Risk factors for androgenetic alopecia: a systematic review and meta-analysis

Abstract

Background

Androgenetic alopecia (AGA), the most common type of hair loss, is influenced by multiple factors. However, the associations between various risk factors and AGA remain controversial, with no systematic review or meta-analysis comprehensively analyzing these risk factors to date.

Methods

PubMed, Embase, and Cochrane databases were searched for observational studies reporting on AGA risk factors from inception through January 5, 2024.

Results

31 studies involving 11,224 AGA cases and 36,825 controls were included. Family history (presence: odds ratio [OR] 2.72, 95% confidence interval [CI] 1.85–3.99; progression: OR 4.24, 95% CI 2.77–6.49) and smoking (presence: OR 1.46, 95% CI 1.06–2.01; progression: OR 1.60, 95% CI 1.29–1.99) were significantly associated with both presence and progression of AGA. A paternal family history of AGA (OR 2.22, 95% CI 1.59–3.10), insulin resistance (standardized mean difference [SMD] 0.40, 95% CI 0.27–0.53), and high fasting insulin levels (SMD 0.48, 95% CI 0.18–0.78) were also identified as factors that increased the risk of the presence of AGA. Alcohol consumption (OR 1.72, 95% CI 1.28–2.32), insufficient/poor quality sleep (OR 1.36, 95% CI 1.18–1.56), and overweight/obesity (OR 2.31, 95% CI 1.01–5.29) were related to the progression of AGA. Hypertension in men showed a significant correlation with AGA (OR 1.60, 95% CI 1.26–2.04).

Conclusion

This meta-analysis identified several risk factors associated with AGA presence and progression, including a family history of AGA, smoking, alcohol consumption, insufficient or poor-quality sleep, overweight/obesity, hypertension in men, insulin resistance, and elevated fasting insulin levels. Future studies should explore the underlying mechanisms linking these factors to AGA.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26258-y.

Introduction

Androgenetic alopecia (AGA), the most prevalent type of hair loss, is characterized by progressive hair follicle miniaturization and disruption of the hair growth cycle, with a shortened anagen phase and an extended telogen phase, eventually leading to baldness [1]. The appearance of hair loss can impose a significant psychological burden on patients, leading to anxiety, low self-esteem, and severe impairment of personal life and social interaction [2].

AGA, a polygenic hereditary condition, is more susceptible to environmental influences than monogenic hereditary diseases, indicating that its presence and progression are not exclusively linked to genetic factors but also to non-genetic factor [3]. Recent researches have demonstrated that poor lifestyle choices, unhealthy dietary habits, and psychological stress may contribute to the presence or worsen the progression of AGA [4–6]. Furthermore, comorbidities associated with AGA are found to exhibit a close and bidirectional relationship with its development and should be regarded as risk factors for AGA [7].However, the impact of each risk factor is controversial, highlighting the need for a meta-analysis.

Although meta-analyses have explored the relationships between AGA and metabolic syndrome [8], cardiovascular and prostate diseases [9–11], quality of life [12], and glucose-lipid and hormone spectrums [13, 14], many important factors remain unaddressed due to the evolution of research advancements. Moreover, previous meta-analyses on smoking have focused exclusively on male AGA patients, underscoring the need for an updated meta-analysis [15].

Thus, in this context, our study aims to conduct a comprehensive exploration of the risk factors associated with the presence and progression of AGA, coupled with a quantitative analysis of certain factors.

Materials and methods

This systematic review was designed and reported in accordance with the PRISMA guidelines. and was preregistered with PROSPERO (CRD42024508848).

Search strategy

A systematic literature search was conducted across the PubMed, Embase, and Cochrane databases from their inception until October 26, 2025. The comprehensive search strategy is detailed in Appendix S1 (see Supporting Information). Zehong Guo and Botian Jiang independently screened the titles and abstracts, whereas a third reviewer Wenzhen Li conducted the full-text assessments. Discrepancies in data extraction were resolved through a comprehensive discussion with the third reviewer.

Study selection

The inclusion criteria for titles and abstracts based on the search results are as follows: (1) The study must be an observational study that reports one or more factors associated with the presence or severity of AGA. (2) Eligible factors include family history, lifestyle or environmental factors and comorbidities. (3) The inclusion of grey literature is permitted. During full-text screening, exclusion criteria encompass: (1) Studies that fail to validate the hair loss type as AGA (2) Studies that do not employ recognized classification metrics such as the Hamilton scale, BASP classification, or Ludwig scale for AGA evaluation. (3) Non-observational study articles. (4) Meta-analyses and systematic reviews (although their references were checked). (5) Articles not written in English or Chinese. (6) The total sample size of the study is less than 100.

Data extraction and quality assessment

Haoyang Li and Jiaxian Zhang extracted data on the authors, publication year, geographical location, study design, sample size, diagnostic criteria for AGA, classification schema for disease severity, participant sex, number of individuals with risk factors in both the AGA cohort and the control group without hair loss (alternatively, in the mild versus moderate-to-severe AGA groups), and information related to the risk factors. For each risk factor, the exposure categories, effect estimates, and 95% confidence intervals (CIs) were ascertained. When possible, only effect estimates adjusted for at least age and family history were extracted (except when family history was considered as a risk factor). The methodological rigor of each study was assessed using the 11-item checklist from the Agency for Healthcare Research and Quality (AHRQ). An AHRQ item scored ‘0’ denoted a negative or indeterminate response, while a ‘1’ indicated a positive response. The articles’ quality was categorized as follows: scores of 0–3 corresponded to low quality, 4–7 to medium quality, and 8–11 to high quality [16].

Statistical analysis

Risk estimates were calculated using risk ratios (RRs) and odds ratios (ORs), which were presented as pooled estimates along with their 95% CIs. We utilized the DerSimonian and Laird random-effects model to pool risk effect estimates from studies involving three or more investigations. Heterogeneity was assessed using the Q test and the I² statistic. Subgroup analyses were performed when substantial heterogeneity was observed for a specific risk factor and more than five studies were included. For subgroup analyses that included fewer than ten studies, where conventional funnel plot tests are underpowered, we assessed publication bias using the Luis Furuya-Kanamori (LFK) index derived from Doi plots. An LFK index within ± 1 was interpreted as no asymmetry, while values beyond ± 2 indicated significant asymmetry. However, due to the limited number of studies comprising each meta-analysis, meta-regression was not conducted. To test for publication bias, funnel plots were examined visually, and Egger’s weighted linear regression was applied to meta-analyses that included five or more studies. The presence of publication bias was evaluated using the p-value from Egger’s test, with a p-value < 0.05 indicating significant funnel plot asymmetry suggestive of potential publication bias.The results are reported as the intercept estimate along with its 95% CI.​

This meta-analysis includes two types of control groups in the studies analyzed, which amis to analyze the presence and progression of AGA separately. The first type compares patients diagnosed with AGA of varying severity to a healthy control group (individuals without AGA). The second type compares individuals with moderate to severe AGA (men classified as Hamilton–Norwood scale IV or above, women as Ludwig scale II or above, and individuals according to the BASP classification of M2 ~ 3, C2 ~ 3, V2 ~ 3, F2 ~ 3, U1 ~ 3) to those with mild AGA (individuals not meeting the aforementioned criteria). Due to the limited number of studies concerning certain factors such as family history, hypertension, and dyslipidemia, research examining both the presence and progression of AGA was combined to assess their overall effect. With respect to insulin resistance and fasting insulin levels, given the limited literature that focuses solely on the incidence of AGA, our study consolidated this research and assessed the correlation between these factors and the presence of AGA. Furthermore, studies by Kim, Salvador, et al., which implemented gender stratification among AGA patients, were considered as two independent studies to comprehensively examine the impact of gender.

Additionally, in selecting specific factors, we aggregated the estimates from the highest study-defined category compared with never-smokers, non-drinkers, and individuals with normal sleep patterns. Although the range for considering increased Body Mass Index (BMI) as a risk factor varied across studies, all definitions aligned with the WHO’s criteria for overweight or obesity (BMI > 24.9 kg/m²), thereby justifying the combination of their respective effect sizes. For the analysis of combined effect sizes, our study independently assessed the correlation between risk factors and both the presence and progression of AGA. All analyses were conducted using the R version 4.3.1.

Results

Study inclusion and characteristics of included studies

During the initial literature search, we identified 14,605 articles for potential inclusion, of which 178 articles, along with 10 additional studies from the reference lists, underwent full-text review. Post full-text review, 36 studies were deemed eligible as they reported on at least one AGA risk factor, of which 31 were ultimately incorporated into the meta-analysis [4, 17–46]. The primary reasons for exclusion at the full-text stage included lack of relevance (n = 87), nonconforming outcomes (n = 20), or unsuitable controls (n = 19) (Fig. 1).

Fig. 1.

Flow diagram of selection procedure of studies assessing the relationship of various demographic characteristics, comorbidities, and lifestyle factors with the risk of AGA. A preferred reporting items for systematic reviews and meta-analyses flow diagram that details the inclusion and exclusion of studies considered for this systematic review

Table 1 presents the characteristics of 31 included studies that explored the risk of presence and progression of AGA with respect to demographic characteristics, comorbidities, or lifestyle factors. A total of 11,224 AGA cases were analyzed, with a comparative control group numbering 36,825. Among the studies selected for this review, 2 was conducted in America, 9 in Europe, 17 in Asia, 1 in Australia, and 2 in Africa. 11 studies were of a case-control design, while 16 were designed as cross-sectional studies.

Table 1.

Characteristics of all studies investigating risk factors for AGA

Author, Year	Study type	Region	Gender	Age Group(case/control)༈Mean ± SD༉	Case (n)	Control (n)	Outcome	Diagnostic criteria for AGA	Demographics	Comorbidities	Lifestyle Factors	Quality Score
Su, 2007 [17]	cross-section study	China	M	65.2 ± 11.2	151	576	Moderate and severe AGA (Hamilton–Norwood classification IV–VII)	Hamilton–Norwood or Ludwig	Family history of AGA	Overweight, Hypertension, Dyslipidemia	Smoking, Alcohol consumption	9
Su, 2013 [18]	cross-section study	China	F	> 30y	3104	22,953	AGA(Ludwig type > I)	Ludwig	Family history of AGA	Overweight, Hypertension, Dyslipidemia	Smoking, Alcohol consumption, UV exposure	9
Cristina, 2017 [19]	cross-section study	Italy	M/F	35.6 ± 14.2	114	237	Moderate and severe AGA(male: Hamilton–Norwood classification IV–VII; female: Ludwig type II and III)	Hamilton–Norwood(male) or Ludwig(female)	Family history of AGA	Overweight, Hypertension, Dyslipidemia	Smoking, Alcohol consumption	7
Ahmed, 2021 [4]	cross-section study	Egypt	M	25.69 ± 2.96/24.95 ± 2.93	500	500	AGA Hamilton–Norwood classification Ⅰ–VII	Hamilton–Norwood	None	None	Smoking	6
Yi, 2020 [20]	cross-section study	China	F	> 18y	944	881	Moderate and severe AGA (BASP classification M2 ~ 3, C2 ~ 3, V2 ~ 3, F2 ~ 3, U1 ~ 3)	BASP	None	None	Smoking, Alcohol consumption, Insufficient, sleeping time with poor quality, Outdoor physical labor	8
Mohammad, 2020 [21]	case-control study	Ireland	M	38.33 ± 8.85/38.40 ± 8.59	256	256	Moderate and severe AGA (Hamilton–Norwood classification IV–VII)	Hamilton–Norwood	None	None	Smoking	8
Kim, 2018 [22]	cross-section study	Korea	M/F	Male : 55.8 ± 12.4/53.9 ± 13.9 Female: 56.7 ± 10.3/55.1 ± 11.8	678	372	AGA(BASP classification M0 ~ 3, C0 ~ 3, V1 ~ 3, F1 ~ 3, U1 ~ 3)	BASP	None	Overweight, Hypertension, Dyslipidemia	Smoking, Alcohol consumption	9
Shi, 2023 [23]	cross-section study	China	M	26.5 ± 6.5/28.7 ± 7.5	592	436	AGA (BASP classification M1 ~ 3, C1 ~ 3, V1 ~ 3, F1 ~ 3, U1 ~ 3)	Hamilton–Norwood	Family history of AGA	Overweight	Smoking, Alcohol consumption, Insufficient sleeping time with poor quality	9
Baik, 2019 [24]	cross-section study	Korea	M	59.7 ± 7.3/55.5 ± 6.9	224	708	AGA (Hamilton–Norwood classification IV–VII)	Hamilton–Norwood	Family history of AGA	None	Smoking, Alcohol consumption	5
Matilainen,2003 [25]	cross-section study	Finland	F	63y	97	221	AGA (Ludwig type II and III)	Ludwig	Family history of AGA	Hypertension, Dyslipidemia	Smoking	6
Hirsso, 2006 [26]	cross-section study	Finland	M	63y	129	92	AGA Hamilton–Norwood classification Ⅲ–VII	Hamilton–Norwood	Family history of AGA	Insulin	Smoking	7
Vora, 2019 [27]	case-control study	Turkey	M	29.3 ± 5.6/31.5 ± 6.4	50	50	AGA Hamilton–Norwood classification Ⅲ–VII	Hamilton–Norwood	None	None	Smoking, Alcohol consumption	7
Su, 2010 [28]	cross-section study	China	M	all the participants: 65.2 ± 11.2	150	520	Moderate and severe AGA (Hamilton–Norwood classification IV–VII)	Hamilton–Norwood	Family history of AGA	None	Smoking	8
Somprasong, 2023 [29]	case-control study	Tailand	M	54.9 ± 13.9/53.6 ± 10.5	223	223	AGA Hamilton–Norwood classification Ⅱ–VII	Hamilton–Norwood	None	Hypertension	Insufficient sleeping time with poor quality	8
Lai, 2013 [30]	cross-section study	China	M	35–65	60	228	Moderate and severe AGA (Hamilton–Norwood classification IV–VII)	Hamilton–Norwood	None	None	Insufficient sleeping time with poor quality	9
Yi, 2020 [31]	cross-section study	China	M	30–40	313	3340	Moderate and severe AGA (Hamilton–Norwood classification IV–VII)	Hamilton–Norwood	None	None	Insufficient sleeping time with poor quality	6
Severi, 2003 [32]	cross-section study	Australia	M	40–69	1040	350	AGA Hamilton–Norwood classification Ⅱ–VII	Hamilton–Norwood	None	Overweight	Smoking, Alcohol consumption	8
Yang, 2014 [33]	cross-section study	China	M	30.2 ± 8.8/ 35.8 ± 11.5	38	104	Moderate and severe AGA (Hamilton–Norwood classification IV–VII)	Hamilton–Norwood	None	Overweight	None	5
Lakhani, 2000 [34]	case-control study	Finland	M	19–50	154	154	AGA Hamilton–Norwood classification >Ⅲ	Hamilton–Norwood or Ludwig	None	Overweight	None	6
Vayá, 2016 [35]	case-control study	Spain	M	43 ± 8/43 ± 10	50	50	AGA	Hamilton–Norwood	None	Overweight, Insulin, Insulin resistance	None	6
Arias, 2010 [36]	case-control study	Spain	M/F	Male:46.3 ± 8.7/43.28 ± 10.3 Female: 48.18 ± 9.7/47.38 ± 6.4	77	77	AGA (male: Ebling scale III-V/Ludwig scale II-III)	Ebling scale or Ludwig	None	Insulin resistance	None	8
Swaroop,2019 [37]	case-control study	India	M	24.18 ± 2.663/25.12 ± 2.344	50	50	AGA Hamilton–Norwood classification Ⅲ–Ⅵ	Hamilton–Norwood	None	Insulin	None	8
González,2009 [38]	case-control study	Mexico	M	25.64 ± 4.16/26.49 ± 3.08	80	80	AGA Hamilton–Norwood classification Ⅲ–VII	Hamilton–Norwood	None	Insulin	None	8
Wu,2023 [39]	case-control study	China	M	36.28 ± 10.49/36.28 ± 10.98	80	60	AGA Hamilton–Norwood classification Ⅰ–VII	Hamilton–Norwood	None	Insulin resistance	None	8
Acibucu, 2010 [40]	case-control study	Turkey	M	36.28 ± 7.74/35.14 ± 6.54	80	48	AGA Hamilton–Norwood classification Ⅲ–Ⅵ	Hamilton–Norwood	None	Insulin resistance	None	8
Bakry, 2014 [41]	case-control study	Egypt	M	40.09 ± 10.57/37.85 ± 9.48	100	100	AGA Hamilton–Norwood classification Ⅲ–Ⅵ	Hamilton–Norwood	None	Insulin resistance	None	8
Su, 2011 [42]	cross-section study	China	M	policeman: 38.6 ± 7.6 / not policeman: 65.2 ± 11.2	91	627	AGA Hamilton–Norwood classification Ⅲ–Ⅵ	Hamilton–Norwood	Family history of AGA	Dyslipidemia	UV exposure	5
Chumlea,2004 [43]	cross-section study	America	M	all the participants: 36.1 ± 8.9	97	61	AGA Hamilton–Norwood classification >Ⅲ	Hamilton–Norwood	Family history of AGA	None	None	7
Zhang, 2024 [44]	cross-section study	China	M	all the participants: 29.5 ± 7.5	all the participants: 1172		Moderate and severe AGA (Hamilton–Norwood classification IV–VII)	Hamilton–Norwood	None	None	Alcohol consumption	5
Wu, 2025 [45]	cross-section study	China	M/F	18–30 y	324	569	AGA Hamilton–Norwood classificationⅡ–Ⅵ	Hamilton–Norwood(male) or Sinclair Scale (female)	Family history of AGA	None	Insufficient sleeping time with poor quality	7
Peng, 2025 [46]	cross-section study	China	F	42.98 ± 13.26/54.61 ± 10.35	206	2802	AGA Savin Scale >Ⅰ-1	Savin Scale	None	Overweight	Smoking, Alcohol consumption, Insufficient sleeping time with poor quality	8

M Male, F Female, AGA Androgenetic alopecia, BMI Body mass index, UV Ultraviolet Rays, BASP Basic and specific

Study quality

The methodological rigor of all the incorporated articles was systematically evaluated, with the respective quality scores detailed in Table 1. This assessment yielded 11 studies classified as high-quality and 20 studies of medium-quality. There were no studies rated as low quality (Table 1).

Family history

A meta-analytic synthesis of 7 studies revealed a marked association between family history and AGA (pooled OR 2.72, 95% CI 1.85–3.99) (Fig. 2A). There was significant heterogeneity among the included studies (I² = 84%). Combining the funnel plot (Fig. S1) and the results of Egger’s test (p = 0.05, the intercept (5.72, 95% CI = -0.07 to 11.51)), it can be demonstrated that the publication bias is not significant. Sensitivity analysis did not change results (Fig. S2).

Fig. 2.

A Forest plot detailing the association of the presence of AGA with family history. B Forest plot detailing the association of the presence of AGA with parental family history. C Forest plot detailing the association of the presence of AGA with alcohol consumption. D Forest plot detailing the association of the severity of AGA with alcohol consumption. E Forest plot detailing the association of the presence of AGA with smoking. F Forest plot detailing the association of the severity of AGA with smoking. G Forest plot detailing the association of the presence of AGA with Insufficient sleeping time and/or poor sleeping quality. H Forest plot detailing the association of the severity of AGA with Insufficient sleeping time and/or poor sleeping quality

Due to the limited number of studies available, subgroup analyses were restricted to the AGA research focus (Fig. S3). Subgroup results reinforced the significant correlation that family history is significantly associated with both the presence and the progression of AGA. Although the subgroup analyses suggested an association, the observed LFK indices (1.92 and 2.67) indicate potential publication bias (Figs. S4-5).

Further investigation was conducted on the specific link between paternal family history and the presence of AGA, including 3 related studies in the analysis. The findings demonstrated a significant association between a paternal history of AGA and the presence of AGA (pooled OR 2.22, 95% CI 1.59–3.10) (Fig. 2B).

Lifestyle factors

Alcohol consumption

In a meta-analytical synthesis of 8 studies, our research found no significant link between alcohol consumption and the presence of AGA. (pooled OR 1.15, 95% CI 0.92–1.43) (Fig. 2C). The heterogeneity of the meta-analysis was moderate (I² = 30%). Although Egger’s test did not reach formal statistical significance (p = 0.10), the intercept (2.48, 95% CI = -0.60 to 5.55) and visual inspection of the funnel plot (Fig. S6) suggested a potential for publication bias, The results of the trim-and-fill analysis (pooled OR 0.97, 95% CI 0.76–1.24) made the conclusion more credible (Fig. S7). The consistent null finding before and after adjustment strengthens the conclusion of no link between alcohol consumption and AGA presence.

Further examination involving an additional 4 studies to assess the relationship between alcohol consumption and the severity of AGA, and the results showed a significant association (pooled OR 1.72, 95% CI 1.28–2.32) (Fig. 2D), with no heterogeneity among the studies (I² = 0%).

Smoking

A meta-analytic synthesis of 11 studies revealed a significant association between smoking and the presence of AGA (pooled OR 1.46, 95% CI 1.06–2.01)( Fig. 2E). However, heterogeneity was significant (I² = 86%). Subgroup analyses by gender, region, and size did not resolve the heterogeneity’s causes and LFK indices revealed considerable asymmetry (LFK > 1) in all but the Asian region subgroups, suggesting widespread publication bias. (Figs. S8-S14). Funnel plots (Fig. S15) and Egger’s test (p = 0.12, the intercept (-1.96, 95% CI = -4.53 to 0.61) suggested that there was no significant publication bias. Sensitivity analysis using the trim-and-fill method, which added 5 studies, suggested that the effect size remained robust (pooled OR 2.14, 95% CI 1.49–3.07) (Fig. S16).

A meta-analytic synthesis of 5 studies revealed a marked association between smoking and the progression of AGA (pooled OR 1.60, 95% CI 1.29–1.99),no heterogeneity was observed among the studies (Fig. 2F).

Sleeping

A meta-analysis of 5 studies was conducted to examine the relationship between insufficient/poor quality sleep and the presence of AGA. The results did not reveal a significant correlation between sleep deprivation/poor sleep quality and the presence of AGA (pooled OR 1.28, 95% CI 0.86–1.93) (Fig. 2G). There was considerable heterogeneity among the studies (I² = 78%). However, the combined effect sizes from 3 additional studies suggested a significant association between insufficient/poor quality sleep and the progression of AGA (pooled OR 1.36, 95% CI 1.18–1.56) (Fig. 2H), with no heterogeneity detected among these studies (I² = 0%).

Comorbidities

Overweight/Obesity and dyslipidemia

The association between overweight/obesity and AGA was analyzed by including 8 studies on presence and 3 on progression. The initial analysis indicated a borderline association between overweight/obesity and the presence of AGA (pooled OR 1.32, 95% CI 1.05–1.66) (Fig. 3A). Although a visual inspection of the funnel plots suggested a slight asymmetry (Fig. S17), Egger’s test results (p = 0.21, the intercept (1.47, 95% CI = -4.53 to 0.61)) did not indicate a significant publication bias. To further verify the robustness of the results, we performed a trim-and-fill analysis. This adjustment yielded a pooled OR of 1.16 (95% CI 0.91–1.48), which is no longer statistically significant. This indicates that the initial association is not robust and should be interpreted cautiously (Fig. S18). No statistically significant association between overweight /obesity and presence of AGA was observed in either the male or female subgroup and the LFK index indicated substantial bias in both subgroups. (Figs. S19-S21). However, the results revealed a significant association between being overweight or obese and the progression of AGA (pooled OR 2.31, 95% CI 1.01–5.29) (Fig. 3B).

Fig. 3.

A Forest plot detailing the association of overweight/obesity with the presence of AGA. B Forest plot detailing the association of overweight/obesity with the severity of AGA. C Forest plot detailing the association of hypertension with AGA. D Forest plot detailing the association of HOMA-IR with the presence of AGA. E Forest plot detailing the association of the level of fasting insulin with the presence of AGA

Another 6 studies were included to analyze the relationship between dyslipidemia and AGA. These analyses, however, found no significant correlation between dyslipidemia and AGA, and there was high heterogeneity among the studies (I² = 81). The instability of the results and the clear presence of publication bias (Figs.S22-24) were noted.

Hypertension

A meta-analysis of 7 studies identified a significant correlation between hypertension and the presence and progression of AGA (pooled OR 1.32, 95% CI 1.03–1.70) (Fig. 3C). Nonetheless, there was notable heterogeneity among these studies (I² = 61%). In order to better understand the sources of heterogeneity, subgroup analyses were conducted, which indicated that gender was the primary contributor to the heterogeneity (Fig. S25). Specifically, subgroup analysis revealed a significant correlation between AGA in men and hypertension (pooled OR 1.60, 95% CI 1.26–2.04), while this correlation was not observed in women (pooled OR 1.02, 95% CI 0.93–1.11); the LFK index indicated no significant asymmetry in either subgroup(Figs. S26-27). To test for publication bias, although a visual inspection of the funnel plot did reveal some asymmetry (Fig. S28), the result of Egger’s test (p = 0.08) did not indicate significant publication bias.

Insulin resistance and fasting insulin levels

A meta-analysis was carried out to investigate the potential associations of insulin resistance and fasting insulin levels with AGA. 5 studies were included for insulin resistance and 9 for fasting insulin levels. The results indicated that both factors were significantly associated with the presence of AGA. Individuals with AGA exhibited notably elevated HOMA-IR (Homeostatic Model Assessment for Insulin Resistance) scores when compared to the control group (standardized mean difference [SMD] 0.40, 95% CI 0.27–0.53) (Fig. 3D). There was moderate heterogeneity among these studies (I² = 44%). Visual inspection of the funnel plot (Fig. S29) and the results of Egger’s test (p = 0.14) indicated an absence of significant publication bias among the studies. In parallel, the analysis showed that fasting insulin levels were significantly increased than non-AGA individuals (SMD 0.48, 95% CI 0.18–0.78) (Fig. 3E). The results were stable, as indicated by the leave-one-out sensitivity analysis. Further scrutiny of the funnel plots did not suggest any meaningful publication bias (Figs. S30-31).

Discussion

In this systematic review and meta-analysis, we identified significant correlations between AGA and several potential risk factors (Table 2). Other understudied potential.

Table 2.

Summary of meta-analysis results and key messages for risk factors in the study

Risk factors	Number of Studies	Pooled Estimate (95% CI) /SMD (95% CI)	I²(%)	Heterogeneity Between Groups P Value	Subgroup analysis	Results of sensitivity analysis	Publication bias	Focus on AGA
AGA family history	7	2.72(1.85–3.99)	84	< 0.001	Yes	Stable	Non-significant	Severity and presence*
AGA parental family history	3	2.22(1.59–3.10)	0	0.77	No	None	None	Presence*
Smoking	11	1.46(1.06–2.01)	86	< 0.001	Yes	Stable	Non-significant	Presence*
Smoking	5	1.60(1.29–1.99)	0	0.88	No	None	None	Severity*
Alcohol consumption	8	1.15 (0.92–1.43)	36	0.19	No	None	Significant	Presence#
Alcohol consumption	4	1.72 (1.28–2.32)	0	0.91	No	None	None	Severity*
Insufficient sleeping time with poor quality	5	1.28 (0.86–1.93)	78	< 0.001	No	None	None	Presence#
Insufficient sleeping time with poor quality	3	1.36 (1.18–1.56)	0	0.78	No	None	None	Severity*
Overweight	8	1.32 (1.05–1.66)	78	< 0.001	Yes	Stable	Non-Significant	Presence#
Overweight	3	2.31 (1.01–5.29)	81	< 0.01	No	None	None	Severity*
Hypertension	7	1.32(1.03–1.70)	61	0.02	Yes	None	Non-significant	Severity and presence*
Dyslipidemia	6	1.14 (0.97–1.34)	81	< 0.01	No	Unstable	Significant	Severity and presence
Insulin resistance	5	0.40 (0.27–0.53)	44	0.11	No	None	Non-significant	Presence*
Fasting insulin levels	9	0.48 (0.18–0.78)	89	< 0.01	Yes	Stable	Non-significant	Presence*

^*Significant correlation

^#Suggestive but not statistically significant

risk factors include unhealthy diets [23, 47], nutritional deficiencies [30], excessive working hours [48], other unhealthy lifestyles, excessive exercise [21, 47, 48], exposure to ultraviolet rays [18, 20, 42], stress and anxiet [29], occupational factors [42, 49], and comorbidities in AGA such as hyperandrogenemia and polycystic ovary syndrome (Tables S1-3) [7].

In AGA, a polygenic hereditary disease, family history plays a critical and clearly established role in its occurrence, particularly concerning paternal lineage, which aligns with the findings of this study [43, 50]. Two studies indicate that males are at higher risk of developing AGA if their mother or maternal grandfather has AGA; this risk increases further if AGA exists in both their paternal and maternal lineages. Nonetheless, the current body of research does not yet provide enough data for a thorough meta-analysis [25, 43]. Our study further validates that family history is a contributing factor to the progression of AGA. Additionally, Su et al.‘s research indicates an increased risk of AGA progression with a family history in first- or second-degree relatives [17]. Regarding lifestyle factors, our study quantitatively supports that smoking and alcohol consumption are significant risk factors for AGA, with the former significantly associated with both the incidence and progression of AGA, while the latter is only associated with its progression. Our findings on smoking’s link to male AGA align with a recent meta-analysis [51], yet no significant association was confirmed between smoking and the occurrence of AGA in women, possibly due to lower smoking rates among females. The mechanisms by which smoking exacerbates hair loss may include impaired follicular blood circulation, increased oxidative stress, production of dihydrotestosterone, protease system imbalance, and induced low estrogen states [52]. Although studies suggest that excessive alcohol consumption may lead to metabolic syndrome [53], which can cause hair loss, our research did not find a significant link between drinking and the presence of AGA. Secondly, insufficient/poor-quality sleep and prolonged UV exposure may exacerbate AGA by increasing scalp oxidative stress or promoting pro-inflammatory factors in the body [54–56].Lastly, the relationship between exercise and AGA still requires further validation; while exercise can raise levels of reactive oxygen species and testosterone, potentially promoting hair loss, aerobic exercise may benefit scalp circulation [57, 58, 59]. Regarding comorbidities of AGA, previous systematic reviews have shown an association between AGA and diseases across various systems,⁷and meta-analyses have indicated that AGA patients have higher blood pressure and BMI compared to the normal population [8]. This study pioneers the use of meta-analysis to investigate the connection between being overweight/obesity, hypertension, dyslipidemia, and the presence and progression of AGA. Firstly, being overweight/obesity can promote the progression of AGA. Overweight/obesity is often associated with a high-fat diet, has been implicated in AGA pathogenesis through the potential epidermalization of hair follicle stem cells, as demonstrated in animal models [60].However, the relationship between dyslipidemia and AGA remains unresolved, aligning with previous meta-analyses on AGA and lipid profiles [14]. Secondly, our analysis substantiates a significant link between hypertension and AGA in men, but not in women. Regarding the relationship between hypertension and AGA, prior studies have proposed two possibilities: one hypothesis suggests that the binding of androgens to vascular receptors may lead to an increase in blood pressure. Another explanation posits that most patients with hypertension have a condition called hyperaldosteronism, which could directly be involved in the presence and progression of hair loss [61–63]. Lastly, the results suggest insulin resistance and high fasting insulin levels as potential risk factors for AGA. Earlier research has demonstrated a strong correlation between diabetes and female pattern hair loss [64], and proposes AGA as an early clinical indicator of insulin resistance [65]. This could stem from insulin resistance disturbing androgen balance and insulin’s role in follicular miniaturization through perifollicular vascular impacts [66, 67].

To mitigate AGA and slow its advancement, forthcoming studies could take several strategic directions. Initially, robust cohort studies are essential to clarify the causal relationships between various risk factors and AGA. Subsequent research should focus on predictive models for AGA to assess individual development and progression risks. Future studies ought to employ randomized controlled trials or analyze real-world data to appraise AGA treatments and outcomes, tailoring to case-specific severity and progression risks.

Our study has several limitations. Primarily, the studies’ moderate quality and the absence of cohort studies might introduce bias, weakening the causal link between risk factors and AGA. To mitigate confounding, we only included effect sizes adjusted for at least age and family history. Furthermore, limiting our review to English and Chinese-language studies could have excluded relevant research. Additionally, we were unable to further explore heterogeneity among certain risk factors due to the small number of included studies. Combining studies with disparate outcomes due to scarce research may compromise analytical rigor. Justifying this, early AGA symptoms’ subtlety contrasts with the focus on moderate to severe cases in most studies. Thus, AGA presence comparisons may inadvertently contrast mild with advanced stages.

Conclusion

In summary, our study suggests that the presence and progression of AGA may be associated with multiple risk factors. These include a family history of AGA, smoking, alcohol consumption, insufficient or poor-quality sleep, overweight/obesity, hypertension in men, insulin resistance, and elevated fasting insulin levels. Furthermore, other risk factors related to AGA require in-depth investigation. Upcoming research should aim to quantify these associations, create AGA risk prediction models, and offer detailed preventative and therapeutic strategies for AGA management.

Abbreviations

AGA

Androgenetic alopecia

OR

Odds Ratio

CI

Confidence Interval

SMD

Standardized Mean Difference

LFK

Luis Furuya-Kanamori

BMI

Body Mass Index

Authors’ contributions

All authors agree with the manuscript content. Yong Miao, Haoyang Li and Jiaxian Zhang led study design and prepared the manuscript. The search strategy development and literature search were performed by Haoyang Li and Yingjie Zhao. Data collection and interpretation were conducted by Haoyang Li, Wenzhen Li, Yingjie Zhao, Zehong Guo, Botian Jiang and Zhan Wang. Statistical analysis was performed by Haoyang Li, Qihong Liang. Shengli An and Qian Qu provided many suggestions during study design and manuscript preparation. All authors read and approved the final manuscript.

Funding

This study was funded by the National Natural Science Foundation of China (Grant No.82372538) and the Natural Science Foundation of Guangdong Province (Grant No.2023A1515012827).

Data availability

All data analysed during this study are included in this published article.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.