Feasibility of SNP Genotyping Using Dried Blood Spot Samples Collected in an Epidemiological Study and Its Integration With Genetic Risk Analysis for Endometriosis

ABSTRACT

Purpose

This study evaluated the feasibility of single‐nucleotide polymorphism (SNP) genotyping using dried blood spot (DBS) samples stored under various conditions, based on the genotyping success rate and concordance with whole blood results. It also examined associations between selected SNPs and endometriosis risk in Japanese women.

Methods

DBS samples from 41 cohort participants and 28 hospital patients were used to assess genotyping feasibility. Five endometriosis‐associated SNPs—rs10965235, rs12700667, rs12024204, rs16826658, and rs801112—were genotyped in 37 cases and 144 controls. Genotype distributions were evaluated for Hardy–Weinberg equilibrium (HWE) using Pearson's χ ² test or, when appropriate, Fisher's exact test, with a significance threshold of p < 0.05. Fisher's exact test was used for association analysis.

Results

SNP genotyping for rs12700667 showed 100% success and complete concordance between DBS and whole blood samples under all storage conditions. Four of five SNPs met HWE, while rs10965235 significantly deviated from it (p = 0.0225). The CC genotype of rs10965235 was potentially associated with lower endometriosis risk (odds ratio: 0.19), although this was not statistically significant after correction.

Conclusions

DBS is a robust DNA source for SNP genotyping under various conditions and suitable for mail‐based epidemiological studies. Population‐specific validation is essential when applying GWAS findings.

1. Introduction

Endometriosis is a chronic gynecological disorder that affects approximately 5%–10% of women of reproductive age and is characterized by the ectopic presence of endometrial‐like tissue outside the uterus [1]. Despite extensive research, the precise etiology of endometriosis remains unclear. The most widely accepted hypothesis involves retrograde menstruation; however, other factors, including genetic predisposition, hormonal imbalances, immune dysfunction, inflammation, and environmental influences, also contribute to its pathogenesis [2, 3, 4]. Recent studies have highlighted the roles of oxidative stress, coelomic metaplasia, benign metastasis, and stem cell involvement in disease progression [5, 6]. Understanding these complex pathogenic pathways is critical for developing effective treatments for endometriosis and improving the quality of life for affected women [1].

Genetic factors play a substantial role in susceptibility to endometriosis, with multiple studies investigating the association between single‐nucleotide polymorphisms (SNPs) and this disease. SNPs are among the most common genetic variations in the human genome, occurring approximately every 100–300 base pairs [7]. These variations can influence disease susceptibility, gene expression, and protein function [8, 9]. SNPs are valuable genetic markers used in genome‐wide association study (GWAS) and genetic epidemiology research [10, 11]. Their stability over time and potential for identifying genetic predispositions make them particularly useful in studying complex diseases, including endometriosis [12, 13]. However, challenges such as population stratification, the need for multiple testing, and functional validation must be addressed to ensure the robustness of SNP‐based studies [14].

SNPs in genes related to estrogen metabolism, inflammation, and cell adhesion have been implicated in the risk of endometriosis [15]. For instance, variants in IL1A (interleukin 1 alpha) have demonstrated strong associations with endometriosis in both Japanese and European populations [15]. Additionally, polymorphisms in the HSD17ß1 (17ß‐hydroxysteroid dehydrogenase type 1), IL10 (interleukin 10), ESR2 (estrogen receptor beta), and PGR (progesterone receptor) genes have been identified as potential genetic markers for endometriosis in various populations [16, 17, 18, 19, 20, 21]. Furthermore, SNPs in WNT4 (Wnt family Member 4), VEZT (vezatin, adherens junction transmembrane protein), and FSHB (follicle stimulating hormone beta subunit) have been linked to endometriosis risk in a Greek cohort, while polymorphisms in CDH1 (cadherin 1, type 1, E‐cadherin [epithelial]) (encoding E‐cadherin) have been associated with the disease in Chinese women [22, 23]. Despite these findings, the complex genetic underpinnings of endometriosis remain incompletely understood, with conflicting results being reported across studies [24].

Large‐scale epidemiological studies require efficient and minimally invasive methods for sample collection, particularly when analyzing genetic markers such as SNPs. Dried blood spot (DBS) on filter paper has emerged as a valuable tool for epidemiological research owing to its ease of collection, cost‐effectiveness, and simple transportation and storage [25]. DBS has been widely used for newborn screening, infectious disease surveillance, and genetic studies, demonstrating the stability of DNA for SNP analysis [26]. This method has facilitated seroepidemiological studies, genotyping, and metabolic disease screening, making it ideal for large‐scale population research [27, 28, 29]. Recent advances in laboratory techniques have further expanded the application of DBS for genetic research, enabling whole‐genome amplification, and precise SNP detection [30].

Despite the increasing utility of DBS in genetic epidemiology, the feasibility of accurately measuring SNPs from DBS samples remains an area of ongoing investigation. The present study aimed to validate the accuracy of SNP measurements using DBS and assess their applicability in large‐scale epidemiological studies. Specifically, we explored the association between SNPs and endometriosis by using DBS‐collected blood samples from participants in an epidemiological survey. Additionally, we collected blood samples from patients who were hospitalized for surgery due to uterine fibroids and endometriosis (ovarian chocolate cysts) at our facility's Department of Obstetrics and Gynecology. These blood samples were used to measure five target SNPs (rs10965235 [31], rs12024204 [32], rs12700667 [32], rs16826658 [31], and rs801112 [33]), which were previously suggested to be associated with endometriosis. This allowed us to examine whether there were any biases in SNP distributions within the studied population. This study provides fundamental data for future genetic research on endometriosis and contributes to our understanding of its genetic basis.

2. Materials and Methods

2.1. Examination of Feasibility of SNP Genotyping Using DBS Samples

2.1.1. Study Participants

This study was approved by the institutional ethics committee (Approval number: 2018‐136, Approval date: 7/11/2018). The study included two groups: a control group and a disease group. The control group consisted of 41 participants from the Gunma Nurses' Health Study (GNHS) sub‐cohort, specifically from the “Epidemiological Study on Sleepiness and Lifestyle Habits in Nurses [34].” The disease group included 13 patients diagnosed with endometriosis and 15 patients diagnosed with uterine fibroids at the Department of Obstetrics and Gynecology, Gunma University Hospital. All participants were of Japanese ethnicity. Written informed consent was obtained from all participants prior to blood collection. Basic demographic data, including age and other clinical information, were obtained from either the epidemiological survey database or the patients' medical records.

2.1.2. Sample Collection and Storage

Blood samples were collected from all participants and applied to filter paper to create DBS samples, simulating the conditions under which DBS samples would be mailed from participants in large‐scale epidemiological studies. In addition, fresh whole blood samples were analyzed immediately after collection to serve as a reference for SNP genotyping. To evaluate the feasibility of SNP genotyping from DBS under different storage conditions, DBS samples were stored at −30°C, 4°C, room temperature, and 70°C for 24, 48, and 72 h. These conditions were selected to test the robustness of SNP detection under extreme temperatures and prolonged storage durations, reflecting potential challenges in real‐world sample collection and transport. At each time point, a 3‐mm‐diameter circular section of the DBS was punched out for DNA extraction. SNP genotyping of rs12700667 was conducted on both DBS and fresh blood to assess concordance between them.

Based on the principle that fluorescence intensity in real‐time polymerase chain reaction (real‐time PCR) reflects the amount of PCR products, the endpoint fluorescence signals were compared across storage conditions to evaluate the stability of the template DNA.

2.2. DNA Extraction and SNP Genotyping

DNA was extracted by mixing either 5 μL of whole blood or a 3‐mm DBS punch with 20 μL of lysis solution (Thermo Fisher Scientific, Waltham, MA, USA) and incubating at room temperature (20°C–25°C) for 3 min. Subsequently, 20 μL of DNA stabilizing solution was added. A 20 μL reaction mixture containing TaqMan Genotyping Assay reagents was used to genotype SNPs via real‐time PCR. Allele‐specific fluorescence signals were measured to determine genotypes.

2.3. Examination of SNPs Associated With Endometriosis Onset

2.3.1. Study Participants

This study was approved by the institutional ethics committee. Participants included 41 epidemiological study participants and 140 patients admitted to our department for surgical treatment. The disease group consisted of 37 patients diagnosed with endometriosis via surgery or MRI‐confirmed endometriotic cysts. The control group included the 41 epidemiological study participants and 103 patients with uterine fibroids or benign ovarian tumors who did not have endometriosis. All participants were of Japanese ethnicity. Written informed consent was obtained from all participants prior to blood collection. Basic demographic data, including age and other clinical information, were obtained from either the epidemiological survey database or the patients' medical records.

2.3.2. SNP Analysis Method

SNP genotyping was conducted using DNA extracted from fresh whole blood samples obtained from all participants. The following SNPs were analyzed: rs10965235, rs12024204, rs12700667, rs16826658, and rs801112. Genotyping was performed using the same TaqMan Genotyping Assay protocol as described above, and allele‐specific fluorescence signals were measured by real‐time PCR to determine SNP genotypes. The TaqMan Genotyping Assay was performed using real‐time PCR, and the fluorescence signal generated by allele‐specific probes was measured to determine SNP genotypes.

2.4. Statistical Analysis Methods

All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

2.4.1. Statistical Comparison of Baseline Characteristics

Descriptive statistics were calculated to summarize the background characteristics of study participants. Continuous variables such as age were expressed as means and standard deviations, while categorical variables including smoking history and alcohol intake were summarized using counts and percentages. Comparisons between the control and disease groups were conducted using Student's t‐test for continuous variables and either Pearson's χ ² test or Fisher's exact test for categorical variables, depending on the expected cell counts. p values < 0.05 were considered statistically significant.

2.4.2. Hardy–Weinberg Equilibrium

To confirm that allele and genotype frequencies were stable within the study population, Hardy–Weinberg equilibrium (HWE) was tested using Pearson's χ ² test at a significance level of 0.05. For SNPs with low expected genotype counts (e.g., rs10965235, where no AA genotype was observed), Fisher's exact test was additionally applied to ensure robustness of the assessment.

2.4.3. Genotype Frequency and Odds Ratio for Endometriosis

For each SNP, a 2 × 2 or 3 × 2 contingency table was constructed to compare genotype frequencies between the case and control groups. To address the presence of zero cells and avoid division‐by‐zero errors, the Haldane–Anscombe correction was applied by adding 1 to all cells in the contingency table. Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated using the corrected counts. The reference genotype was set as the homozygous wild‐type allele (e.g., AA), which was compared with the heterozygous and homozygous variant genotypes (e.g., AG, GG). The logarithm of the OR was used to compute standard errors (SEs), and Wald tests were employed to calculate two‐sided p‐values. Because five independent SNPs were analyzed, multiple testing was controlled for by Bonferroni correction; the adjusted significance threshold was set at α = 0.05/5 = 0.010. Accordingly, p < 0.010 was regarded as statistically significant. Given the limited sample size and the low frequency of some genotypes, logistic regression analysis incorporating potential covariates such as age and body mass index (BMI) was not feasible. Therefore, Fisher's exact test was used as a robust alternative to assess genotype–disease associations.

2.4.4. Comparison of Risk Allele Frequencies

To evaluate the consistency of our findings with previous GWAS, we compared the risk allele frequencies observed in our study with those reported in GWAS datasets. For each SNP, we analyzed allele frequencies separately in the control and disease groups. The risk allele frequencies obtained in our study were compared against the frequencies reported in GWAS for both groups. The comparisons included the SNPs rs10965235, rs12024204, rs12700667, rs16826658, and rs801112 [31, 32, 33].

Risk allele frequencies in our study were calculated on the basis of genotyping results from real‐time PCR using the TaqMan Genotyping Assay. The statistical significance of differences in allele frequencies between our dataset and the GWAS data was assessed using Pearson's χ ² test for independence. For each SNP, two contingency tables (2 × 2) were constructed:

1. Comparison of control group frequencies (GWAS vs. our study).

2. Comparison of disease group frequencies (GWAS vs. our study).

Each contingency table contained the number of individuals with the risk allele and those without it, assuming a hypothetical population size of 1000 for standardization. If the expected cell count was too low, Fisher's exact test was considered as an alternative. To account for multiple comparisons across the ten 2 × 2 tables (five SNPs × two groups), Bonferroni correction was applied, and a corrected p‐value threshold of 0.005 (0.05/10) was used to determine statistical significance.

Accordingly, p‐values below 0.005 were considered statistically significant, indicating a meaningful difference between our study's allele frequencies and those reported by GWAS. These analyses were intended to determine whether the observed discrepancies were due to methodological differences, population‐specific genetic variability, or sample selection bias.

3. Results

3.1. Baseline Characteristics of Study Population

Baseline characteristics of the participants are presented in Table 1. The mean age was significantly lower in the disease group (38.2 ± 5.91 years) than in the control group (42.1 ± 10.28 years, p = 0.0033), primarily due to the inclusion of older participants in the GNHS subgroup. The GNHS controls, who were population‐based, had a higher mean age (49.8 ± 10.20 years), whereas the hospital‐based controls had a mean age of 39.0 ± 8.60 years, which was more comparable to that of the disease group. Body mass index (BMI) was comparable between the overall control and disease groups (22.5 ± 3.48 vs. 22.1 ± 4.48, p = 0.6160), and no significant differences in this variable were observed in subgroup comparisons. Smoking history and alcohol intake did not differ significantly between the groups. Smoking was reported in 16.7% of controls and 21.1% of cases (p = 0.6430), while alcohol intake was reported in 10.4% of controls and 18.4% of cases (p = 0.1654).

TABLE 1.

Statistical comparison of baseline characteristics.

	Control group (n = 144)	Disease group a (n = 37)	p b , c
Age (years) (mean ± SD)	42.1 ± 10.28	38.2 ± 5.91	0.0033
Age class, n (%)
20–29	12 (8.3)	4 (10.8)	0.0144
30–39	51 (35.4)	15 (40.5)
40–49	55 (38.2)	18 (48.6)
50—	26 (18.1)	0 (0.0)
BMI (mean ± SD)	22.5 ± 3.48	22.1 ± 4.48	0.6160
Smoking history, n (%)	24 (16.7)	8 (21.1)	0.6430
Alcohol intake, n (%)	15 (10.4)	7 (18.4)	0.1654

Abbreviation: BMI, body mass index; NA, Not applicable.

^a Not assessed in the disease group; variable was only collected in the control group.

^b Statistical comparison was not performed due to missing data.

^c p‐values were calculated using Student's t‐test or Welch's t‐test for continuous variables, depending on the equality of variances, and χ ² test (or Fisher's exact test when appropriate) for categorical variables.

3.2. Evaluation of Fluorescence Signal Stability in DBS‐Based SNP Genotyping Under Various Storage Conditions

To evaluate the feasibility of SNP genotyping from DBS under various storage conditions, fluorescence signal derived from the TaqMan Genotyping Assay was measured using real‐time PCR at different time points (0, 24, 48, and 72 h) and temperatures (30°C, 4°C, room temperature, and 70°C). The results, as illustrated in Figure 1, indicate that the fluorescence signal remained relatively stable across storage conditions, with minor variations over time. At 30°C, the signal intensity fluctuated within a narrow range, suggesting minimal degradation under low‐temperature conditions. Similarly, at 4°C and room temperature, fluorescence signal values remained stable, with only minor fluctuations within an acceptable range. However, at 70°C, a gradual decline in fluorescence signal intensity was observed over time, suggesting potential degradation of the target DNA at elevated temperatures. The fluorescence signal was highest in fresh whole blood samples (measured immediately after collection), which served as a reference for comparison with DBS samples stored under different conditions. Notably, SNP genotyping for rs12700667 was successful in all samples, with a 100% detection rate and complete concordance between DBS‐derived and whole blood‐derived genotypes across all storage conditions and time points. Although this validation was limited to a single SNP, the results provide strong support for the robustness and accuracy of DBS‐based SNP genotyping under various storage conditions. These findings imply that, while DBS samples can maintain DNA integrity for SNP genotyping under standard storage conditions, exposure to extreme heat may impact measurement reliability.

FIGURE 1.

Stability of fluorescence signal in SNP genotyping from DBS samples under various storage conditions. This figure illustrates the stability of fluorescence signals obtained from SNP genotyping using dried blood spot (DBS) samples stored under different conditions. The fluorescence signal was measured via the TaqMan Genotyping Assay for rs12700667, with fresh whole blood serving as a reference. DBS samples were stored at four different temperatures (−30°C, 4°C, room temperature, and 70°C) for 24, 48, and 72 h before genotyping analysis. The fluorescence intensity remained stable across most storage conditions, particularly at −30°C, 4°C, and room temperature, indicating minimal DNA degradation. However, a gradual decline in fluorescence signal was observed at 70°C, suggesting potential degradation at elevated temperatures. These findings demonstrate the feasibility of SNP genotyping from DBS samples when stored under appropriate conditions, reinforcing its applicability in large‐scale epidemiological studies.

Overall, the stability of the TaqMan assay‐derived fluorescence signal in DBS samples highlights the feasibility of this method for epidemiological studies, provided appropriate storage conditions are maintained.

3.3. Analysis of Hardy–Weinberg Equilibrium in Control Group

To ensure that the control group was representative of the general population and that the genetic variants under study followed expected Mendelian inheritance patterns, we conducted an HWE test using Pearson's χ ² test with a significance level of 0.05. For rs10965235, where the AA genotype was not observed, Fisher's exact test was additionally performed due to the low expected count.

Table 2 summarizes the HWE test results for the five SNPs analyzed in the control group (n = 144). Among these, four conformed to HWE (p > 0.05), but rs10965235 showed a significant deviation from it (p = 0.0225), suggesting a potential bias due to sample size, genotyping error, or population substructure. Certain genotypes—such as AA in rs10965235, GG in rs12024204, AA in rs12700667, and CC in rs801112—were observed at low frequencies, which may limit the sensitivity of HWE tests. These rare occurrences should be interpreted with caution in the context of sample size limitations.

TABLE 2.

Hardy–Weinberg equilibrium test for control group.

SNP ID	Chromosomal location	Gene	Alleles	Genotype	Frequency (n, %)	p a
rs10965235	9p21.3	CDKN2B‐AS1	C/A	AA	0 (0.0%)	0.023
				CA	46 (31.9%)
				CC	98 (68.1%)
rs12024204	1p22	Intergenic	G/A	AA	96 (66.7%)	0.608
				GA	42 (29.2%)
				GG	6 (4.2%)
rs12700667	7p15.2	Intergenic	A/G	GG	95 (66.0%)	0.216
				AG	41 (28.5%)
				AA	8 (5.6%)
rs16826658	1p36.23–p35.1	Intergenic	G/T	TT	37 (25.7%)	0.491
				GT	76 (52.8%)
				GG	31 (25.6%)
rs801112	1q42.11–q42.3	Intergenic	C/T	TT	102 (70.8%)	0.381
				CT	40 (27.8%)
				CC	2 (1.4%)

Abbreviations: HWE, Hardy–Weinberg equilibrium; SNP, single nucleotide polymorphism.

^a Hardy–Weinberg equilibrium was assessed using Pearson's χ ² test or Fisher's exact test as appropriate, with a significance threshold of p < 0.05.

3.4. Association Between SNPs and Endometriosis Risk

To investigate the association between specific SNPs and susceptibility to endometriosis, we calculated ORs and 95% CIs using contingency table analysis with the Haldane–Anscombe correction. The results are summarized in Table 3.

TABLE 3.

Association between SNPs and endometriosis risk.

SNP ID	Genotype	Control group (n = 144) (n, %)	Disease group (n = 37) (n, %)	Odds ratio a , b	95% CI	p c
rs10965235	A	0 (0.0%)	1 (2.7%)	Ref.
	CA	46 (31.9%)	13 (35.1%)	0.24	0.049–1.213	0.085
	CC	98 (68.1%)	23 (62.2%)	0.19	0.041–0.917	0.039
rs12024204	AA	96 (66.7%)	26 (70.3%)	Ref.
	GA	42 (29.2%)	10 (27.0%)	0.92	0.418–2.020	0.834
	GG	6 (4.2%)	1 (2.7%)	1.03	0.201–5.230	0.975
rs12700667	GG	95 (66.0%)	26 (70.3%)	Ref.
	AG	41 (28.5%)	11 (29.7%)	1.02	0.470–2.195	0.968
	AA	8 (5.6%)	0 (0.0%)	0.40	0.048–3.258	0.388
rs16826658	TT	37 (25.7%)	9 (24.3%)	Ref.
	GT	76 (52.8%)	20 (54.1%)	1.04	0.444–2.418	0.934
	GG	31 (21.5%)	8 (21.6%)	1.07	0.387–2.952	0.898
rs801112	TT	102 (70.8%)	27 (73.0%)	Ref.
	CT	40 (27.8%)	10 (27.0%)	0.99	0.450–2.165	0.974
	CC	2 (1.4%)	0 (0.0%)	1.23	0.123–12.248	0.862

Abbreviation: CI, confidence interval.

^a Odds ratios and 95% confidence intervals were calculated using the Haldane–Anscombe correction.

^b Reference genotype used to calculate odds ratios.

^c p‐values were Bonferroni‐corrected for five independent SNP comparisons. A corrected significance threshold of p < 0.01 was applied.

Although rs10965235 showed a statistically significant association before correction (OR = 0.19, 95% CI: 0.041–0.917, p = 0.039), this association did not remain significant after Bonferroni correction (corrected threshold p < 0.01). The CA genotype also showed a tendency toward reduced risk (OR = 0.24, 95% CI: 0.049–1.213, p = 0.085), but this was also not statistically significant after correction. Given the limited number of individuals carrying the AA genotype (observed in only one case and absent in the control group), the association for this genotype could not be reliably assessed. These results should be interpreted with caution and warrant validation in larger, independent cohorts. For rs12024204, the GA genotype was associated with an OR of 0.92 (95% CI: 0.418–2.020, p = 0.834), and the GG genotype had an OR of 1.03 (95% CI: 0.201–5.230, p = 0.975), indicating no significant association with endometriosis. Regarding rs12700667, the AG genotype had an OR of 1.02 (95% CI: 0.470–2.195, p = 0.968), and the AA genotype showed an OR of 0.40 (95% CI: 0.048–3.258, p = 0.388). These results do not support a significant relationship of these SNPs with disease risk. For rs16826658, the GT genotype had an OR of 1.04 (95% CI: 0.444–2.418, p = 0.934), and the GG genotype showed an OR of 1.07 (95% CI: 0.387–2.952, p = 0.898), again indicating no statistically significant difference. Lastly, for rs801112, the CT genotype had an OR of 0.99 (95% CI: 0.450–2.165, p = 0.974), and the CC genotype had an OR of 1.23 (95% CI: 0.123–12.248, p = 0.862), suggesting no meaningful association with endometriosis. Taking these findings together, none of the five SNPs analyzed in this study demonstrated a statistically significant association with endometriosis risk after correction for multiple comparisons. These findings highlight the importance of cautious interpretation in small‐sample genetic studies and the need for replication in larger, well‐powered cohorts.

3.5. Comparison of Risk Allele Frequencies Between GWAS and the Current Study

To assess the consistency of SNP associations with endometriosis across populations, we compared the risk allele frequencies observed in our study with those reported in previous GWAS [31, 32, 33]. These reference GWAS cohorts varied in ancestry and sample size: rs10965235 and rs16826658 were identified in Japanese studies involving approximately 2400 participants [31], while rs12024204 and rs12700667 were derived from European cohorts with over 10 000 individuals [32]. rs801112 was also reported in a Japanese GWAS comprising more than 7000 participants [33]. Table 4 presents the frequency of risk alleles in both control and disease groups from our study alongside GWAS findings.

TABLE 4.

Comparison of risk allele frequencies between GWAS and our study.

SNP ID	Risk allele	Control group			Disease group
		GWAS a	Our study	p b , c	GWAS	Our study	p c
rs10965235	C	0.802	0.840	0.0309	0.854	0.797	0.0010
rs12024204	G	—	0.188	NA	0.540	0.162	< 0.0001
rs12700667	A	—	0.198	NA	0.740	0.149	< 0.0001
rs16826658	G	0.517	0.479	0.0980	0.573	0.486	0.0001
rs801112	C	0.099	0.153	0.0004	0.167	0.135	0.0529

Abbreviation: GWAS, genome‐wide association study; NA, not available.

^a Data not available in the reference GWAS dataset.

^b GWAS data for this SNP were not reported in the reference datasets.

^c p‐values were Bonferroni‐corrected for five independent SNP comparisons. A corrected significance threshold of p < 0.001 was applied.

Among the five SNPs analyzed, rs10965235 (C allele) exhibited a similar frequency pattern between GWAS and our study. In the control group, the C‐allele frequency was 0.802 in GWAS [31], while our study reported a slightly higher frequency of 0.840. In the disease group, GWAS reported a frequency of 0.854 [31], while our study found a slightly lower frequency of 0.797. The p‐value for the disease group comparison was 0.0010, which remained statistically significant after Bonferroni correction (p < 0.005). In contrast, the control group comparison (p = 0.0309) was not significant after correction. For rs12024204 (G allele), there was a notable discrepancy between GWAS and our study. For the disease group in GWAS, a frequency of 0.540 was reported [32], whereas our study found a significantly lower frequency of 0.162 (p < 0.0001). This difference was statistically significant and remained so after Bonferroni correction. In the control group, our study detected a G‐allele frequency of 0.188, although no corresponding GWAS control data were available for direct comparison. A similar trend was observed for rs12700667 (A allele), where the GWAS disease group exhibited a frequency of 0.740 [32], whereas our study reported a substantially lower frequency of 0.149 (p < 0.0001). This difference remained significant after Bonferroni correction. Meanwhile, in the control group, our study found a frequency of 0.198, although GWAS data for controls were not available. For rs16826658 (G allele), GWAS reported a control frequency of 0.517 and a disease frequency of 0.573 [31]. Our study yielded similar values, with a control group frequency of 0.479 and a disease group frequency of 0.486. The p‐value for the disease group comparison was 0.0001, which remained statistically significant after Bonferroni correction. However, the control group comparison (p = 0.0980) was not statistically significant after correction. Finally, for rs801112 (C allele), GWAS reported a control group frequency of 0.099 and a disease group frequency of 0.167 [33]. In our study, we observed slightly higher frequencies, with 0.153 in the control group and 0.135 in the disease group. The control group comparison showed a statistically significant difference (p = 0.0004), which remained significant after Bonferroni correction, while the disease group comparison (p = 0.0529) did not reach significance.

Overall, while some SNPs showed risk allele frequencies similar to those reported in previous GWAS, others—particularly rs12024204 and rs12700667—exhibited substantial differences. After Bonferroni correction adjusted the significance threshold to p < 0.005, significant differences in risk allele frequencies were observed for rs10965235 (disease group only), rs12024204 (disease group), rs12700667 (disease group), rs16826658 (disease group), and rs801112 (control group). These discrepancies may be attributable to population‐specific genetic variations or methodological differences in genotyping and sample selection. Further studies with larger sample sizes and diverse populations are necessary to validate these findings and clarify the genetic architecture of endometriosis.

4. Discussion

This study was primarily intended to evaluate the feasibility and robustness of SNP genotyping using DBS samples in the context of large‐scale epidemiological research. In practical scenarios such as cohort studies, DBS samples are often collected at home by the participants themselves and submitted by mail. This introduces variability in storage duration and temperature, making it essential to validate that SNP genotyping remains reliable under such conditions. Our findings clearly demonstrate that DBS is a practical and reliable DNA source for SNP genotyping. DBS samples, which can be collected in a minimally invasive manner and stored long‐term, offer practical advantages for large‐scale epidemiological research where conventional blood collection may pose logistical challenges [25, 26]. Our analysis showed that SNPs could be reliably detected from DBS across various storage durations and temperatures. Fluorescence signal evaluation indicated that DNA integrity was well preserved under standard conditions, although signal degradation was observed following prolonged exposure to high temperatures (70°C). Furthermore, genotyping of rs12700667—the SNP selected to assess technical feasibility—was successful in 100% of DBS samples across all storage conditions, demonstrating the robustness and reliability of DBS‐based SNP detection. These findings support the utility of DBS as a reliable DNA source, aligning with its established role in newborn screening and large cohort studies. The simplicity, stability, and scalability of DBS sampling make it a valuable tool for future genetic epidemiology.

To ensure the reliability of genotype frequency distributions, HWE analysis was conducted. Following the recommendation that exact tests are preferable for small samples and rare alleles [35, 36], we assessed equilibrium with Pearson's χ ² test and, where any expected cell count was < 5, confirmed the result with Fisher's exact test. Four of the five SNPs conformed to Hardy–Weinberg expectations; however, rs10965235 showed a nominal deviation (p = 0.0225), attributable to the absence of the AA genotype in our modest control sample. In the original Japanese GWAS that first reported this variant, rs10965235 passed HWE quality control in a much larger cohort of 5292 controls, indicating that it is unlikely to actually deviate from HWE [31]. Moreover, subsequent meta‐analyses have shown that this SNP is monomorphic (all individuals are CC) in populations of European ancestry [12], reflecting marked inter‐ethnic differences in allele frequency rather than genotyping error. Taking these findings together, our identified deviation from HWE probably reflects stochastic variation caused by the very low frequency of the A allele in this small Japanese control set. The finding that the other four SNPs did not depart from HWE supports the overall representativeness of the control group and the reliability of genotyping. Nevertheless, rare genotypes (e.g., rs12024204 GG, rs12700667 AA, rs801112 CC) were by their very nature infrequently found, limiting the sensitivity of HWE tests; deviations from HWE—or lack thereof—should therefore be interpreted with caution in small samples [36].

Endometriosis is a complex disease with a strong genetic component, as demonstrated by multiple GWAS [7, 32, 37]. Several SNPs have been consistently associated with endometriosis risk. Among them, rs12700667 is located upstream of HOXA10 (homeobox A10), a gene involved in endometrial receptivity, and NFE2L3 (nuclear factor, erythroid 2 like 3), which plays a role in placental development. Previous studies have suggested that this SNP is significantly associated with endometriosis risk across multiple populations [32, 38]. rs10965235 is located within CDKN2B‐AS1 (cyclin‐dependent kinase inhibitor 2B antisense RNA 1), a long non‐coding RNA situated at the 9p21.3 endometriosis risk locus [15]. Subsequent studies have shown that CDKN2B‐AS1 may regulate CDKN2A/B (cyclin‐dependent kinase inhibitor 2A/2B) gene expression through chromatin interactions, potentially affecting cell cycle progression and endometrial proliferation [39]. Other SNPs examined in this study also play potential roles in susceptibility to endometriosis. rs16826658 is positioned within the linkage disequilibrium block containing the WNT4 gene, a key regulator of reproductive system development. WNT4 is critical for ovarian function and steroidogenesis, and its dysregulation has been linked to endometriosis [12]. rs801112 is located downstream of the G‐protein‐related gene RHOU (Ras Homolog Family Member U), which is involved in cellular signaling and cytoskeletal dynamics, processes that may contribute to ectopic endometrial cell survival and migration. rs12024204 has been reported to be associated with the onset of endometriosis, although its functional role remains unclear [38]. Given these genetic implications, this study aimed to evaluate the feasibility of SNP genotyping from DBS samples and to investigate the association between selected SNPs and the risk of endometriosis. By integrating genetic analysis with epidemiological data, we sought to demonstrate the utility of DBS in large‐scale population‐based studies and to clarify the potential role of specific SNPs in the pathogenesis of endometriosis.

In contrast to previous GWAS, our analysis identified rs10965235 as a potential protective variant against endometriosis, with the CC genotype being present in affected individuals at a significantly lower frequency (OR = 0.19, p = 0.039). While earlier reports associated the C allele with increased risk, our findings suggest that this discrepancy may reflect differences in sample composition, case definition, or regional genetic structure rather than ethnic background alone. rs10965235 was originally identified in a GWAS of Japanese women [31], whereas our sample was also exclusively Japanese but drawn from a smaller, hospital‐based cohort. Factors such as sample size, regional variation within Japan, and case‐selection criteria could therefore underlie the observed difference. Crucially, rs10965235 deviated from HWE in controls (p = 0.0225), most likely because the AA genotype was entirely absent and the cohort was small. Deviations of this nature are well documented in simulations and empirical studies of limited sample sets [35] and may inflate apparent effect sizes or generate false positives [36]. Hence, the apparent protective effect of the CC genotype should be regarded as preliminary. The AA genotype was observed in only one case and was absent in the control group, further limiting interpretability. Low‐frequency genotypes can distort association statistics, emphasizing the need for replication in larger, independent cohorts before firm conclusions are drawn. Moreover, rs12700667, an SNP previously reported to be associated with endometriosis, did not demonstrate a significant association in our cohort. Similarly, rs12024204, rs16826658, and rs80111 showed no significant relationship with disease risk. These discrepancies may be attributable to genetic heterogeneity across populations, limited statistical power, or unaccounted gene–environment interactions. Given the small sample size, our findings should be regarded as exploratory. Unlike large‐scale GWAS, this study does not provide confirmatory evidence of genetic susceptibility but serves as a population‐specific reference for hypothesis generation. Taken together, our results emphasize the need for population‐specific validation and underscore the importance of conducting large‐scale studies involving individuals from diverse ethnic backgrounds to elucidate the genetic architecture of endometriosis more comprehensively.

Our analysis also revealed significant differences in allele frequencies compared with the findings in previous GWAS datasets [31, 32, 33]. In particular, for rs12024204 and rs12700667, both identified in European cohorts [32], the risk allele frequencies in our disease group were markedly lower than those reported in the original. These findings suggest notable population‐specific differences in genetic architecture and remained statistically significant even after Bonferroni correction. Notably, rs10965235, rs16826658, and rs801112 were originally identified in Japanese GWAS involving 2400 to over 7000 participants [31, 33], while rs12024204 and rs12700667 were reported in large European cohorts including over 10 000 individuals [32]. This distinction suggests that differences in allele frequencies cannot be attributed solely to ethnic background. Despite all samples being from a single country (Japan), regional genetic substructure or differences in study design—such as recruitment strategy or case definition—may have contributed to these discrepancies. In particular, rs801112, although identified in a Japanese GWAS, exhibited a frequency difference in the control group that remained statistically significant after correction, but not in the disease group. This pattern may reflect stochastic variation or selection‐related biases that disproportionately affect control populations. Together, these findings underscore the importance of using fine‐scale, population‐specific reference data when interpreting genetic association results, even within seemingly homogeneous ethnic groups.

This study has several limitations that warrant consideration. First, the relatively small sample size, particularly in the disease group, may have reduced the statistical power to detect modest genetic associations, especially for genotypes with low frequencies. We therefore calculated the statistical power for each of the five SNPs based on the actual observed allele frequencies and sample sizes. The estimated power was consistently low, ranging from approximately 5%–11%, indicating that our study was underpowered to detect small genetic effects. This underscores the limitations of our study in detecting modest associations and reinforces the importance of sample size considerations in future genetic epidemiology research. Although statistical corrections such as the Haldane–Anscombe method and Fisher's exact test were applied to improve estimate stability, these approaches cannot fully address the uncertainty inherent in small samples. Furthermore, the small number of cases and low genotype counts precluded the use of multivariable logistic regression analysis to adjust for potential confounders such as age or BMI. As a result, we relied on univariate analyses using Fisher's exact test, which do not allow covariate adjustment but are appropriate for small and sparse datasets. The observed age difference between the control and disease groups was largely due to the inclusion of older participants from the population‐based GNHS cohort. However, the age distribution among hospital‐based participants appeared more similar between groups. While formal covariate adjustment was not feasible, these observations suggest that the risk of confounding due to age or other baseline characteristics is likely to be limited. Second, while DBS samples offer a stable and minimally invasive DNA source, potential degradation under extreme storage conditions—such as prolonged exposure to high temperatures—remains a concern [30]. Lastly, our analysis focused solely on SNP associations and did not incorporate potential epigenetic modifications or gene–environment interactions, which are increasingly recognized as important contributors to the pathogenesis of endometriosis [8, 40].

In conclusion, this study demonstrates the feasibility and robustness of DBS‐based SNP genotyping and its applicability to large‐scale epidemiological research. The inclusion of both hospital‐based and population‐based participants contributed to sample diversity and helped mitigate selection bias. Although our limited sample size precluded confirmation of previously reported SNP–disease associations, we observed notable differences in allele frequencies compared with prior GWAS findings. In particular, rs10965235 showed a potentially inverse association with endometriosis risk, although this finding should be interpreted with caution due to deviation from HWE and limited sample size. These results highlight the importance of population‐specific validation when applying GWAS findings, even within ethnically homogeneous populations. Based on the demonstrated feasibility of DBS genotyping, we plan to utilize our existing large‐scale cohort of Japanese women to further investigate the association between candidate SNPs and the development of endometriosis.

Consent

All authors reviewed and approved the final version of the manuscript and consent to its publication.

Conflicts of Interest

The authors declare no conflicts of interest. This study was approved by the institutional ethics committee, and informed consent was obtained from all participants. Dr. Akira Iwase is the Editor‐in‐Chief of Reproductive Medicine and Biology and a co‐author of this article. To minimize bias, he was excluded from all editorial decision‐making related to the acceptance of this article for publication.

Kitahara Y., Ideno Y., Tomiyoshi K., et al., “Feasibility of SNP Genotyping Using Dried Blood Spot Samples Collected in an Epidemiological Study and Its Integration With Genetic Risk Analysis for Endometriosis,” Reproductive Medicine and Biology 24, no. 1 (2025): e12675, 10.1002/rmb2.12675.

Data Availability Statement

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.