Joint Effects of Smoking and Gene Variants Involved in Sex Steroid Metabolism on Hot Flashes in Late Reproductive-Age Women

Abstract

Background:

Although smoking has a known association with hot flashes, the factors distinguishing smokers at greatest risk for menopausal symptoms have not been well delineated. Recent evidence supports a relationship between menopausal symptoms and variants in several genes encoding enzymes that metabolize substrates such as sex steriods, xenobiotics, and catechols. It is currently not known whether the impact of smoking on hot flashes is modified by the presence of such variants.

Objective:

The objective of the study was to investigate the relationship between smoking and hot flash occurrence as a function of genetic variation in sex steroid-metabolizing enzymes.

Methods:

A cross-sectional analysis of data from the Penn Ovarian Aging study, an ongoing population-based cohort of late reproductive-aged women, was performed. Smoking behavior was characterized. Single-nucleotide polymorphisms in five genes were investigated: COMT Val158Met (rs4680), CYP1A2*1F (rs762551), CYP1B1*4 (Asn452Ser, rs1800440), CYP1B1*3 (Leu432Val, rs1056836), and CYP3A4*1B (rs2740574).

Results:

Compared with nonsmokers, European-American COMT Val158Met double-variant carriers who smoked had increased odds of hot flashes [adjusted odds ratio (AOR) 6.15, 95% confidence interval (CI) 1.32–28.78)]; European-American COMT Val158Met double-variant carriers who smoked heavily had more frequent moderate or severe hot flashes than nonsmokers (AOR 13.7, 95% CI 1.2–154.9). European-American CYP 1B1*3 double-variant carriers who smoked described more frequent moderate or severe hot flashes than nonsmoking (AOR 20.6, 95% CI 1.64–257.93) and never-smoking (AOR 20.59, 95% CI 1.39–304.68) carriers, respectively. African-American single-variant CYP 1A2 carriers who smoked were more likely to report hot flashes than the nonsmoking carriers (AOR 6.16, 95% CI 1.11–33.91).

Conclusion:

This is the first report demonstrating the effects of smoking within the strata of gene variants involved in sex steroid metabolism on hot flashes in late reproductive-age women. The identification of individuals with a genetic susceptibility to smoking-related menopausal symptoms could contribute to interventions targeted at reducing reproductive morbidity both in the menopause and across the reproductive life course.

Despite aggressive public health efforts to curb tobacco use, smoking remains a prevalent behavior in this country, especially in women. Recent estimates from the Centers for Disease Control and Prevention indicate that 18.3% of the adult female population smoke, a figure that represents more than 20 million women (1). Smoking has a known association with earlier age at menopause, increased odds of hot flashes, and risk of postmenopausal osteoporosis (2–9). Multiple reports have described a 2- to 3-fold increase in reporting of any and of severe hot flashes in smokers compared with nonsmokers (3, 7). Tobacco smoke is thought to promote estrogen depletion; women who smoke have lower serum estrogen levels and higher concentrations of urinary estrogen metabolites than nonsmokers (10).

The metabolism of estrogens involves oxidation by cytochrome P450-containing enzyme (CYP) enzyme families, CYP1 and CYP3, as an initial step (phase I enzymes) (11, 12) (Supplemental Fig. 1, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org). Phase I metabolites are further processed through several distinct phase II pathways: glucuronidation by uridine diphosphate-glucuronosyltransferase, sulfation by sulfotransferase, and O-methylation by catechol O-methyltransferase (COMT) (12–16).

The intrinsic factors that distinguish smokers at greatest risk for menopausal morbidity and symptoms have not been well delineated. Tobacco smoke represents a mixture of more than 4000 chemicals, several capable of disrupting endocrine pathways that could affect menopausal end points (17). Nicotine may influence the likelihood of hot flashes by inhibiting aromatization of androgens to estrogens (18, 19) and stimulating release of dopamine (20, 21). Polyaromatic hydrocarbons bind the aryl hydrocarbon receptor, a member of the helix-loop-helix family of transcription factors, to create a complex that recognizes DNA-specific CYP promoter sequences to up-regulate transcription (9, 17, 22–25).

A growing body of evidence supports a relationship between menopausal symptoms and variants in genes involved in sex steroid metabolism. It has been demonstrated that Chinese women who are carriers of CYP 1A1 rs2606345 variants report diminished vasomotor symptoms compared with noncarriers (26). An association between CYP 1B1 variants in African-American (AA) women and symptom reporting has been described by several groups (26–28). Despite evidence that smoking and genetic variation in sex steroid metabolism impact menopausal end points separately, data addressing possible interactions between variants in sex steroid-metabolizing genes and smoking behavior on the risk of menopausal symptoms are lacking.

The goal of this investigation was to assess the odds and severity of hot flashes in reproductively aging smokers as a function of genotypes in enzymes involved in sex steroid metabolism. Our primary hypothesis was that specific variants in these enzymes could modify the increased likelihood hot flashes observed in smokers. Support for this hypothesis derives from the following considerations: 1) the antiestrogenic effect of tobacco smoke is in part mediated by induction of CYP enzymes, 2) wild-type alleles of genes encoding steroid metabolizing enzymes that are inducible by tobacco smoke may display altered inducibility when variants in these genes are present, and 3) a combined effect of smoking and gene variants in steroid-metabolizing enzymes could develop if the variant differentially metabolizes nonsteroidal hormones (i.e. neurotransmitters) that mediate unique elements of hot flash development.

Materials and Methods

Study cohort

Participants from the Penn Ovarian Aging Study, a longitudinal population-based study of hormonal changes and symptoms in late reproductive-aged women, were studied. Eligibility criteria for the study at enrollment included age 35–47 yr, menstrual cycles in the normal range (22–35 d) for 2 months before enrollment, an intact uterus, and at least one ovary (2, 29–31). Exclusion criteria included use of hormonal or psychotropic medications, hysterectomy, pregnancy or lactation, serious health problems known to compromise ovarian function, and drug or alcohol abuse in the year before enrollment. The study was approved by the Institutional Review Board of the University of Pennsylvania, and written informed consent was obtained from all participants. The cohort, which has been described in detail elsewhere (2, 29–31), recruited 436 women by random digit dialing and stratified enrollment to achieve balance in proportions of women by race (218 AA and 218 European-American). Clinical data for this report were collected from participants 11 yr after study initiation. At this point, 296 participants had complete clinical, smoking, and genetic data and had not met any of the stated exclusion criteria.

Data collection

During each study assessment period, two visits were conducted, each occurring between d 1 and 6 of two consecutive menstrual cycles or approximately 4 wk apart in noncycling women. At each visit, a trained research coordinator conducted a standardized interview, collected blood samples, and measured height and weight to determine body mass index (BMI). The interview focused on overall health, and the interviewer collected information regarding demographics; reproductive and menstrual history; menopausal symptoms; medication; history of depressive disorders; and behaviors such as smoking, alcohol, and caffeine consumption.

Laboratory measures

Nonfasting blood samples for hormone assays were collected between d 2 and 6 of the menstrual cycle in two consecutive cycles (or at monthly intervals in noncycling women) during each assessment period. The samples were centrifuged and frozen in aliquots at −80C. Estradiol was measured by RIA using Coat-A-Count commercial kits (Diagnostic Products Corp., Los Angeles, CA). Assays were performed in duplicate and repeated if values differed by more than 15%. The intraassay and interassay coefficients of variation were less than 8% and less than 20%, respectively, for concentrations of 50 to 500 pg/ml; the analytical sensitivity was 15 pg/ml.

DNA was extracted from peripheral blood leukocytes using the QIAamp 96 DNA buccal swab biorobot kit and performed on a 9604 Biorobot (QIAGEN, Inc., Valencia, CA) and amplified by PCR. Genotypes were determined using previously described methods (32, 33). Five functionally relevant single-nucleotide polymorphisms (SNP) in four genes were selected for study. All gene variants were involved in the downstream metabolism of estrogen with sufficient allele frequency to provide adequate power to test for possible smoking-gene interactions. The SNP included the following: COMT Val158Met (rs4680), CYP1A2*1F (rs762551), CYP1B1*4 (Asn452Ser, rs1800440), CYP1B1*3 (Leu432Val, rs1056836), and CYP3A4*1B (rs2740574). Genotype coding was based on knowledge of the predicted function of the variants as well as the frequency of genotypes of interest. To identify potential deviations from the Hardy Weinberg equilibrium, exact tests of expected genotype proportions in each racial group were calculated.

Clinical measures

Assessment of the presence of hot flashes was ascertained from a validated symptom list embedded in the structured interview that was administered at each assessment. Participants were queried on the occurrence of hot flashes within the past month, the frequency of hot flashes, and the severity of hot flashes on a 4-point scale ranging from 0 (none) to 4 (severe).

Subjects were primarily classified by smoking status as current smokers or nonsmokers. Smoking status was further characterized as three categories: current, former or never smoker. Smoking heaviness was captured by the following categories: heavy smokers (at least 20 cigarettes daily), light smokers (less than 20 cigarettes daily), and nonsmokers. Menstrual cycle lengths were determined from the menstrual date at each study interview and the two previous menstrual dates recorded in the interview at each assessment period. Confirmatory dates were obtained from the menstrual diaries recorded by the participants for one menstrual cycle at each assessment period. The definitions of menopausal stages were adapted from the staging system for reproductive aging in women that we have validated in previous reports (30, 34, 35). At each assessment, the participant was assigned to one of the following categories based on menstrual bleeding patterns at that assessment: 1) premenopausal, regular menstrual cycles in the 22- to 35-d range; 2) late premenopausal, change in cycle length 7 d in either direction compared with the participant's personal baseline at enrollment in the cohort and observed for one cycle in the study; 3) early transition, change in cycle length 7 d in either direction from the participant's personal baseline at enrollment in the cohort and observed for at least two cycles in the study; 4) late transition, 3–11 months amenorrhea during the study; or 5) postmenopausal, 12 months amenorrhea with no hysterectomy. For the purposes of this investigation, these five categories were collapsed into three stages: premenopausal, transition, and postmenopausal.

Alcohol consumption was queried at each follow-up assessment. Study participants were asked how many alcoholic beverages they consume each day with a recall period of the prior 6 months. Subject responses were dichotomized based on whether one more drinks were consumed daily.

Subjects completed standard self-report questionnaires to assess anxiety. The Zung Anxiety Index is a validated, 20-item, self-report measure that is sensitive to the frequency of affective and somatic anxiety symptoms and was used to assess anxiety in the Penn Ovarian Aging Study (36).

Racial categorization (AA/Black or European-American/white/Caucasian) was determined by self-report upon cohort entry.

BMI was calculated as weight (kilograms) divided by the square of height (meters). Height was measured without shoes to the nearest 0.25 in.; weight was measured in light clothing to the nearest 0.50 lb. Each measurement was taken twice, and the average of the two measures was used in the calculation of BMI at each study period. The following BMI categories were generated for statistical analysis: less than 25 kg/m², 25–29.9 kg/m², 30–34.9 kg/m², 35 kg/m² or greater.

Statistical analysis

Comparisons were stratified by race. Continuous data were analyzed using the Mann-Whitney U test or Kruskal Wallis test as appropriate. Categorical data were analyzed with the χ² test. Odds ratios were calculated to perform initial tests of association between smoking and any hot flashes and hot flash severity dichotomized as moderate or severe compared with none or mild. Adjusted odds ratios for the association between smoking and symptoms and for the association between each genotype and symptoms were calculated using logistic regression modeling.

Multivariable logistic regression was also used to determine odds ratios of any hot flashes and moderate or severe hot flashes in the combined presence of smoking and selected SNP and controlling for multiple confounders. For each SNP studied, models with and without interaction terms were compared to determine whether the association between smoking and hot flashes was modified by genetics. Linear combinations of coefficients from the models with interaction terms permitted an assessment of the adjusted odds ratio of hot flashes in smokers compared with nonsmokers in strata of each variant (noncarrier, single carrier, double carrier). Separate models were fit and interactions tested for smoking as a categorical variable (never, former, current; nonsmoker, light smoker, heavy smoker). Hot flash models that included gene variants were initially adjusted for BMI, alcohol consumption, menopausal stage, and age; a final set of models was adjusted for those covariates and anxiety.

Hormone measurements were reported as geometric means. The two hormone values obtained in each assessment period were averaged for each participant, with the mean value of the two hormone measurements for that assessment period used in the analysis. In the cases in which two hormone values were not obtained in an assessment period, the single value was used for the mean. Statistical analyses were performed using STATA 9 software (College Station, TX).

Results

Demographic and clinical data for the cohort are presented in Table 1. The median age of the group was 50.6 yr and was nearly equal in the proportion of AA (47%) and European-American women (EA) (53%). Smoking was prevalent in all women studied (31.8%); a higher proportion of AA than EA women reported current smoking (38.1 and 26.1%, respectively), but heavy smoking was more prevalent in EA women (21% compared with 14.4% in AA). AA participants were also more likely than EA to report hot flashes and had significantly higher BMI. No racial differences in menopausal stage, current age, anxiety, or alcohol consumption were found.

Table 1.

Demographic and clinical data

Factor	European-Americans (n = 157)	African-Americans (n = 139)	P value
Median age (yr) (interquartile range)	50.7 (48.1–53.3)	50.4 (47.3–53.5)	0.6
Smoking status			0.08
Current smokers	26.1% (41)	38.1% (53)
Former smokers	40.8% (64)	35.3% (49)
Never smokers	33.1% (52)	26.6% (37)
Heavy smokers	21% (33)	14.4% (20)	0.07
Menopausal stage			0.3
Premenopausal	6.7% (10)	3.9% (5)
Menopausal transition	56% (84)	63.9% (83)
Postmenopausal	37.3% (56)	32.3% (42)
Alcohol consumption (one or more alcoholic beverages daily)	17.8% (28)	11.5% (16)	0.1
Median BMI (interquartile range)	27 (24.4–31.5)	32.8 (26.9–38.2)	<0.0001
Feelings of anxiety	51.6% (81)	48.9% (68)	0.6
Presence of hot flashes	54.1% (85)	68.4% (97)	0.01
Presence of moderate or severe hot flashes	30.6% (48)	41% (57)	0.06

Some proportions may total more than 100% due to rounding.

Table 2 lists carrier frequency for each SNP across racial categories. As anticipated, the SNP carrier frequency varied significantly across AA and EA women. However, no deviation from expected Hardy-Weinberg proportions was demonstrated in either racial group.

Table 2.

Gene variant functions and carrier frequencies by race

Gene and SNP nucleotide designation	Gene function/variant function	European-Americans		African-Americans
		Carrier frequency (n)	Proportion missing (n)a	Carrier frequency (n)	Proportion missing (n)a
COMT Val158Met (rs4680)	Metabolism of estrogens and neurotransmitters/decreased enzyme activity	0.747 (115)b	0.019 (3)	0.571 (76),b P = 0.001c	0.043 (6)
CYP1A2*1F (rs762551)	Metabolism and hydroxylation of estrogens, activity induced by tobacco and estrogens/increased inducibility and metabolism of estrogens	0.890 (137)b	0.019 (3)	0.860 (117),b P = 0.4c	0.022 (3)
CYP1B1*3 (Leu432Val, rs1056836)	Metabolism and hydroxylation of estrogens, highly expressed in the ovary, induced by tobacco and involved in metabolic activation of smoking toxins/increased metabolic activity	0.673 (103)b	0.026 (4)	0.941 (128),b P < 0.0001c	0.022 (3)
CYP1B1*4 (Asn452Ser, rs1800440)	Metabolism and hydroxylation of estrogens/increased metabolic activity	0.303 (46)b	0.026 (4)	0.096 (13),b P < 0.0001c	0.022 (3)
CYP3A4*1B (rs2740574)	Metabolism and hydroxylation of estrogens, activation of tobacco-related toxins/increased metabolic activity	0.844 (114)b	0.013 (2)	0.085 (13),b P < 0.0001c	0.029 (4)

^aProportion of subjects for whom genotyping could not be performed.

^bIndicates no deviation from Hardy-Weinberg proportions at the P < 0.05 level.

^cP value for difference in carrier frequency across race.

Univariate analysis confirmed smoking as a risk factor for the presence and severity of hot flashes. Overall, smokers were twice as likely as nonsmokers to report hot flashes [odds ratio (OR) 1.95, 95% confidence interval (CI) 1.12–3.45)] and were more than twice as likely to report moderate to severe hot flashes (OR 2.17, 95% CI 1.27–3.71). Stratifying on race and controlling for multiple covariates attenuated these associations (Table 3). Although AA smokers had a greater than 2-fold increased odds of moderate or severe hot flashes compared with nonsmokers (OR 2.2, 95% CI 1.03–4.71), no significant smoking and hot flash associations were observed in EA women.

Table 3.

Associations between smoking, hot flashes, and hot flash severity

Current smoking status	HF	No HF	OR of HF (95% CI)	Adjusted OR of HF (95% CI)a	Moderate or severe HF	Minimal or no HF	OR of moderate or severe HF (95% CI)	Adjusted OR of moderate or severe HF (95% CI)a
EA	(n = 85)	(n = 72)			(n = 48)	(n = 109)
No	69.4% (59)	79.2% (57)	Referent	Referent	64.6% (31)	78% (85)	Referent	Referent
Yes	30.6% (26)	20.8% (15)	1.67 (0.76–3.76), P = 0.2	1.56 (0.71–3.42), P = 0.3	35.4% (17)	22% (24)	1.94 (0.85–4.34), P = 0.08	1.89 (0.84–4.24), P = 0.1
AA	(n = 95)	(n = 44)			(n = 57)	(n = 82)
No	56.8% (54)	72.7% (32)	Referent	Referent	50.9% (29)	69.5% (57)	Referent	Referent
Yes	43.2% (41)	27.3% (12)	2.04 (0.89–4.86), P = 0.07	1.84 (0.79–4.29), P = 0.2	49.1% (28)	30.5% (25)	2.2 (1.03–4.71), P = 0.03	2.03 (0.93–4.42), P = 0.08

HF, Hot flashes.

^aAdjusted for age, BMI, alcohol consumption, and menopausal stage.

Associations between genotype and hot flash occurrence and severity are presented in Table 4. Data regarding the association between hot flashes and candidate genotypes have previously been published from this cohort using repeated-measures extensions of logistic regression to properly consider longitudinal data and symptom reporting (27). The OR reported herein are based on one cycle of reporting from the Penn Ovarian Aging Study and are adjusted for BMI, current age, menopausal stage, alcohol consumption, and smoking status. In this assessment, hot flashes were reported less frequently in AA carriers of a single copy of CYP 1B1*4 than noncarriers (adjusted OR 0.11, 95% CI 0.02–0.48). Conversely, AA carriers of two copies of COMT Val158Met were greater than 4 times as likely as noncarriers to report moderate or severe hot flashes (adjusted OR 4.20, 95% CI 1.13–15.59).

Table 4.

Association of genotypes and hot flashes, adjusted OR with 95% CI

Gene variant	Variant vs. reference genotype comparison	European-Americans		African-Americans
		HF	Moderate and severe HF	HF	Moderate and severe HF
COMT Val158Met	+/+, +/− vs. −/−	0.71 (0.31–1.60)	1.21 (0.48–3.04)	1.30 (0.59–2.88)	1.44 (0.67–3.09)
	+/− vs. −/−	0.63 (0.23–1.69)	1.17 (0.45–3.09)	0.97 (0.42–2.23)	1.03 (0.45–2.37)
	+/+ vs. −/−	0.75 (0.32–1.79)	1.30 (0.44–3.84)	6.67 (0.80–55.67)	4.20 (1.13–15.59)a
CYP3A4*1B	+/+, +/− vs. −/−	1.33 (0.38–4.70)	0.68 (0.16–2.86)	1.10 (0.39–3.13)	1.09 (0.39–3.04)
	+/− vs. −/−	1.06 (0.28–4.0)	0.42 (0.08–2.21)	1.31 (0.42–4.06)	1.03 (0.39–3.53)
	+/+ vs. −/−	b	b	0.91 (0.29–2.82)	1.17 (0.39–3.53)
CYP1A2*1F	+/+, +/− vs. −/−	0.96 (0.32–2.88)	1.60 (0.41–6.18)	1.00 (0.32–3.10)	1.69 (0.53–5.43)
	+/− vs. −/−	1.47 (0.45–4.85)	1.72 (0.42–7.13)	1.31 (0.39–4.39)	1.91 (0.56–6.58)
	+/+ vs. −/−	0.77 (0.24–2.42)	1.50 (0.37–6.07)	0.77 (0.23–2.56)	1.49 (0.43–5.16)
CYP1B1*3	+/+, +/− vs. −/−	1.18 (0.58–2.42)	0.88 (0.41–1.91)	2.24 (0.49–10.24)	6.57 (0.73–58.82)
	+/− vs. −/−	1.18 (0.56–2.51)	0.82 (0.36–1.86)	1.38 (0.61–13.89)	5.34 (0.80–51.40)
	+/+ vs. −/−	1.18 (0.42–3.30)	1.10 (0.37–3.31)	2.92 (0.61–13.89)	7.25 (0.79–65.89)
CYP1B1*4	+/+, +/− vs. −/−	1.39 (0.66–2.92)	0.77 (0.35–1.79)	0.11 (0.02–0.48)c	0.51 (0.79–65.89)
	+/− vs. −/−	1.38 (0.64–2.94)	0.82 (0.36–1.88)	0.11 (0.02–0.48)	0.51 (0.79–65.89)
	+/+ vs. −/−	1.70 (0.09–33.51)	b	b	b

Odds ratios adjusted for age, BMI, alcohol consumption, and menopausal stage.

^aP = 0.03.

^bEstimate uninterpretable due to insufficient data.

^cP = 0.003.

Table 5 describes the adjusted OR of any hot flashes and moderate or severe hot flashes in the combined presence of smoking and selected SNP in EA women. In this group, double-variant carriers of COMT Val158Met who smoked demonstrated increased odds of any hot flashes (adjusted OR 6.15, 95% CI 1.32–28.78) compared with double-variant carriers who were nonsmokers; moderate or severe hot flashes were also increased in COMT Val158Met double-variant carriers who smoked, an association that reached borderline significance (adjusted OR 4.35, 95% CI 0.95–19.93). EA CYP 1B1*3 double-variant carriers demonstrated a greater than 20-fold increased odds of moderate to severe hot flashes in smokers compared with nonsmokers (adjusted OR 20.6, 95% CI 1.64–257.93). In CYP1B1*4 wild types, smoking was associated with increased odds of hot flashes (adjusted OR 2.63, 95% CI 1.02–6.78), whereas in CYP1B1*4 carriers, the association between smoking and hot flashes was not present (adjusted OR 0.71, 95% CI 0.12–4.32). AA single-variant carriers of CYP 1A2*1F who smoked demonstrated increased odds of reporting hot flashes (adjusted OR 6.16, 95% CI 1.11–33.91) compared with single-variant carriers who did not smoke (Table 6).

Table 5.

Hot flashes in smokers by genotype, European-Americans

	Any HF (n)		Adjusted OR of HF in smokers vs. nonsmokers (95% CI)	Moderate or severe HF (n)		Adjusted OR of moderate and severe HF in smokers vs. nonsmokers (95% CI)
	+	−		+	−
SNP COMT Val158Met
−/−	23	16	1.68 (0.26–10.70)	10	29	0.63 (0.06–6.71)
+/−	40	35	0.72 (0.22–2.35)	22	53	1.68 (0.48–5.81)
+/+	20	20	6.15 (1.32–28.78)	14	26	4.35 (0.95–19.93)
			Pint = 0.1			Pint = 0.3
CYP3A4*1B
−/−	73	67	1.49 (0.65–3.43)	41	99	1.87 (0.78–4.46)
+/−	7	5	2.63 (0.15–44.97)	3	9	2.65 (0.11–66.8)
+/+	1	0	a	1	0	a
			Pint = 0.7			Pint = 0.8
CYP1A2*1F
−/−	9	8	2.79 (0.27–28.98)	4	13	11.1 (0.65–189.29)
+/−	38	22	1.47 (0.42–5.17)	19	41	2.91 (0.84–10.04)
+/+	36	41	1.43 (0.46–4.44)	23	54	1.07 (0.31–3.73)
			Pint = 0.9			Pint = 0.2
CYP1B1*3
−/−	26	24	0.49 (0.11–2.10)	16	34	1.58 (0.37–6.79)
+/−	44	36	1.51 (0.5–4.59)	22	58	1.22 (0.37–4.05)
+/+	13	10	a	8	15	20.6 (1.64–257.93)
			Pint = −			Pint = 0.09
CYP1B1*4
−/−	55	51	2.46 (0.96–6.30)	33	73	2.63 (1.02–6.78)
+/−	26	18	0.52 (0.11–2.46)	13	21	0.71 (0.12–4.32)
+/+	1	1	a	0	2	a
			Pint = 0.09			Pint = 0.2

Adjusted for age, BMI, alcohol consumption, and menopausal stage. P_int is the P value for significance of interaction between smoking and genotype status. P_int− indicates that the P value for interaction could not be calculated. HF, Hot flashes.

^aEstimate uninterpretable due to insufficient data.

Table 6.

Hot flashes in smokers by genotype, African-Americans

	Any hot flashes (n)		Adjusted OR of hot flashes in smokers (95% CI)	Moderate or severe hot flashes (n)		Adjusted OR of moderate or severe hot flashes in smokers (95% CI)
	+	−		+	−
SNP COMT Val158Met
−/−	37	20	0.95 (0.29–3.08)	20	37	1.99 (0.62–6.46)
+/−	40	21	2.89 (0.74–11.25)	24	37	1.78 (0.52–0.61)
+/+	13	2	a	10	5	1.11 (0.1–12.17)
			Pint−			Pint = 0.9
CYP3A4*1B
−/−	14	7	0.64 (0.09–4.34)	8	13	0.76 (0.11–5.06)
+/−	42	17	1.83 (0.48–7.02)	23	36	2.14 (0.62–7.34)
+/+	35	20	3.35 (0.86–13.11)	24	31	2.51 (0.74–8.55)
			Pint = 0.4			Pint = 0.6
CYP1A2*1F
−/−	12	7	1.13 (0.13–9.97)	5	14	0.64 (0.07–6.19)
+/−	49	17	6.16 (1.11–33.91)	30	36	2.67 (0.82–8.70)
+/+	31	20	0.88 (0.25–3.15)	20	31	1.78 (0.5–6.39)
			Pint = 0.2			Pint = 0.5
CYP1B1*3
−/−	4	4	a	1	7	a
+/−	30	19	a	19	30	a
+/+	58	21	a	35	44	a
			Pint−			Pint−
CYP1B1*4
−/−	88	35	1.71 (0.67–4.36)	52	71	1.86 (0.81–4.26)
+/−	4	9	2.35 (0.12–46.80)	3	10	2.6 (0.13–51.81)
+/+	0	2	a	0	2	a
			Pint = 0.9			Pint = 0.9

Adjusted for age, BMI, alcohol consumption, and menopausal stage. P_int is the P value for significance of interaction between smoking and genotype status. P_int− indicates that the P value for interaction could not be calculated.

^aEstimate uninterpretable due to insufficient data.

Further adjusting these models for feelings of anxiety made little impact on the stratified OR observed in the original models except for the adjusted OR of severe hot flashes in EA carriers of two copies of CYP 1B1*3, which changed from 20.6 (95% CI 1.64–257.93) to 14.09 (95% CI 1.11–178.94) (Supplemental Tables 1 and 2).

No significant interactions of current smoking and gene variant status on the odds or severity of hot flashes were found.

Hot flash models were then fit that considered smoking as three categories (never, former, current) and in terms of heaviness of exposure (nonsmoker, current light smoker, current heavy smoker) in combination with gene variants. Using this approach, an association between hot flash severity in current smokers compared with never smokers in EA CYP 1B1*3 double-variant carriers was demonstrated (adjusted OR 20.59, 95% CI 1.39–304.68). In addition, EA COMT Val158Met double-variant carriers who smoked heavily were nearly 14 times more likely than nonsmoking double-variant carriers to experience moderate to severe hot flashes (adjusted OR 13.7, 95% CI 1.2–154.9). The other genotype stratum-specific effects of smoking on hot flashes described in the models with smoking as a dichotomous exposure were not observed in the updated models. No new effects were observed and no significant interactions between smoking behavior and gene variants on hot flashes were found.

Geometric mean estradiol levels (with 95% CI) were compared between smokers and nonsmokers in all study participants and between smokers and nonsmokers in AA and EA participant separately, with no significant differences observed (Supplemental Table 3). The race-specific geometric mean estradiol levels for each stratum of smoking (current or nonsmoker) and gene variant were also evaluated (Supplemental Tables 4 and 5). The estradiol concentrations across variant strata within the categories of smoking and race were not significantly different.

Discussion

In demonstrating a difference in the odds and severity of hot flashes in smokers as a function of genetic variability in sex steroid metabolism, our data support the presence of genetic susceptibility to environmental toxins in this facet of reproductive aging. Strong evidence exists of a relationship between smoking and increased frequency of hot flashes (2–4, 6, 37). It has been postulated that this relationship is due in part to the antiestrogenic impact of tobacco smoke components that induce enzymes responsible for steroid hormone metabolism (9–11). To our knowledge, genetic susceptibility of reproductively aging women to smoke-induced hot flashes has not been previously reported.

When focusing on the relationship of genotypes and symptoms alone, an association between CYP 1B1*4 carrier status and reduced hot flash occurrence in AA women was observed. Conversely, AA double-variant carriers of COMT Val158Met were significantly more likely than noncarriers to report moderate or severe hot flashes. These data have not been previously reported and expand on the literature addressing the association between steroid-metabolizing genotypes and menopausal symptoms.

In carriers of COMT Val158Met, CYP 1B1*3, and CYP 1A2*1F, significant race-specific associations between smoking and hot flash occurrence and/or severity were demonstrated.

COMT, CYP 1B1, and CYP 1A2 function at unique steps along the estrogen metabolic pathway (Supplemental Fig. 1). CYP 1A2 is primarily involved in 2-hydroxylation and 4-hydroxlyation of estrogens to the corresponding catecholestrogens (10, 12). CYP 1B1 hydroxylates estrogens to 4-hydroxyestradiol catecholestrogens (10, 12). COMT metabolizes catecholestrogens through the very efficient addition of methyl groups to these intermediate metabolites, converting them to methoxyestrogens (10, 38). The enzyme exists in membrane-bound and soluble forms that, although encoded by the same gene, differ in size by 50 amino acids (membrane bound COMT contains 50 more amino acids than the 221 present in soluble COMT). Substitution of Val118 by Met108 in soluble COMT due to a transition of guanine to adenine at codon 158 produces a protein that has 3- to 4-fold lower enzyme activity (38).

Increased hot flashes in smokers who are CYP 1B1*3 carriers could arise from an amplification of initial phase metabolism of estrogens (39). CYP 1B1*3 carriers have increased enzyme activity to hydrolyze estrogens; catalytic activity of CYP 1B1*3 is also more inducible by smoking than is seen in the nonvariant allele (39). Estrogens should be inactivated more rapidly in the presence of this variant, especially in smokers, resulting in greater odds of hot flashes.

CYP 1A2 activity is inducible by tobacco smoke and published reports have demonstrated that CYP 1A2*1F carriers who smoke have increased enzyme activity and inducibility (40, 41). Increased enzyme activity in this abundant hepatic source of estrogen metabolism could hasten hormone degradation, increasing the likelihood of hot flashes.

The mechanism for increased hot flashes in carriers of COMT Val158Met in combination with smoking is likely multifactorial. In carriers of COMT Val158Met, metabolism of estrogens is expected to be less efficient; in carriers who smoke, catecholestrogen accumulation would be anticipated due to induction of phase I CYP enzymes that metabolize estrogens to catecholestrogens, magnified by reduced degradation of the catecholestrogen load due to diminished COMT activity (38, 42, 43).

Despite less efficient metabolism of estrogens and catecholestrogens expected in COMT Val158Met carriers, we found that double-variant carrier status was a risk factor for hot flashes in two groups: AA women independent of smoking and other confounders and EA women in combination with current smoking behavior. When the capacity for O-methylation is reduced or inhibited by an excess catecholestrogen load, the half-life of catecholestrogens increases (43). In general, catecholestrogens are capable of binding the estrogen receptor but with less affinity and potency than estradiol. 4-Hydroxyestradiol, however, appears to have a lower rate of receptor dissociation than does estradiol (44, 45). Thus, it is possible that up-regulated catecholestrogen production in smokers further accumulate due to a relative block in further metabolism in COMT Val158Met carriers; these weak estrogens could outcompete low-level estradiol concentrations for occupancy of estrogen receptors and remain bound to the receptor, enhancing a hypoestrogenic state and increasing vasomotor symptoms. In addition, COMT Val158Met carriers, especially those who smoke, may have increased variability in estradiol levels, and this may also enhance the odds of having hot flashes. As has been proposed elsewhere (26), these findings support a model of hot flash risk that expands on the view that diminished estrogen levels per se are associated with increased symptoms and suggests that the type of estrogen and the variation in estrogens and their metabolites play critical roles in symptom generation.

A second pathway explaining increased hot flashes in smokers who are COMT Val158Met carriers may involve alterations in neurotransmitter metabolism within the dopaminergic system. COMT is the primary enzyme involved in the inactivation of dopamine (46). Nicotine stimulates release of dopamine from neurons in the ventral tegmental area of the brain, an action felt to be critical for its rewarding and reinforcing effects (20, 47, 48). Dopamine has long been considered a candidate neurotransmitter for the development of hot flashes. Support for dopamine's role in hot flashes derives from several unique lines of evidence. Dopamine agonist treatment of prolactinomas in asymptomatic postmenopausal women has been associated with the onset of hot flashes (49), suggesting that either the addition of dopamine promotes symptoms or that the presence of hyperprolactinemia suppresses symptoms. Multiple reports, including randomized controlled trials have described a role for the drug veralipride, a dopamine antagonist, in the treatment of hot flashes (50, 51). The evidence demonstrating a role for elevated dopamine levels in the development of hot flashes lends biological credence to our findings that COMT Val158Met carriers who smoke have increased frequency and severity of these symptoms. Elevated synaptic levels of dopamine that occur in all smokers should be further augmented in women who are carriers of COMT Val158Met, and this could promote the development of hot flashes.

Among the strengths of this investigation include the use of a racially balanced study population permitting race-specific analyses unbiased by differential distribution of gene variant frequencies in AA and EA women. The five variants studied were selected based on their roles in metabolism of sex steroid hormones or neurotransmitters with plausible roles in hot flash generation. Although many genotypes involved in sex steroid metabolism could be investigated, evaluating the role of five variants placed an appropriate limit on the number of hypotheses tested.

We believe that the odds OR for hot flashes observed represent results for the combination of smoking and genotypes independent of other factors that have a known relationship with symptoms. A rational approach was used to derive final models that were adjusted for age, menopausal stage, alcohol consumption, and BMI. BMI, for instance, was an important confounder of our results and appeared to exert the most impact in the models in EA women. Additional factors that have been associated with hot flashes in separate reports such as parity, hypertension/hypertensive medications, FSH levels, and polycystic ovary syndrome history either were not relevant to the cohort (polycystic ovary syndrome diagnosis) or did not confound the effects observed (data not shown) and were excluded from the final models. As pertains to reproductive hormones (estradiol and FSH), there was a concern that that adding these hormones to the models would result in a biased estimation of the direct effect that genotypes had on our outcomes due to the role of the polymorphisms in estrogen metabolism. The impact of adding anxiety to the models was minimal except for the effect on the odds of severe hot flashes in EA smokers who were carriers of two copies of CYP 1B1*3. However, although the anxiety-adjusted OR was attenuated toward the null, it remained statistically significant (adjusted OR 14.09, 95% CI 1.11–178.94).

Despite these strengths of the work, some limitations exist. The nature of our study does not permit a characterization of smoking behavior as concerns subtle variations in heaviness of smoking or periods of cessation across the menopausal transition. It is likely that a complex relationship exists between menopausal stage, mood, hot flashes, and patterns of smoking. In addition, genetic variation in the metabolism of hormones and neurotransmitters that mediate the rewarding effects of tobacco could influence smoking behaviors across the menopausal transition. COMT Val158Met carrier status, for instance, has been linked to successful smoking cessation in EA women (21). Nevertheless, a relationship between COMT Val158Met carrier status and smoking behavior was not found in our study participants (data not shown).

The multiple comparisons made in this analysis introduce the possibility that false-positive associations were detected. We have reviewed the approach of Wacholder et al. (52) to determining false-positive reporting probabilities in molecular epidemiology studies and believe that, based on key criteria (magnitude of associations, functional relevance of SNP studied, and prior probability of associations), that our significant findings are noteworthy.

Conversely, although several genotype stratum-specific effects of smoking on hot flashes were demonstrated, these associations were not uniformly observed across racial groups. It is possible that uniform associations were not found when the frequency of a genotype was low in a particular race, limiting the power to detect possible significant associations. In addition, we performed race-specific comparisons of hot flashes between smokers and nonsmokers within three strata of selected gene variants. Carrier frequencies varied across groups and some of our strata had numbers that were too sparse to reliably find associations in both racial groups studied. As for statistical power to examine whether the association between smoking and hot flashes is modified by genetics, we conducted post hoc power calculations using Power V3, available from the National Cancer Institute (http://dceg.cancer.gov/tools/design/power). Assumptions included hot flash prevalence of 40–60%, smoking prevalence of 38%, type I error of 5%, and sample size of 158 (parameters consistent with the AA subgroup). For a SNP with a high-risk allele frequency of 25%, this study had 41% power to detect an interaction OR of 5 or greater. The size of our study population may have limited both our ability to detect more associations than were described and to identify the presence of formal interactions between variants and smoking if they truly exist.

Differences between estradiol concentrations in smokers and nonsmokers were not observed. In addition, estradiol concentrations across variant strata within categories of smoking and race were not significantly different. These results are consistent with other reports that have either observed null effects of smoking or steroid hormone-metabolizing gene variants on estrogen concentrations in the menopause and menopausal transition (3, 27, 28, 53). Our results also lend support to the premise that the relationship between hot flashes and smoking behaviors may involve pathways in which nonsteroidal hormones such as dopamine play an important role.

In summary, we present novel data supporting the presence of joint effects of smoking and specific gene variants involved in sex steroid metabolism on the occurrence and severity of hot flashes in reproductively aging women. In presenting genetic characteristics of smokers at greatest risk for hot flashes, studies such as ours may help providers better allocate resources for smoking cessation in menopausal women. Interpreting the impact of pervasive exposures on menopausal symptoms within an appropriate genetic context could enhance individualized menopausal care and suggest approaches to stratifying smokers into risk categories for additional menopause-related symptoms and morbidities. The associations described herein may also apply to the risk of reproductive dysfunction and infertility in younger age groups exposed to environmental toxins with similar properties as tobacco smoke. Future investigations that expand on these results in additional populations could contribute to innovative approaches that address reproductive morbidity across the female life course.