Common genetic variation of the IGF2 gene and epithelial ovarian cancer risk in Chinese population

The demographic and clinical characteristics of the participants are summarized in Table  1 . A total of 213 patients with EOC and 203 control individuals were included in the study. The two groups were well-matched in terms of age, hypertension history, and diabetes status. Histologically, the majority of EOC cases (99.06%) were of the serous subtype. Among the patients with available clinicopathological information, 24 were classified as FIGO stage II, 162 as stage III, and 26 as stage IV. Regarding tumor differentiation, 10 cases were reported as well differentiated, while 183 exhibited moderate or poor differentiation. Table 1 Demographical information and clinical parameters of cases and controls Characteristics Case, n (%) Control, n (%) P value Overall 213 203 Age 1 (years) 50 (46–57) 47 (41–53) 0.118 Hypertension 0.432   Yes 9 (4.23) 12 (5.91)   No 204 (95.77) 191 (94.09) Diabetes 0.519   Yes 5 (2.35) 3 (1.48)   No 208 (97.65) 200 (98.52) FIGO stage   I 0 (0)   II 24 (11.27)   III 162 (76.06)   IV 26 (12.21)   Unknown 1 (0.46) Grade of differentiation (FIGO)   I (well) 10 (4.72)   II (moderate) + III (poor) 183 (86.32)   Unknown 19 (8.96) Histology   Serous 211 (99.06)   Non-serous 2 (0.94) 1 Median (25th-75th percentiles) n number; FIGO  International Federation of Gynecology and Obstetrics Demographical information and clinical parameters of cases and controls 1 Median (25th-75th percentiles) n number; FIGO  International Federation of Gynecology and Obstetrics All four SNPs were in Hardy–Weinberg equilibrium (HWE) in the control group, with results detailed in Supplementary Table S3. The allele and genotype frequencies of the IGF2 polymorphisms in both case and control groups are presented in Table  2 . A significant difference in genotype distribution was observed for rs11602347 C/G between patients and controls (CC: 9.04% vs. 0.7%; CG: 31.16% vs. 48.25%; GG: 59.80% vs. 51.05%; P  < 0.0001). Similarly, the genotype frequencies of rs4244808 T/G also differed significantly between groups (TT: 46.86% vs. 45.41%; TG: 31.43% vs. 43.24%; GG: 21.71% vs. 11.35%; P  = 0.010). In contrast, no significant differences were found in the allele frequencies of any of the four SNPs, nor in the genotype distributions of the other two SNPs. Table 2 Allele and genotype frequencies in cases and controls SNP Cases, n (%) Controls, n (%) P -value rs1004446G/A G 297 (70.38) G 276 (67.98) 0.455 A 125 (29.62) A 130 (32.02) Missing 4 Missing 0 GG 109 (51.66) GG 94 (43.31) 0.465 GA 79 (37.44) GA 88 (43.35) AA 23(10.90) AA 21 (10.34) Missing 2 Missing 0 rs10840447A/G A 51 (12.09) A 46 (11.33) 0.485 G 371 (87.91) G 360 (88.67) Missing 4 Missing 0 AA 2 (0.95) AA 2 (0.99) 0.926 AG 47 (22.27) AG 42 (20.69) GG 162 (76.78) GG 159 (78.33) Missing 2 Missing 0 rs11602347C/G C 98 (24.62) C 71 (24.83) 0.952 G 300 (75.38) G 215 (75.17) Missing 28 Missing 120 CC 18 (9.04) CC 1 (0.7) < 0.0001 CG 62 (31.16) CG 69 (48.25) GG 119 (59.80) GG 73 (51.05) Missing 14 Missing 60 rs4244808T/G T 219 (62.57) T 248 (67.03) 0.211 G 131 (37.43) G 122 (32.97) Missing 76 Missing 36 TT 82 (46.86) TT 84 (45.41) 0.010 TG 55 (31.43) TG 80 (43.24) GG 38 (21.71) GG 21 (11.35) Missing 38 Missing 18 Allele and genotype frequencies in cases and controls The associations between IGF2 polymorphisms and EOC risk are summarized in Table  3 . After adjustment for age, hypertension, and diabetes, the G allele of rs11602347 was associated with a significantly decreased risk of EOC under both dominant and additive genetic models (CG + GG vs. CC: adjusted OR = 0.09, 95% CI: 0.01–0.37, Benjamini-Hochberg adjusted P  = 0.002; GG vs. CC: adjusted OR = 0.11, 95% CI: 0.01–0.47, Benjamini-Hochberg adjusted P  = 0.008). For rs4244808, the GG genotype was associated with a significantly increased risk of EOC in the recessive model (GG vs. TG + TT: adjusted OR = 2.51, 95% CI: 1.35–4.65, Benjamini-Hochberg adjusted P  = 0.016). No significant associations were observed for the remaining SNPs. Table 3 Genotype among cases and controls and their association with EOC Risk Genetic model Genotype Cases, n (%) Controls, n (%) Crude OR (95% CI) P value Adjusted OR (95% CI) P value Benjamini-Hochberg adjusted P value rs1004446 Dominant GG 109 (51.66) 94 (43.31) Reference GA + AA 102 (48.34) 109 (56.69) 0.81 (0.55–1.19) 0.276 0.76 (0.50–1.14) 0.182 0.370 Recessive GA + GG 188 (89.10) 182 (89.66) Reference AA 23 (10.90) 21 (10.34) 1.06 (0.57–1.98) 0.855 0.92 (0.47–1.80) 0.814 0.910 Additive GG 109 (57.98) 94 (59.89) Reference GA 79 (42.02) 88 (40.11) 0.77 (0.51–1.17) 0.221 0.75 (0.49–1.15) 0.185 0.370 GG 109 (82.58) 94 (94.78) Reference AA 23 (17.42) 21 (5.22) 0.95 (0.49–1.81) 0.864 0.84 (0.41–1.70) 0.620 0.902 rs10840447 Dominant AA 2 (0.95) 2 (0.99) Reference AG + GG 209 (99.05) 201 (99.01) 1.03 (0.15–7.45) 0.969 1.14 (0.16–8.26) 0.899 0.910 Recessive AG + AA 49 (23.22) 44 (21.67) Reference GG 162 (76.78) 159 (78.33) 0.92 (0.58–1.45) 0.706 0.91 (0.56–1.47) 0.695 0.910 Additive AA 2 (4.08) 2 (4.55) Reference AG 47 (95.92) 42 (95.45) 1.12 (0.15–8.30) 0.912 1.19 (0.16–8.90) 0.863 0.910 AA 2 (1.18) 2 (1.24) Reference GG 162 (98.82) 159 (98.76) 1.02 (0.14–7.32) 0.985 1.12 (0.15–8.23) 0.910 0.910 rs11602347 Dominant CC 18 (9.04) 1 (0.7) Reference GC + GG 181 (90.96) 142 (99.3) 0.11 (0.0- 0.43) < 0.001 0.09 (0.01–0.37) < 0.001 0.002 Recessive GC + CC 80 (40.2) 70 (48.95) Reference GG 119 (59.80) 73 (51.05) 1.35 (0.88–2.08) 0.174 1.34 (0.84–2.15) 0.224 0.398 Additive CC 18 (22.5) 1 (1.43) Reference GC 62 (77.5) 69 (98.57) 0.08 (0.01–0.32) < 0.001 0.07 (0.01–0.29) < 0.001 0.001 CC 18 (13.14) 1 (1.35) Reference GG 119 (86.86) 73 (98.65) 0.13 (0.01–0.54) 0.002 0.11 (0.0-0.47) 0.001 0.008 rs4244808 Dominant TT 82 (46.86) 84 (45.41) Reference TG + GG 93 (53.14) 101 (54.59) 0.94 (0.62–1.43) 0.782 0.89 (0.57–1.38) 0.595 0.902 Recessive TG + TT 137 (78.29) 164 (88.65) Reference GG 38 (21.71) 21(11.35) 2.17 (1.21–3.87) 0.008 2.51 (1.35–4.65) 0.004 0.016 Additive TT 82 (59.85) 84 (51.22) Reference TG 55 (40.15) 80 (48.78) 0.70 (0.45–1.11) 0.134 0.62 (0.38–1.01) 0.053 0.141 TT 82 (68.33) 84 (80.00) Reference GG 38 (31.67) 21 (20.00) 1.85 (1.00-3.42) 0.047 1.47 (1.054–2.07) 0.028 0.090 OR odds ratio; CI confidence interval For all SNPs except rs11602347, ORs were calculated using standard logistic regression. For rs11602347, ORs were calculated using Firth’s penalized logistic regression to reduce sparse-data bias Additive model: refers to a genotypic ordinal additive model (three genotype levels), not a per-allele additive model Genotype among cases and controls and their association with EOC Risk OR odds ratio; CI confidence interval For all SNPs except rs11602347, ORs were calculated using standard logistic regression. For rs11602347, ORs were calculated using Firth’s penalized logistic regression to reduce sparse-data bias Additive model: refers to a genotypic ordinal additive model (three genotype levels), not a per-allele additive model SNP × age interaction tests were first performed to assess age-related effect modification on IGF2 SNP-EOC risk associations (Table  4 ), followed by age-stratified analyses based on the median age of controls (Table  5 ). Significant SNP×age interactions were identified for rs10840447 (adjusted interaction OR = 1.06, 95% CI: 1.01–1.10, P  = 0.011) and rs4244808 (adjusted interaction OR = 1.10, 95% CI: 1.01–1.21, P  = 0.030), with no significant interactions for rs1004446 ( P  > 0.05). Exploratory stratified analyses showed that a decreased risk of EOC was associated with the GG genotype of rs10840447 among individuals aged ≤ 47 years (adjusted OR = 0.47, 95% CI: 0.24–0.92, P  = 0.028). Conversely, an increased risk was observed with the TG + TT genotype of rs4244808 in subjects older than 47 years (adjusted OR = 3.52, 95% CI: 1.28–10.04, P  = 0.015). However, no significant association was found between the genotypes of rs1004446 or rs11602347 and EOC susceptibility in either age group. Table 4 Interaction effects between four SNPs and age SNP Interaction OR (95% CI) Interaction P -value Age main effect OR (95% CI) Age main effect P -value rs1004446 (GA + AA vs. GG)   Crude model 0.98 (0.94–1.02) 0.369 1.07 (1.04–1.11) < 0.001   Adjusted 1 model 0.98 (0.94–1.02) 0.394 1.08 (1.05–1.11) < 0.001 rs10840447 (GG vs. AA + AG)   Unadjusted 1.05 (1.01–1.10) 0.013 1.02 (0.99–1.06) 0.227   Adjusted 1 model 1.06 (1.01–1.10) 0.011 1.03 (0.99–1.06) 0.147 rs11602347 (GC + GG vs. CC)   Unadjusted NA   Adjusted 1 model NA rs4244808 (TG + TT vs. GG)   Unadjusted 1.10 (1.01–1.21) 0.032 1.06 (1.04–1.09) < 0.001   Adjusted 1 model 1.10 (1.01–1.21) 0.030 1.07 (1.04–1.10) < 0.001 OR odds ratio; CI confidence interval 1 Adjusted for hypertension and diabetes Interaction effects between four SNPs and age OR odds ratio; CI confidence interval 1 Adjusted for hypertension and diabetes Table 5 Association between the four IGF2 SNPs and EOC risk stratified by age Genotype Age group Cases, ( n ) Controls, ( n ) Crude OR (95% CI) P value Adjusted OR (95% CI) P value rs1004446 ≤ 47 34/38 66/56 0.76 (0.42–1.36) 0.355 0.72 (0.37–1.39) 0.323 GA + AA/GG > 47 68/71 43/38 0.85 (0.49–1.47) 0.551 0.86 (0.49–1.51) 0.608 rs10840447 ≤ 47 48/24 98/24 0.49 (0.25–0.95) 0.033 0.47 (0.24–0.92) 0.028 GG/AA + AG > 47 114/25 61/20 0.67 (0.34–1.30) 0.234 1.43 (0.73–2.82) 0.298 rs11602347 ≤ 47 57/9 86/0 NA NA NA NA GC + GG/CC > 47 124/9 56/1 0.25 (0.03–1.99) 0.156 0.23 (0.03–1.91) 0.175 rs4244808 ≤ 47 13/48 16/100 1.69 (0.75–3.80) 0.199 2.04 (0.80–5.21) 0.136 TG + TT/GG > 47 25/89 5/64 3.60 (1.31–9.90) 0.009 3.52 (1.28–10.04) 0.015 OR odds ratio; CI confidence interval Association between the four IGF2 SNPs and EOC risk stratified by age OR odds ratio; CI confidence interval We further performed exploratory analyses to evaluate the association between variant genotypes and clinicopathological characteristics of EOC, including FIGO stage and tumor differentiation (Table  6 ). After adjusting for age, hypertension, and diabetes, the GG genotype of rs11602347 was significantly associated with an increased odds of advanced FIGO stage among cases (III + IV vs. I+ II: adjusted OR = 2.67, 95% CI: 1.08–6.62, P  = 0.034). In contrast, the GG genotype of rs4244808 was associated with a reduced odds of advanced FIGO stage among cases (III + IV vs. I+ II: adjusted OR = 0.30, 95% CI: 0.10–0.90, P  = 0.032). No significant associations were observed for the rs1004446 or rs10840447 polymorphisms with these clinicopathological features. Table 6 Association between the four IGF2 SNPs and EOC risk stratified by clinicopathological characteristics Variable rs1004446G/A Crude OR (95% CI) P value Adjusted OR (95% CI) P value GA + AA GG FIGO stage   I+ II 14 10 Reference Reference   III 76 85 0.64 (0.27–1.52) 0.309 0.64 (0.26–1.55) 0.318   IV 11 14 0.56 (0.18–1.74) 0.316 0.80 (0.23–2.77) 0.722   III + IV 87 99 0.63 (0.27–1.49) 0.286 0.65 (0.27–1.56) 0.332 Tumor differentiation   Well 8 2 Reference Reference   Moderately and poorly 85 96 0.22 (0.05–1.07) 0.042 0.24 (0.05–1.19) 0.081   Variable rs10840447A/G Crude OR (95% CI) P value Adjusted OR (95% CI) P value GG AA + AG FIGO stage   I+ II 17 7 Reference Reference   III 124 37 1.38 (0.53–3.58) 0.507 1.19 (0.45–3.15) 0.719   IV 20 5 1.65 (0.44–6.15) 0.456 1.03 (0.24–4.41) 0.965   III + IV 144 42 1.41 (0.55–3.63) 0.473 1.17 (0.45–3.06) 0.751 Tumor differentiation   Well 7 3 Reference Reference   Moderately and poorly 138 43 1.38 (0.34–5.55) 0.654 1.09 (0.26–4.55) 0.902   Variable rs11602347C/G Crude OR (95% CI) P value Adjusted OR (95% CI) P value GG CC + GC FIGO stage   I+ II 9 14 Reference Reference   III 92 59 2.43 (0.91–6.75) 0.049 2.49 (1.00- 6.23) 0.051   IV 18 6 4.67 (1.34–16.24) 0.013 4.39 (1.11–17.35) 0.035   III + IV 110 65 2.63 (1.08–6.42) 0.029 2.67 (1.08–6.62) 0.034 Tumor differentiation   Well 6 4 Reference Reference   Moderately and poorly 105 65 1.08 (0.29–3.96) 0.911 1.08 (0.29–4.01) 0.912   Variable rs4244808T/G Crude OR (95% CI) P value Adjusted OR (95% CI) P value GG TG + TT FIGO stage   I+ II 7 9 Reference Reference   III 28 109 0.33 (0.11–0.96) 0.036 0.31 (0.10–0.92) 0.034   IV 3 18 0.21 (0.05–1.03) 0.046 0.36 (0.07–1.91) 0.229   III + IV 31 127 0.31 (0.11–0.91) 0.026 0.30 (0.10–0.90) 0.032 Tumor differentiation   Well 2 5 Reference Reference   Moderately and poorly 33 118 0.70 (0.13–3.77) 0.676 0.71 (0.13–3.89) 0.696 OR odds ratio; CI confidence interval; FIGO International Federation of Gynecology and Obstetrics Association between the four IGF2 SNPs and EOC risk stratified by clinicopathological characteristics OR odds ratio; CI confidence interval; FIGO International Federation of Gynecology and Obstetrics Linkage disequilibrium (LD) and haplotype analyses for the IGF2 polymorphisms (rs1004446, rs10840447, rs11602347, and rs4244808) were conducted using SHEsis software ( http://analysis.bio-x.cn/myAnalysis.php ). As presented in Supplementary Fig. S1, weak linkage disequilibrium was observed among the four SNPs. Furthermore, haplotype analysis revealed that the GGGG* haplotype (rs1004446/rs10840447/rs11602347/rs4244808) was associated with an elevated risk of EOC (OR = 2.16, 95% CI: 1.04–4.48, P  = 0.036), as detailed in Table  7 . Table 7 Haplotype analysis of polymorphisms (rs1004446, rs10840447, rs11602347 and rs4244808) at the IGF2 gene locus in the case and control groups Haplotype Case (freq) Control (freq) Fisher’s P Pearson’s P OR (95% CI) rs1004446/rs10840447/rs11602347/rs4244808 A G G G* 51.12 (0.150) 32.33 (0.128) 0.394 0.394 1.23 (0.76–1.98) A G G T* 40.65 (0.120) 44.64 (0.177) 0.060 0.060 0.64 (0.41–1.02) G A C G* 12.94 (0.038) 2.90 (0.011) 0.044 0.044 3.47(0.96–12.54) G A G G* 25.39 (0.075) 19.87 (0.079) 0.900 0.900 0.96 (0.52–1.77) G G C G* 3.26 (0.010) 7.99 (0.032) 0.055 0.055 0.30 (0.08–1.10) G G C T* 41.24 (0.121) 37.01 (0.147) 0.410 0.410 0.82 (0.51–1.32) G G G G* 28.29 (0.083) 10.39 (0.041) 0.036 0.036 2.16 (1.04–4.48) G G G T* 122.88 (0.361) 90.77 (0.360) 0.846 0.846 1.04 (0.74–1.46) OR odds ratio; CI confidence interval All frequencies < 0.03 will be ignored in haplotype analysis Haplotype analysis of polymorphisms (rs1004446, rs10840447, rs11602347 and rs4244808) at the IGF2 gene locus in the case and control groups OR odds ratio; CI confidence interval All frequencies < 0.03 will be ignored in haplotype analysis

This hospital-based case-control study included 213 patients with epithelial ovarian cancer (211 serous) and 203 healthy volunteers, who were consecutively recruited from Hunan Cancer Hospital in Changsha, China. All EOC patients were newly diagnosed and histologically confirmed, while those with secondary or recurrent EOC were excluded. Controls were healthy volunteers recruited from the hospital physical examination center, with no personal history of cancer, clinical symptoms, or genetic disorders. Tumor differentiation grade and stage were classified according to the International Federation of Gynecology and Obstetrics (FIGO) criteria. The study protocol was approved by the Ethics Committee of Central South University, and written informed consent was obtained from all participants. Four SNPs (rs1004446, rs10840447, rs11602347, and rs4244808) were selected for this study based on the following two criteria: (1) a minor allele frequency (MAF) greater than 0.05 in the CHB population, and (2) a pairwise linkage disequilibrium (LD) threshold of r² < 0.8, as analyzed using Haploview software (version 4.2). The protocol for genomic DNA extraction using the Wizard ® Genomic DNA Purification Kit (Promega, USA) followed the procedure described in our previous study [ 12 ]. Genotyping was performed using the Sequenom MassARRAY platform (BioMiao Biological Technology Co., Ltd., Beijing, China). Primer and probe sequences are available in our earlier publication [ 12 ]. For quality control, approximately 10% of samples were randomly selected and re-genotyped, yielding a concordance rate greater than 99.9%. Genotyping call rates for each SNP in cases and controls are listed in Supplementary Table S1. The overall genotype-level call rate across all samples exceeded 90%. Rs11602347 and rs4244808 had relatively lower genotyping call rates, with lower rates observed in controls for rs11602347 and in cases for rs4244808. Age, hypertension, and diabetes were comparable between participants with missing genotypes and those without, with no evidence that missingness was related to case–control status. Minor missing data occurred due to competitive inhibition within the high-throughput genotyping reaction system; specifically, as multiple SNPs were genotyped simultaneously, competitive binding among primers in the reaction system led to relatively low binding efficiency between the specific primers of some SNPs and the DNA template, resulting in genotyping signals below the instrument’s detection threshold. This is a common technical limitation in high-throughput genotyping experiments rather than a systematic bias. We cannot fully exclude potential impacts of missing data on the robustness of the results. To verify the reliability of the missing data, we performed sensitivity analysis using rs11602347 and rs4244808 (the two SNPs with the highest missing rates), comparing allele frequency, genotype frequency, Hardy–Weinberg equilibrium (HWE), and disease risk between samples with and without missing genotypes. All odds ratios in this sensitivity analysis were adjusted for age, hypertension, and diabetes. The statistical conclusions were consistent across all dimensions, with only minor fluctuations in numerical values and significance levels, and no reversal of conclusions or change in effect direction, with the detailed analysis results shown in Supplementary Table S2, suggesting that missing data were unlikely to introduce substantial systematic bias. All statistical analyses were performed using Stata software (version 18.0; STATA Corp., College Station, TX, USA) and R 4.5.2. A two-sided P-value < 0.05 was considered statistically significant. Differences in the distribution of categorical variables and genotype frequencies between cases and controls were evaluated using the Pearson χ² test, except when expected cell counts were small (e.g., for rs11602347), in which case Fisher’s exact test was used instead. The goodness-of-fit χ² test was applied to assess deviations from Hardy–Weinberg equilibrium among controls. Associations between polymorphisms and EOC risk were estimated by calculating odds ratios (ORs) and their corresponding 95% confidence intervals (CIs). The crude OR was calculated using the Woolf approximation method. For all SNPs except rs11602347, adjusted ORs were derived from standard logistic regression models controlling for age, hypertension, and diabetes. For rs11602347, the CC genotype was extremely rare in the control group (only one subject). Although Firth’s penalized logistic regression was used to alleviate bias related to sparse data, the effect estimates remain unstable and should be interpreted with considerable caution. The Benjamini–Hochberg (BH) false discovery rate (FDR) method was applied to correct for multiple comparisons across all 4 SNPs and the three genetic models.

In the present study, we found preliminary and hypothesis-generating evidence for the first time that rs11602347 C > G and rs4244808 T > G were associated with a significantly decreased and increased risk of EOC, respectively. Younger individuals with the rs10840447 GG genotype have a reduced risk of EOC, while older individuals with the rs4244808 TG + TT genotype have an elevated risk of EOC. Furthermore, among EOC cases, the variant GG genotype of rs11602347 was associated with increased odds of advanced FIGO stage, whereas the GG genotype of rs4244808 was associated with reduced odds of advanced FIGO stage. Additionally, the GGGG* haplotype (rs1004446/rs10840447/rs11602347/rs4244808) was related to an increased risk of EOC. IGF2 is a multifunctional regulator of cell proliferation, primarily secreted by the liver. As one of the earliest identified endogenous imprinted genes, its expression and regulation do not follow Mendelian inheritance patterns [ 19 – 21 ]. With cellular aging, aberrant compartmentalized methylation on the maternal chromosome leads to a loss imprinting (LOI). This LOI results in IGF2 overexpression and confers a heightened responsiveness to IGF2 signaling pathways [ 22 ]. Physiologically, the IGF axis plays crucial roles in ovarian physiology, including folliculogenesis, ovulation, follicle growth and oocyte maturation [ 23 , 24 ]. However, in malignant cells, IGF2 is frequently overexpressed through diverse mechanisms [ 6 , 25 ] fetal promoter reactivation, maternal LOI, hypermethylation of the IGF2 / H19 locus, and downregulation of tumor suppressors like p53 and WT1. Consequently, elevated IGF2 expression has recently been established as a predictor of poor prognosis in epithelial ovarian cancer and is associated with its most aggressive subtypes. Furthermore, serum IGF2 levels constitute a component of a four-marker panel for the early detection of this disease [ 26 , 27 ]. SNPs can play a significant role in the pathogenesis of various diseases by altering gene transcription efficiency or protein structure and function. The rs1004446 SNP, located within the differentially methylated region (DMR) of the IGF2 , is a widely studied locus and has been implicated in the risk of multiple complex diseases across numerous large-scale genetic and clinical studies. In T1D research, Concannon et al. initially identified an association between rs1004446 and the disease at a genome-wide significant level ( P = 2.5 × 10⁻⁹) through a genome-wide linkage scan and family-based association analysis [ 16 ]. Subsequent studies further confirmed that the A allele of rs1004446 significantly reduces the risk of developing T1D (HR = 0.71, P = 0.006) [ 28 ]. Moreover, among children carrying high-risk HLA genotypes, this locus was significantly associated with a reduced risk of developing islet autoantibodies (HR = 0.77, 95% CI: 0.66–0.90) [ 29 ]. In oncology-related studies, rs1004446 also demonstrates considerable influence. A large nested case-control study revealed that the A/G variant at this locus decreases the risk of endometrial cancer (OR = 0.80, 95% CI: 0.68–0.94) [ 15 ]. Furthermore, a study on prostate cancer (PCa) mortality risk found that rs1004446 is in high linkage disequilibrium with another SNP, rs3741211 (r² = 1). Each additional risk allele was associated with a significantly increased risk of PCa-specific mortality (HR = 1.19, Ptrend = 0.003); however, this variant was not associated with tumor characteristics such as stage or Gleason score [ 30 ]. At the molecular phenotype level, a study on circulating IGF levels indicated that the copy number of the C allele of rs1004446 is positively correlated with plasma concentrations of IGF1 and IGF2, suggesting a potential role in regulating IGF expression [ 31 ]. In a Chinese population, this polymorphism was also associated with alterations in triglyceride levels among patients with coronary artery disease (CAD) and hypertension [ 32 ]. The GG genotype was significantly more frequent in CAD patients with hypertension ( P = 0.027) [ 32 ], implying that it may represent a genetic susceptibility factor for dyslipidemia in this subgroup. Additionally, cross-age group analysis showed that the minor allele of rs1004446 is consistently associated with a reduced risk of diabetes in both younger and older populations (HR = 0.67) [ 33 ]. However, in our study, stratification by tumor differentiation revealed that among patients with moderately to poorly differentiated ovarian cancer, those carrying the rs1004446 A allele showed a suggestive significant association with a reduced risk of ovarian cancer. However, after adjusting for age, hypertension, and diabetes, this association was no longer statistically significant. These four SNPs within the IGF2 gene region are listed in Supplementary Table S4. To our knowledge, no previous studies have reported the associations of IGF2 SNPs (rs10840447, rs11602347, and rs4244808) with disease susceptibility, and no functional data are currently available for these variants. In this study, we demonstrate for the first time that the rs10840447 G > A variant is not significantly associated with EOC risk, whereas the rs11602347 C > G variant significantly decreases EOC risk and the rs4244808 T > G variant significantly increases it. Age-stratified analysis revealed that the rs10840447 GG genotype was associated with reduced EOC risk in the younger subgroup. Interestingly, among older individuals, carriers of the rs4244808 T allele exhibited significantly elevated EOC risk. This discrepancy may be attributed to several factors. The rs4244808 G risk allele might be associated with more aggressive tumor phenotypes, leading to earlier mortality; thus, carriers of this allele could be under-represented in the older age group due to selective survival, resulting in an apparent protective effect of the T allele among older patients. Alternatively, the molecular pathway involving this SNP may be specifically relevant in younger patients, whereas EOC pathogenesis in older individuals may operate through distinct mechanisms, diminishing or reversing the SNP’s effect. Furthermore, age-dependent differences in environmental exposures, such as smoking, diet, or infections, could interact with the SNP, thereby modulating its penetrance and direction of association. Exploratory stratified analysis based on tumor biological characteristics revealed that among EOC patients, carriers of the rs11602347 GG genotype had significantly higher odds of advanced FIGO stage, whereas those with the rs4244808 GG genotype exhibited significantly lower odds of advanced FIGO stage. Notably, these findings show opposite effect directions compared with overall risk: the rs11602347 G allele was protective for EOC susceptibility but associated with advanced stage, whereas the rs4244808 GG genotype increased overall risk yet was linked to lower odds of advanced stage disease. These patterns suggest that the IGF2 variants may play distinct roles in disease initiation versus progression. In addition, limited sample sizes in subgroup analyses may reduce statistical stability and contribute to inconsistent results. These findings are preliminary and should be interpreted cautiously. Potential explanations for these findings include the high heterogeneity of EOC, which comprises distinct histological subtypes with varying molecular profiles. The rs11602347 C and rs4244808 G alleles may primarily drive tumorigenesis in low-grade tumors but have limited relevance in high-grade disease. Unmeasured confounding factors strongly associated with FIGO stage might also contribute to these patterns. Moreover, the potential under-representation of certain patient subgroups in high-grade cohorts, owing to more aggressive disease progression and higher mortality, may introduce selection bias. Our results underscore the complexity of tumor biology and genetics, suggesting that the effects of genetic variants such as these SNPs are not uniform but may depend on contextual factors such as age or tumor-specific features including grade and stage. This study has several limitations that should be acknowledged. First, the statistical power was limited for certain genetic variants due to a relatively small sample size, particularly in subgroup and stratified analyses. Second, although controls were healthy individuals from the physical examination center rather than hospitalized patients, we cannot fully rule out the potential influence of unmeasured lifestyle or environmental factors. Nevertheless, the use of healthy physical examination controls largely minimizes selection bias related to metabolic or endocrine disorders. Third, residual confounding represents another major limitation. Due to the retrospective design and limited availability of detailed clinical data, we were unable to adjust for established EOC risk factors beyond age, hypertension, and diabetes, including parity, oral contraceptive use, endometriosis, family history of cancer, body mass index (BMI), and hormonal exposures. The absence of these variables means that unmeasured confounding could have substantially affected the magnitude and potentially the direction of the reported associations. Therefore, our findings should be considered hypothesis-generating rather than confirmatory. Fourth, the dominant, recessive, and additive models assume specific inheritance patterns that may not fully reflect the true biological mechanisms of IGF2 variants. The extreme sparsity of certain genotypes (e.g., the CC genotype of rs11602347 in controls) can lead to unstable odds ratio estimates, even with Firth’s penalized logistic regression, warranting cautious interpretation. Fifth, the four IGF2 polymorphisms were selected solely based on minor allele frequency (MAF > 0.05) and linkage disequilibrium (r 2  < 0.8) in the CHB population, without prior functional validation. No mechanistic or expression experiments were conducted to confirm the biological effects of these variants. Thus, we cannot confirm whether the studied SNPs are causal or merely tag SNPs linked to true functional variants. Finally, as the study population was exclusively of Chinese ancestry and predominantly serous EOC subtype, the generalizability of the findings to other ethnic groups and other EOC histological subtypes remains uncertain and warrants further validation. In summary, this hospital-based case-control investigation provides preliminary and hypothesis-generating evidence that specific IGF2 polymorphisms (rs1004446, rs10840447, rs11602347, and rs4244808) are associated with serous EOC risk and may be involved in the pathogenesis of the disease. As the vast majority of cases were serous epithelial ovarian cancer, the present findings are primarily relevant to this histological subtype and may not be generalizable to other EOC subtypes. Further validation through larger, multi-ethnic prospective studies is essential to corroborate these associations and to explore their potential clinical implications.

Ovarian cancer (OC) is a highly aggressive gynecological malignancy, ranking as the eighth most commonly diagnosed cancer and the eighth leading cause of cancer-related death among women worldwide [ 1 ]. Global estimates indicate 313,959 new cases and 207,252 deaths attributable to ovarian cancer in 2020, with incidence and mortality rates continuing to rise [ 1 ]. Histologically, epithelial ovarian cancer (EOC) represents the most prevalent subtype, accounting for over 90% of cases [ 2 ]. Despite improvements in cytoreductive surgery, platinum-based chemotherapy, and targeted maintenance therapy, the 5-year survival rate remains below 29%, largely due to frequent late-stage diagnosis and high rates of recurrence [ 3 ]. The identification of biomarkers for early detection may improve diagnostic timeliness, reduce mortality, and potentially facilitate targeted therapeutic strategies for ovarian cancer. Insulin-like growth factor 2 (IGF2), encoded by a 150-kb locus on chromosome 11, is a secreted peptide consisting of 67 amino acids that is highly expressed primarily during early embryonic and fetal development [ 4 ]. This genomic region contains two imprinted genes, IGF2 and H19 , which are paternally and maternally expressed, respectively [ 5 ]. While Igf2 encodes the IGF2 protein, H19 is transcribed to produce a long non-coding RNA that acts as a negative regulator of cell proliferation [ 6 ]. Although H19 has been implicated as a putative tumor suppressor, its functions in carcinogenesis are multifaceted and influenced by contextual cues [ 7 ]. The H19 locus encodes miR-675, a microRNA that post-transcriptionally represses IGF1 receptor (IGF1R) expression, leading to attenuation of IGF2 signaling via the IGF1R pathway [ 8 ]. IGF2 has been shown to play important roles in cell proliferation, survival, metabolism, and differentiation [ 9 ], dysregulation is frequently observed in various malignancies, including ovarian, breast, pancreatic, colorectal, and esophageal cancers, as well as osteosarcoma [ 4 , 5 ]. Moreover, accumulating evidence indicates that IGF2 not only promotes metastatic dissemination but also contributes to tumor immunoevasion [ 9 ] and therapeutic resistance [ 10 – 12 ]. Single nucleotide polymorphisms (SNPs), the most prevalent form of genetic variation, can influence gene expression and function, serving as a major source of individual phenotypic diversity. Genetic polymorphisms have been implicated in the pathogenesis of numerous diseases, including a wide spectrum of cancers. For instance, the variant rs55705857 has been identified as a causal allele associated with molecular pathways driving low-grade glioma, conferring a sixfold increased risk [ 13 ]. A recent study reported that rs8570 G > C, located in the 3’-untranslated region (3’UTR) of RRAS2 , is linked to elevated RRAS2 expression and plays a key role in the development of multiple breast cancer subtypes, particularly post-partum-associated triple-negative breast cancer [ 14 ]. The IGF2 polymorphism rs1004446 has been associated with risks of endometrial cancer and type 1 diabetes (T1D) [ 15 , 16 ]. Additionally, a genome-wide association study (GWAS) conducted in a Japanese population identified two novel variants, rs79219469 C > T in Lnc02183 and rs567534295 C > T in BRCA1 , associated with ovarian cancer [ 17 ]. Regarding IGF2 , our previous investigation revealed that four SNPs (rs2525885, rs2839698, rs3741206, rs3741219) within the IGF2 / H19 gene locus are candidate susceptibility variants for EOC risk in a Chinese population. However, since known common genetic variants account for less than 4% of EOC risk [ 18 ], additional susceptibility loci likely remain to be discovered. To further advance the understanding of genetic susceptibility to EOC, we conducted a case-control study to investigate the associations of four IGF2 polymorphisms (rs1004446, rs10840447, rs11602347, and rs4244808) and their haplotypes with EOC risk. Furthermore, we evaluated whether these variants were associated with clinicopathological characteristics. Our findings could provide new insights for the prevention and treatment of EOC.

Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.