A literature search was conducted in all available public and scientific databases, including Embase, NCBI, Google Scholar, Medline, and Science Direct, from inception to April 2023 using the following keywords to identify all articles on the association of 238G/A, 308G/A, and 1031 T/C with PCOS: TNF-alpha, TNF-alpha with polycystic ovarian syndrome, PCOS-TNF-alpha, and PCOS. Only English-language articles were considered. In order to avoid duplicated studies, authors’ names were searched and screened in all the databases to identify appropriate studies, titles, abstracts, and full texts. In addition, the reference lists of the identified articles were also screened.

The studies had to meet all of the following inclusion criteria: (i) evaluate the relationship between the TNF-α 238G/A, 308G/A, and 1031T/C polymorphisms and PCOS; (ii) include patients with PCOS and controls with eligible genotypic and phenotypic distributions of TNF-α 238G/A, 308G/A, and 1031T/C polymorphisms; (iii) both cases and controls are of the same ethnicity; and (iv) the full text was available in English. The exclusion criteria were as follows: (i) no control group; (ii) low 95% confidence interval or odds ratio (OR); (iii) studies with overlapping data; and (iv) animal studies. Most studies on Caucasians have not examined the association between PCOS and TNF-α 238G/A, 308G/A, and 1031T/C polymorphisms. We identified very few studies on Caucasians, and many were excluded due to the lack of proper inclusion criteria.

The following data were extracted from the articles selected for inclusion: publication year, first author, origin, ethnicity, number of cases, genotyping methods, and controls registered. The conflict and disagreement from the selected articles were removed from the study.

The data were analyzed using the Review Manager 5.4 and MetaGenyo software. In order to determine whether the study was significant, we determined whether the p-value was significant at p < 0.005 using genetic variations such as allele comparison, dominant regression, over-dominant, and recessive. The consistency of the findings across all studies was evaluated using the inconsistency index (I²), which ranges from 0 to 100. The inconsistency index is crucial in determining the homogeneity (0% significance) and heterogeneity indications, which are responsible for most variations30. The degree of heterogeneity among the studies was assessed using Q-statistics and the chi-square test. The z-test was used to calculate ORs, and shared results among studies were considered statistically significant at p < 0.05. A sensitivity analysis was performed to assess the relative contributions of the included studies to the total estimates, removing one study at a time. Funnel plots and Egger’s linear regression test were used to detect publication bias. The log standard error was plotted against the odds ratios for each study. The heterogeneity between the eligible analyses performed using Egger’s test, Q-test, and inconsistency index statistics was considered statistically significant if p < 0.005.

Table 2

Contents	Study	Ethnicity	AA_Cases/Controls	GA_Cases/Controls	GG_Cases/Controls	Total Cases/Controls	Hardy-Weinberg equilibrium p-value
TNF-Alpha 238G/A	Sampurna et al. 202131	Asian	32/30	44/45	24/25	100/100	0.3283
	Bhatnagar et al. 201932	Asian	2/0	50/83	148/117	200/200	0.0002
	Kordestani et al. 201833	Asian	1/3	6/6	104/96	111/105	0
	Wen et al. 201334	Asian	0/0	7/4	137/68	144/72	0.8084
	Xie et al. 201635	Asian	0/0	3/3	99/93	102/96	0.8764
	Kordestani et al. 201833	Asian	1/3	6/6	104/96	111/105	0
TNF-Alpha 308G/A	Sampurna et al. 202131	Asian	32/30	44/45	24/25	100/100	0.3283
	Bhatnagar et al. 201932	Asian	2/0	50/83	148/117	200/200	0.0002
	Sampurna et al. 202131	Asian	25/30	50/47	25/23	100/100	0.5798
	Azeez et al. 202136	Asian	1/1	25/15	32/14	58/30	0.2054
	Alwan et al. 202137	Asian	9/7	15/13	56/50	80/70	0.0007
	Li et al. 201738	Asian	3/1	33/24	357/356	393/381	0.387
	Bhatnagar et al. 201932	Asian	0/2	79/63	121/135	200/200	0.0671
	Deepika et al. 201339	Asian	3/3	10/10	270/293	283/306	0
	Mao et al. 200040	Asian	1/4	29/13	88/37	118/54	0.0889
	Milner et al. 199941	Caucasian	2/3	23/42	59/63	84/108	0.1939
	Peng et al. 201042	Asian	1/2	11/27	118/146	130/175	0.557
	Vural et al. 201043	Asian	3/3	16/15	78/77	97/95	0.0549
	Wen et al. 201334	Asian	0/0	14/7	89/52	103/59	0.6281
TNF-Alpha 1031T/C	Sampurna et al. 202131	Asian	38/35	38/40	24/25	100/100	0.055
	Bhatnager et al. 201932	Asian	0/2	48/69	152/129	200/200	0.0272
	Deepika et al. 201339	Asian	6/5	170/139	107/162	283/306	0
	Yun et al. 201144	Asian	2/0	71/22	144/122	217/144	0.321
	Hazwanie et al. 201545	Asian	3/91	8/45	1/9	12/145	0.2922
	Alkhuriji et al. 202046	Asian	1/9	17/46	50/82	68/137	0.4666

This study aimed to identify associations between TNF-α gene polymorphisms and PCOS. The searches of the various databases, including Google Scholar, NCBI, and Science Direct, identified six studies with 768 cases and 678 controls for the TNF-α 238G/A polymorphism47, 32, 33, 31, 34, 35, 11 studies with 1646 cases and 1578 controls for the TNF-α 308G/A polymorphism32, 31, 34, 39, 48, 41, 36, 38, 43, 37, 40, and six studies with 880 cases and 1032 controls for the TNF-α 1031T/C polymorphism32, 31, 39, 46, 45, 44. Based on the data collected from these 23 studies, PCOS was associated with the TNF-α 238G/A, 308G/A, and 1031T/C polymorphism (Figure 1). The data selected for the analysis is shown in Table 2. Among the selected studies, 22 were conducted in Asians and one in Caucasians. Begg’s funnel plot, funnel plot, and Egger’s test were performed for statistical evidence, and heterogeneity was observed from all selected articles.

Figure 1

Table 3

Model	Ethnicity	Number of studies		Test of association			Test of heterogeneity
		-	Odd ratio	95% confidence interval	p-value	Model	p-value	I2	p-value (Egger's test)
Allele contrast (A vs. a)	Overall	6	1.2860	[1.0001; 1.6537]	0.049953	Fixed	0.3215	0.1457	0.957
Recessive model (AA vs. Aa+aa)	Overall	6	1.5287	[1.1134; 2.0989]	0.008691	Fixed	0.3803	0.0463	0.2936
Dominant model (AA+Aa vs. aa)	Overall	3	0.9344	[0.5287; 1.6516]	0.815448	Fixed	0.3419	0.0683	0.8213
Overdominant (Aa vs. AA + aa)	Overall	5	0.6539	[0.4800; 0.8908]	0.007073	Fixed	0.2978	0.1835	0.3097

Table 4

Model	Ethnicity	Number of studies		Test of association			Test of heterogeneity		Publication bias
	-	-	Odd ratio	95% confidence interval	p-value	Model	p-value	I2	p-value (Egger's test)
Allele contrast (A vs. a)	Overall	11	1.0494	[0.8912; 1.2356]	0.562921	Fixed	0.2733	0.1789	0.3883
Recessive model (AA vs. Aa+aa)	Overall	11	1.0159	[0.8364; 1.2338]	0.873805	Fixed	0.3078	0.143	0.1283
Dominant model (AA+Aa vs. aa)	Overall	10	1.2340	[0.8015; 1.9000]	0.339546	Fixed	0.7397	0	0.4684
Overdominant (Aa vs. AA + aa)	Overall	11	1.0380	[0.8515; 1.2654]	0.711939	Fixed	0.3892	0.057	0.1445

Table 5

Model	Ethnicity	Number of studies		Test of association			Test of heterogeneity		Publication bias
	-	-	Odd ratio	95% confidence interval	p-value	Model	p-value	I2	p-value (Egger's test)
Allele contrast (A vs. a)	Overall	6	1.0641	[0.6570; 1.7235]	0.800633	Random	0	0.8621	0.3096
Recessive model (AA vs. Aa+aa)	Overall	6	0.912	[0.4971; 1.6733]	0.766151	Random	0	0.8524	0.5966
Dominant model (AA+Aa vs. aa)	Overall	6	1.5185	[0.6634; 3.4755]	0.322802	Random	0.097	0.4635	0.3609
Overdominant (Aa vs. AA + aa)	Overall	6	1.3060	[0.7446; 2.2907]	0.35179	Random	0	0.8456	0.9502

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

The meta-analysis included 23 total studies to assess the association of PCOS with TNF-α 238G/A, 308G/A, and 1031T/C polymorphisms. We combined all the collected data with 3288 controls and 3294 cases. The TNF-α 238G/A polymorphism showed no significant associations in the allelic, recessive, dominant, and over-dominant models (p > 0.05; Table 3). In addition, subgroup analyses with allelic, recessive, dominant, and over-dominant models were nonsignificant (p > 0.05). Similarly, the TNF-α 308G/A polymorphism showed no significant associations in the allelic, dominant, overdominant, and recessive models (p > 0.05; Table 4). Again, subgroup analyses with the allelic, dominant, recessive, and over-dominant models were nonsignificant (p > 0.05). However, the TNF-α 1031T/C polymorphism showed significant associations with the allelic, recessive, and over-dominant models (p < 0.05; Table 5).

Each variable was checked for apparent publication bias due to sample size constraints and reporting bias. A forest plot represented the heterogenicity, and a sensitivity analysis was conducted to identify studies contributing to the total estimates. Statistically constant results were obtained from the collected data (Figure 2, Figure 3, Figure 4, Figure 5, Figure 6, Figure 7), and significant results (Figure 8, Figure 9, Figure 10) were obtained when the funnel plot was statistically analyzed (Figure 11, Figure 12, Figure 13).

PCOS is the most commonly occurring endocrine disorder among adult women. In PCOS, ovarian dysfunction causes other metabolic disorders, which may result in chronic inflammation. TNF-α, an inflammatory cytokine found in the human ovaries, ovarian follicular fluid, and oocytes, causes many inflammatory disorders and impacts follicular atresia, physiological ovulation dysfunction, ovarian apoptosis, anovulation, and steroid secretion. The pathophysiological and etiological mechanisms of PCOS are highly complex. Inflammation and immune-regulatory genes are responsible for PCOS development and progression. However, many studies have not succeeded in identifying a significant relationship between inflammatory pathways and PCOS development.

Our study comprised 2764 cases with PCOS and 3218 controls obtained from published studies exploring associations between PCOS and TNF-α 238G/A, 308G/A, and 1031T/C polymorphisms. In our study, the TNF-α 238G/A polymorphism showed no association with PCOS in all genetic models (allele, recessive, over-dominant, and dominant), suggesting that TNF-α 238G/A may not contribute to PCOS development. There is no conclusive proof associating the TNF-α 308G/A polymorphism with PCOS. However, the TNF-α 1031T/C polymorphism was significantly associated with PCOS in the allelic, recessive, and over-dominant models but not in the dominant model. Therefore, these results suggest that the TNF-α 1031T/C polymorphism may not be associated with PCOS.

Our meta-analysis has generated some important findings, but it is not without limitations. Since we only considered research studies published in English that were abstracted and indexed in online databases, we might have missed relevant studies published in other languages. We also did not investigate the possible influence of gene-environment interactions, and a subgroup analysis was not performed because there was a dearth of appropriate studies and data. Nonetheless, our meta-analysis does have some strengths. Firstly, our qualitative and quantitative analysis, such as funnel plot and Egger’s linear regression, did not demonstrate clear evidence of publication bias. Therefore, we can conclude that our findings have strong statistical support. Secondly, a strict data extraction and analysis procedure was performed to draw satisfactory and reliable conclusions from the study.

In conclusion, our meta-analysis assessed possible associations of TNF-α 238G/A, 308G/A, and 1031T/C polymorphisms with PCOS using valuable statistical data from significant and nonsignificant studies. Overall, our meta-analysis showed that the TNF-α 238G/A and 308G/A polymorphisms may not be associated with PCOS, whereas the TNF-α 1031T/C polymorphism may be associated with PCOS. Future studies should explore possible associations and interactions between TNF-α polymorphisms and PCOS using large datasets comprising gene and environment data to confirm our findings.

PCOS: Polycystic ovarian syndrome, TNF-α: Tumor necrosis factor-alpha, SNPs: single nucleotide polymorphisms, TACE: TNF-α converting enzyme

Gowtham, Harini, Thirumalai, and Pragya acknowledge CARE for fellowship.

All authors significantly contributed to this work, read and approved the final manuscript.

None.

Data and materials used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Not applicable.

Not applicable.

The authors declare that they have no competing interests.