infertility, derived from the reproductive health questionnaire. Question RHQ074 (“Have you ever attempted to conceive for at least one year without success?”) served as a proxy for infertility history. Participants who responded “yes” were classified as the infertility group, while those who responded “no” were categorized as the fertility history group. Other responses were treated as missing data. Covariate selection was informed by prior research and clinical practice guidelines. Continuous variables included age (years), while categorical variables encompassed race/ethnicity (Mexican American, Other Hispanic, non-Hispanic White, non-Hispanic Black, other races), education level (< 9th grade, 9-11th grade, high school graduate or GED, some college or AA degree, college graduate or higher), marital status (married/cohabiting, widowed/divorced/separated, never married), family income-to-poverty ratio (PIR: low  3.5), BMI (underweight < 18.5 kg/m², normal 18.5–24.9 kg/m², overweight/obese ≥ 25 kg/m²), smoking status (never smoked, former smoker, current smoker), alcohol use (yes/no), hypertension (yes/no), diabetes (yes/no), age at menarche (< 10 years, 10–14 years, 15–20 years), history of pelvic inflammatory disease treatment (yes/no), contraceptive use (yes/no), and female hormone use (yes/no). Alcohol use was assessed by asking if a person had consumed at least 12 drinks in the past year, with responses of “yes” classified as drinkers and “no” as non-drinkers. Hypertension was defined as either an average systolic blood pressure (SBP) ≥ 140 mmHg or diastolic blood pressure (DBP) ≥ 90 mmHg across three measurements, or a physician-diagnosed hypertension. Diabetes was defined as a physician diagnosis, use of insulin or oral hypoglycemic agents, fasting blood glucose (FBG) ≥ 126 mg/dL, or HbA1c ≥ 6.5% [ 17 ]. This study utilized the appropriate NHANES sample weights and incorporated the complex multistage sampling design in the analysis. Continuous variables are presented as means ± standard deviations (SD), and group comparisons were performed using t-tests (for normally distributed data) or Kruskal-Wallis tests (for non-normally distributed data). Categorical variables are reported as counts and percentages, with differences evaluated using chi-square tests or Fisher’s exact tests. Stratified multivariable weighted logistic regression was employed to examine the association between seven insulin resistance (IR) indicators and infertility. Model 1 was unadjusted; Model 2 was adjusted for age, race, education level, marital status, and PIR; and Model 3 included additional adjustments for BMI, smoking status, alcohol consumption, hypertension, diabetes, age at menarche, history of pelvic inflammatory disease treatment, contraceptive use, and history of female hormone therapy. The diagnostic value was evaluated using receiver operating characteristic (ROC) curves, with the area under the curve (AUC) used to quantify the predictive ability of IR-related indicators for infertility. Finally, smoothing curve fitting and threshold effect analysis were performed to further explore the relationship between high diagnostic performance IR indicators and infertility. All statistical analyses were conducted using R (version 4.3.3) and Empower ® software ( https://www.empowerstats.net/cn/ ). A p-value of < 0.05 was considered statistically significant.

This study is the first comprehensive analysis of the relationship between seven insulin resistance (IR) indicators—METS-IR, TyG, TyG-WC, TyG-BMI, TyG-WHtR, HOMA-IR, and TG/HDL—and the risk of infertility in U.S. women of reproductive age. Our findings reveal a significant positive correlation between four IR indicators (METS-IR, TyG-BMI, TyG-WC, and TyG-WHtR) and infertility risk. Specifically, as the levels of TyG-WC and TyG-WHtR increase, infertility risk increases linearly. In contrast, METS-IR and TyG-BMI show a nonlinear positive correlation with infertility, with the most pronounced effect observed when METS-IR is < 49.92 and TyG-BMI is < 290.18. Among the seven IR indicators, METS-IR is the most reliable predictor of infertility risk, followed by TyG-BMI, TyG-WC, and TyG-WHtR. In contrast, the diagnostic efficacy of TG/HDL, HOMA-IR, and TyG for predicting infertility is comparatively lower. An increasing body of evidence underscores the critical role of insulin resistance (IR) in female infertility and reproductive health. The metabolic disturbances linked to IR are central to the pathogenesis of polycystic ovary syndrome (PCOS). A cross-sectional study found a significant positive correlation between IR levels and the severity of menstrual irregularities in PCOS patients, with menstrual dysfunction being a major contributor to infertility in women [ 18 ]. Animal model studies further demonstrate that insulin-resistant mice exhibit substantial abnormalities in spindle formation and chromosome alignment in oocytes, suggesting that IR impairs reproductive health by disrupting the normal meiotic processes of oocytes [ 19 ]. In women without PCOS, IR also plays a crucial role in infertility and reproductive health. Research indicates that approximately 20.5% of non-PCOS infertile women have IR, which is significantly associated with reduced pregnancy rates during ovulation induction treatments. Women with IR show more limited follicular development and significantly lower chances of conception compared to their non-IR counterparts [ 20 ]. Furthermore, studies reveal that in non-PCOS women, IR leads to a delayed response to ovulation induction and impaired oocyte maturation, with a significantly lower proportion of viable embryos available for cryopreservation compared to women without IR [ 9 ]. IR adversely affects female fertility through several mechanisms. First, IR compromises oocyte quality by increasing oxidative stress and impairing mitochondrial function. OU et al. [ 19 ] demonstrated in an IR mouse model that oocyte glutathione levels were diminished, antioxidant defense mechanisms weakened, and intracellular reactive oxygen species (ROS) levels significantly elevated. This oxidative stress imbalance damages the oocyte membrane, proteins, and DNA, disrupting its structural integrity and impairing its quality and function [ 21 , 22 ]. Furthermore, elevated ROS levels induce mitochondrial DNA damage, decrease mitochondrial membrane potential, and inhibit normal energy metabolism, thereby reducing oocyte vitality and developmental potential [ 23 , 24 ]. In addition to affecting oocytes, IR impairs glucose transport in the endometrium, disrupting energy metabolism through downregulation of glucose transporter protein 4 (GLUT4) [ 25 ]. Moreover, IR triggers the release of pro-inflammatory cytokines, such as TNF-α and IL-6, leading to chronic inflammation, activation of oxidative stress, and endothelial dysfunction [ 26 , 27 ]. These pathological processes collectively diminish endometrial receptivity, thereby reducing the success rate of embryo implantation. Finally, IR exacerbates infertility risk by interfering with hormone secretion and embryo implantation. Hyperinsulinemia stimulates the ovaries and adrenal glands to secrete excessive androgens, directly inhibiting follicular development, causing follicular atresia and luteal dysfunction, and further impairing embryo implantation potential [ 28 , 29 ]. IR surrogate markers, including METS-IR, TyG, TyG-WC, TyG-BMI, TyG-WHtR, HOMA-IR, and TG/HDL, have been demonstrated as effective tools for assessing the severity of IR [ 11 – 15 ]. Consistent with prior research, this study identified a significant positive association between TyG-BMI and infertility risk, while HOMA-IR showed no significant correlation with infertility [ 16 ]. Notably, our findings indicate that women in the third quartile of the TyG index had a 1.86-fold increased risk of infertility compared to those in the first quartile. Additionally, women in the third quartile of the TG/HDL ratio exhibited a 2.56-fold higher infertility risk than those in the first quartile. However, the diagnostic efficacy of TyG for infertility remains limited, a finding supported by the research of Zhuang et al. [ 30 ]. Our study is the first to establish a significant positive correlation between METS-IR, TyG-WC, TyG-WHtR, and infertility, a relationship that remains significant after adjusting for multiple confounding factors. The interaction test results indicated that the positive correlation between METS-IR, TyG-BMI, TyG-WC, TyG-WHtR, and infertility was not significantly influenced by age group (all P-values > 0.05), further supporting the broad applicability of our conclusions across different age groups. Furthermore, ROC curve analysis confirms that, among seven IR markers, METS-IR is the most reliable predictor of infertility risk. TyG-BMI, TyG-WC, and TyG-WHtR, which incorporate anthropometric data, exhibit comparable diagnostic efficacy. METS-IR, an innovative tool introduced by Bello-Chavolla et al. in 2018, assesses insulin sensitivity without requiring insulin measurement. Its reliability has been validated using the high-insulin, normal-glucose clamp technique. Compared to the TG/HDL ratio and TyG index in the Mexican population, METS-IR demonstrates superior performance in predicting the incidence of type 2 diabetes [ 11 ]. A prospective cohort study demonstrated that METS-IR outperforms HOMA-IR in predicting the incidence of chronic kidney disease [ 31 ]. Numerous studies have demonstrated that METS-IR is a highly effective tool for predicting IR, type 2 diabetes, and the risk of metabolic syndrome [ 32 , 33 ]. The TyG-BMI, TyG-WC, and TyG-WHtR indices integrate insulin resistance with various anthropometric measurements, such as BMI, waist circumference, and waist-to-height ratio, offering a comprehensive approach to assessing metabolic health [ 34 ]. A cross-sectional study conducted on the Korean adult population shows that, compared to TyG, TyG-WC, and TyG-WHtR, TyG-BMI demonstrates superior accuracy in predicting IR [ 35 ]. These indices, which integrate anthropometric measurements with insulin resistance, have proven highly valuable in predicting the risk of metabolic syndrome, cardiovascular disease, and diabetes, offering essential tools for early detection and risk assessment [ 36 – 38 ].These indices not only quantify insulin resistance but also provide valuable insights into how body fat distribution and metabolic abnormalities influence fertility. We hypothesize that METS-IR offers a more comprehensive and accurate assessment of insulin resistance compared to TyG, TyG-WC, TyG-BMI, TyG-WHtR, HOMA-IR, and TG/HDL by incorporating multiple factors, such as fasting glucose, fasting triglycerides, BMI, and HDL-C. This multidimensional approach facilitates a more precise evaluation of insulin resistance, thereby enhancing the prediction of infertility risk. The strong predictive capacity of METS-IR for metabolic dysregulation, combined with its practicality and ease of use, makes it a valuable tool for the clinical assessment of infertility risk and the guidance of early interventions. Interestingly, we observed that the AUC values for BMI and WC were slightly higher than those for TyG-BMI and TyG-WC. This difference may stem from the fact that BMI and WC more directly capture the negative impact of fat mass on fertility. As alternative markers of IR, TyG-BMI and TyG-WC may be more closely associated with specific infertility subtypes, particularly polycystic ovary syndrome (PCOS) [ 39 ]. We hypothesize that infertility may also be attributed to non-metabolic factors, such as tubal abnormalities and endometriosis, which are linked to obesity but have a weaker association with IR. Consequently, BMI and WC may be more broadly applicable. In resource-limited settings, we propose that BMI and WC could serve as more cost-effective screening tools, especially in populations with high obesity rates. For patients with suspected metabolic-related infertility, TyG indices may provide useful insights; however, for infertility caused by non-metabolic factors, obesity-related markers may have broader applicability. Moreover, investigating IR-related markers in normal BMI patients is essential for understanding their association with infertility, as studies indicate that normal BMI patients with IR produce fewer high-quality embryos during IVF [ 40 ]. Nevertheless, this study did not find a significant correlation between IR-related markers and infertility in the normal BMI group, which may be attributed to factors such as sample size, population heterogeneity, methods of assessing IR, and the multifactorial nature of infertility. Future research should stratify and validate these findings according to underlying causes and assess the differential efficacy of these markers in various infertility subtypes to guide personalized evaluations. This study has several limitations. First, due to its cross-sectional design, it cannot establish a causal relationship between IR and infertility, and therefore, the results should be interpreted with caution. Second, the diagnosis of infertility is based on survey data derived from self-reports and interview responses, which may introduce recall bias or misclassification errors. Moreover, the NHANES database only includes data from the U.S. population, limiting comparisons with other populations. As a result, future research should further examine the relationship between IR and infertility across different populations and regions. Finally, this study did not consider other potential variables, such as genetic factors, environmental influences, lifestyle choices, and cultural background, which may impact the results. Despite these limitations, the findings remain robust and provide valuable scientific evidence to inform public health policies and management strategies related to infertility.

Infertility is defined as the inability of a couple to conceive after 12 months or more of regular, unprotected intercourse [ 1 ]. Globally, approximately 10–15% of couples experience infertility, with female factors contributing to about 40% of cases [ 2 , 3 ]. In recent years, lifestyle changes and other contributing factors have led to a continuous increase in infertility rates, making it a significant public health challenge that impacts both population health and social development [ 4 ]. Consequently, investigating the risk factors of infertility is essential for reducing its prevalence and economic burden. Insulin resistance (IR) refers to the diminished responsiveness of target tissues to physiological levels of insulin, leading to a reduced biological effect compared to normal insulin function [ 5 ]. As a key mechanism underlying obesity and metabolic syndrome, IR is recognized as a potential risk factor for female infertility [ 6 , 7 ]. Studies indicate that approximately 40-60% of women with polycystic ovary syndrome (PCOS) exhibit IR, which not only disrupts metabolic homeostasis but may also significantly impair fertility, contributing to infertility [ 8 ]. Furthermore, research has shown that IR is associated with in vitro fertilization (IVF) outcomes in non-PCOS women, particularly in relation to a decreased proportion of mature oocytes and poor embryo quality, further emphasizing the critical role of IR in infertility among non-PCOS women [ 9 ]. The hyperinsulinemic-euglycemic clamp technique is considered the gold standard for assessing IR [ 10 ]. However, due to its invasiveness, complexity, and time-consuming nature, several alternative IR assessment indices have been developed that are more convenient and widely applicable. These include the metabolic score for insulin resistance (METS-IR) [ 11 ], the triglyceride/glucose index (TyG) [ 12 ], the homeostasis model assessment of insulin resistance (HOMA-IR) [ 13 ], the triglyceride/high-density lipoprotein ratio (TG/HDL) [ 14 ], and various indices based on the product of TyG and anthropometric measurements [ 15 ], such as the triglyceride-glucose-waist circumference (TyG-WC), triglyceride-glucose-body mass index (TyG-BMI), and triglyceride-glucose-waist-height ratio (TyG-WHtR). All of these have been demonstrated to be effective tools for evaluating the severity of insulin resistance. Previous studies have established a significant correlation between TyG-BMI and infertility [ 16 ]. However, the relationship between other IR indicators and infertility remains unexplored. This study is the first to comprehensively examine the association between seven IR indicators—namely, METS-IR, TyG, TyG-WC, TyG-BMI, TyG-WHtR, HOMA-IR, and TG/HDL—and infertility risk, aiming to identify the most reliable predictors of infertility. By identifying potential metabolic biomarkers, this research provides valuable insights for the early detection and intervention of infertility.