Overall, of all the factors analysed in this study, age, ethnicity, BMI, cycle length, and parity had a significant relationship with serum AMH despite adjusting for the influence of all significant factors.
Demographic factors
We found that with a unit increase in age there was a 3.5% reduction in serum AMH level. This is similar to findings from previous studies which suggests that serum AMH levels reduce with age and hence is a reliable marker of ovarian ageing [19, 20]. However, on multivariable analysis, there was a reduction in percentage reduction in serum AMH by age by 0.4%. Women 35 years and older had a rate of decrease in serum AMH with age that was 6 times the rate of decrease noted in those less than 35 years.
Yoruba women in this study were found to have a 15.5% higher serum AMH level than women of other ethnic groups. This is similar to findings from previous studies that demonstrated a significant ethnic and racial variation in serum AMH levels in women in the same population [15, 16]. This could be due to some lifestyle and genetic variations among these ethnic groups [17]. However, given the lack of adequate representation of the major ethnic groups in this study, a population study of a larger sample size, with better representation of the different ethnic groups will give a better understanding of the influence of ethnicity on serum AMH in an ethnically diverse population like Nigeria.
Lifestyle factors
All women recruited in this study were non-smokers (active). However, about 9.9% of the study population reported exposure to second-hand smoking. There was no significant association between passive smoking and serum AMH. This is in line with the findings by Plante et al. and further lends credence to their conclusion that the influence of smoking on the ovarian reserve is most likely from a direct influence that is gotten from active and current smoking [21]. However, given the small proportion of participants exposed to passive smoking, a larger study with a larger proportion of second-hand smokers will better depict the extent of the impact this has on ovarian reserve.
There was no association found between alcohol intake and serum AMH level, similar to what was found in previous studies [9, 22]. Although this was contrary to findings by Bressler et al., who found a 26% lower AMH in binge drinkers among African-American women [10]. This contradiction is likely because the majority of women who reported alcohol consumption in our study are social drinkers and reported infrequent alcohol intake. Therefore, the effect of alcohol on AMH is likely quantity dependent.
There was no significant relationship between caffeine intake and serum AMH. This finding is in line with the study done by Kinney et al. where no association was found between caffeine consumption and AFC, FSH, estradiol, and inhibin [22]. However, it is important to note that our study did not make a more objective quantification of the amount of caffeine consumed in these women. An objective measure of the quantity of caffeine consumed by each woman will be a better predictor of the influence of caffeine on serum AMH levels.
However, we found a positive relationship between BMI and serum AMH, with an increase in BMI, leading to a corresponding increase in serum AMH levels. After adjusting for age, ethnic group, cycle length, and parity, a unit increase in BMI resulted in a 1.4% increase in serum AMH. This contradicts the findings by previous studies in southwest Nigeria, in which researchers found no significant relationship between BMI and serum AMH [14, 23]. Freeman et al. found a negative relationship in women of late reproductive age outside Nigeria [13]. The discrepancies between these findings are likely due to the inclusion of women with polycystic ovary syndrome (PCOS) in this study, which is often associated with high BMI and high serum AMH [24]. Also on sub-analysis by age group, BMI remained significant in women less than 35 years only.
Reproductive factors
There was no association demonstrated between age at menarche and serum AMH in this study, which is in line with findings from the study by Dólleman et al. [9]. However, Bragg et al. reported that independent of age, smoking, and body mass index; women with early age at menarche had significantly higher AMH as young adults [10], whereas another study found that infertile women with early age at menarche were at a significantly higher risk of diminished ovarian reserve in the long-term [11]. The contradiction between these studies and ours is most likely due to the difference in population, as the initial study focused more on young adult women as opposed to women within the entire reproductive age, whereas the latter was done in infertile women only.
There was a significant positive relationship between cycle length in days and serum AMH in the study population. An increase in cycle length by a day was associated with a 3.6% increase in serum AMH independent of age, ethnicity, BMI, and parity. This is most likely due to the inclusion of women with PCOS in the study group. Anovulatory women with PCOS have relatively longer cycle lengths with relatively high random serum AMH levels [24]. However on sub-analysis, by age group, this relationship was not statistically significant in women 35 years of age and older.
Cycle regularity was not associated with serum AMH, contradicting findings by Dólleman et al. where cycle irregularity was associated with a significantly less serum AMH level [9]. This finding is unexpected as irregular menstrual cycles are usually found in women with PCOS in whom it is associated with high serum AMH or older women with menopause-related irregularities who are more likely to have low levels of AMH. Women in this study who reported irregular menstrual cycle had an average age of 30 years; therefore, cycle irregularity was likely, not due to menopause-related irregularity.
However, the experience of dysmenorrhea was not associated with serum AMH levels in this population. This is contrary to a previous study in which women with severe dysmenorrhoea had significantly lower serum AMH levels [25]. Their findings could be linked to endometriosis which some studies have suggested could significantly decrease serum AMH and can also present with severe dysmenorrhea [26]. The lack of association in our study could be due to the lack of accountability for the severity of dysmenorrhea experienced by women in this study. Also, from clinical records, none of the women recruited was diagnosed with endometriosis.
Furthermore, there was a negative relationship between parity and serum AMH with an increase in parity being associated with a 5.3% decrease in serum AMH after adjusting for age, ethnicity, BMI, and cycle length. This contradicts the findings by Dólleman and colleagues who found an increase in age-specific AMH was associated with an increase in parity [9]. However, similar findings were reported by Bragg et al. who found that women with 2 and 3 or more parous experiences had significantly lower serum AMH levels when compared with nulliparous women [10]. However, when the analysis was adjusted for age, the negative relationship persisted in multivariable regression and sub-analysis of women younger than 35 years and women 35 years and older. Also, given that women who were currently on contraceptives were excluded in this study, this effect is likely not due to the higher likelihood of contraceptive use by women with higher parity.
Having a history of infertility did not have a significant relationship with serum AMH levels in this study. This is similar to findings by Hvidman et al. who reported no difference in AMH and antral follicular count in fertile and infertile women of similar age groups [27]. This is not unexpected as various factors are responsible for infertility in women, outside of their intrinsic ovarian reserve. However, there was a significant negative relationship between the duration of infertility and serum AMH, which implies that with an increase in the duration of infertility in years, there is a corresponding reduction in serum AMH levels. This is contrary to findings by Oke et al. in Southwest Nigeria, in which authors reported no statistically significance [23]. Our finding is likely due to the influence of age on serum AMH, as older women are more likely to have experienced infertility for a longer period. This is evidenced by the loss of statistical significance after adjusting for age in regression analysis.
Lastly, past hormonal contraceptive use was not associated with serum AMH levels in this study. Although various studies have reported a decrease in AMH with the use of hormonal contraception, this suggests that past use of hormonal contraceptives does not permanently affect ovarian reserve [28,29,30]. This is similar to findings by Van den Berg et al. who found a 23% increase in serum AMH after discontinuation of oral contraceptive pills [31].
Strengths and limitations
The strength of this study is in its evaluation of 11 independent variables and their association with serum AMH in the same sample of Nigerian women in a clinical setting, where the results are most applicable.
Although the sample size was statistically adequate for the objectives of this study, a large, prospective population-based study is essential to better study the demographic, lifestyle, and reproductive factors that are associated with serum AMH as a marker of ovarian reserve and functioning. The relative homogeneity in terms of ethnicity and the hospital-based nature of this study makes it deficient in determining the true extent to which these factors influence AMH in a culturally and ethnically diverse population like Nigeria.
Also, the study did not measure objectively the amount of alcohol and coffee intake and did not document the severity of dysmenorrhea in the women in this study. This could contribute to the lack of association demonstrated; therefore, future research should aim to objectively qualify and quantify these factors to arrive at a better conclusion.
Furthermore, endometriosis could not be appropriately ruled out as laparoscopy and histology are not routinely done in women who had presented with infertility, although clinical history and records revealed no history suggestive of endometriosis in women in this study.