The Norwegian Women and Cancer Study (NOWAC) is a cohort study based on questionnaires mailed to women who are 30–70 years old 19. Participants are randomly drawn from the Central Population Register. From 1991 up until June 2007, 171 977 women had been enrolled in NOWAC, of whom 167 058 are still participating. Questionnaire information on diet, lifestyle and the use of medication is available, with 1–2 repeat measurements at 4–6 year intervals. The NOWAC postgenome cohort consists of 49 233 participants born between 1943 and 1957, who contributed a blood sample between 2003 and 2006. Written informed consent is obtained from each participant and the collection and storing of questionnaire information and blood samples is approved by The Regional Committee for Medical Research Ethics and the Norwegian Data Inspectorate. Statistics Norway obtains updated information on deaths and migration and performs the sampling of women, thereby providing a complete follow-up of participants.

The present study is a cross-sectional analysis within the NOWAC postgenome cohort (Figure 1). Of the 20 391 women who answered an eight-page questionnaire in the autumn of 2004 (response rate 81%), 17 932 agreed to donate a blood sample. Women, randomly drawn in groups of 500, were mailed a blood collection kit and an accompanying two-page questionnaire in April 2005. One reminder was mailed after three weeks to non-responders. The overall response rate was 74%. The two-page questionnaire included questions regarding menopausal status, smoking, weight, height, use of HRT, oral contraceptives or other medication, omega-3 intake, intake of soy or other dietary supplements, and details concerning blood specimen collection (date, hour, posture). Our present study included 445 responders from one group of 500 women (89%); 3.2% declined to participate, 0.8% had died or migrated and 7.0% did not respond. Six samples were excluded due to incompletely filled blood collection tubes. Additionally, 14 women were excluded due to a lack of information concerning menopausal status, use of HRT or type of HRT used. This left plasma sample measurements for 425 women.

The blood samples were drawn at the women's local general physician's offices, using the blood collection kit. For collection of plasma and buffy coat we used a Vacuette^® Coagulation Tube (Greiner Bio-One GmbH, Kremsmünster, Austria) containing citrate buffer 0.109 mol/L (3.2%); 1 part citrate to 9 parts blood. For collection of RNA we used a PAXgene™ Blood RNA tube (PreAnalytiX GmbH, Hombrechticon, Switzerland), which is a BD Vacutainer™ containing a proprietary reagent that immediately stabilizes intracellular RNA. The samples were mailed overnight to the Institute of Community Medicine at the University of Tromsø, Norway. The women were requested not to have their blood samples drawn on Thursdays and Fridays, in order to avoid a weekend mail delay. The blood samples were generally received by the NOWAC biobank staff within 1–2 days (92%). Upon arrival, the Vacuette^® Coagulation Tubes were centrifuged at 3000 rpm for 15 minutes. Plasma (2 × 1.8 mL) and buffy coat (1.0 mL) were frozen at -20°C and subsequently transferred to -70°C within one week. PAXgene™ Blood RNA tubes were frozen directly at -20°C and transferred to -70°C without pre-processing. The results of gene expression analyses will be published at a later stage.

All the hormone analyses were performed at the Department of Medical Biochemistry, University Hospital of North Norway, Tromsø, Norway. Plasma levels of estradiol (E₂), progesterone (P₄) and Follicle Stimulating Hormone (FSH) were measured by immunometry, using an electrochemiluminescence immunoassay (ECLIA) on Modular Analytics E170 (Roche Diagnostics GmbH, Mannheim, Germany). Plasma levels of Sex Hormone Binding Globulin (SHBG) were measured by chemiluminescent immunometric assay (CLIA) on Immulite^® 2000 (Diagnostic Products Corporation, Los Angeles, CA, USA). The respective detection limits and analytic coefficients of variation (CV) were 0.018 nmol/L and 5.2% for E₂; 0.100 IU/L and 2.3% for FSH; 0.095 nmol/L and 6.9% for P₄; and 0.02 nmol/L and 5.0% for SHBG. For the sake of convenience, SHBG will be referred to as a hormone throughout this paper. According to the laboratory, the postmenopausal reference values for FSH and E₂ were FSH > 26 IU/L and E₂ < 0.20 nmol/L. Three measurements of P₄ were below the detection limit and values were defined as half of the detection limit (0.048 nmol/l). Two measurements of SHBG were above the calibration range, and values were defined as the upper limit of the range (180 nmol/L).

The analysis of Modular Analytics E170 of SHBG is not validated for citrate plasma by the manufacturer. The Department of Medical Biochemistry performed a small verification analysis, using serum and citrate plasma from 21 healthy volunteers (data not published). The results indicated a good correlation between measurements in serum and citrate plasma for all hormones measured (0.9899 ≤ r² ≤ 0.9997).

We used SPSS^® 14.0 for Windows for the statistical analyses. Geometric mean plasma levels across different categories of HRT use or BMI were compared using univariate analysis of covariance (ANCOVA) through the general linear model approach. Additionally, we used multiple linear regression to test the association between BMI (continuous variable) and hormone levels. Covariates tested for potential confounding: age, alcohol consumption (units per week), parity and BMI (ANCOVA across HRT categories). In the analysis of the association between BMI and hormone level among HRT users, we also included HRT category as a potential confounder. With the exception of age, only covariates that contributed significantly to the model were included in the final analysis. Time since menopause was excluded as a covariate, due to 25% missing values among postmenopausal women. Sidak corrected post hoc comparisons were used to determine which group means differed. Levene's homogeneity-of-variance test was used to check the equality of group variances. The association between natural log-transformed hormone levels and time since menopause were tested by partial correlation, controlling for BMI. Difference in hormone levels according to time since last HRT dose (0 or 1 day) or fasting (≥10 hours since last meal 10), and difference in SHBG level between the use of oral and other HRT regimens were analysed with Student's t-test for independent samples. We used the McNemar's test for correlated proportions to check for differences in sensitivity and specificity between the two questionnaires 20. All p-values are two-tailed and the level of statistical significance is 5%.

BMI was categorized as underweight (<18.5 kg/m²), overweight (≥25.0 kg/m²) and obesity (≥30.0 kg/m²) 21. The two lowest categories (underweight and normal weight, <25.0 kg/m²) were merged, due to there being few underweight women.

Menopausal status at blood draw was determined for each woman, based on her answers in the two-page questionnaire as to whether she still had regular menstrual periods, whether the periods were irregular or whether they had stopped. Women were classified as postmenopausal if their periods had stopped and premenopausal if their periods were regular. Women with irregular menses were classified as postmenopausal if they were 53 years or older. This cut-off point was used in a previous NOWAC report 14, based on the definition used in the MWS 15. The eight-page questionnaire additionally included questions regarding the reason why periods had stopped (natural stop, bilateral oophorectomy, hysterectomy or other reasons) and the age when periods had stopped. When classifying according to plasma levels, we used the postmenopausal reference values,

both

and

either

or

HRT use was categorized according to E₂ content: no HRT, HRT without E₂ (i.e. estriol, tibolone and other progestogens), E₂ for vaginal application, E₂ patches (all dosages), oral 1 mg E₂ (continuous and sequential preparations), and oral 2 mg E₂ (continuous, but also sequential preparations with 1 mg E₂ in 6 out of 28 tablets).

To assess the validity of HRT use, we compared plasma E₂ levels among HRT users with the 95% CI for plasma E₂ levels among postmenopausal non-users, and we examined to what extent plasma E₂ levels among non-users exceeded 0.20 nmol/L.

Of 331 postmenopausal women, 66 were current HRT users. Among oral preparations, 84% were combinations of E₂ and norethisterone actetate (NETA), while 16% were E₂-only preparations. Most patches were E₂-only preparations, except one woman using a transdermal combination of E₂ and NETA. Three of the 66 users showed signs of non-compliance, based on their reported date of last HRT dose. Six women did not report the date.

Table 2 shows the distribution of women and geometric mean plasma levels of hormones for each HRT user category. The plasma levels of E₂ increased with increasing E₂ dose (Table 2, Figure 3). Moreover, use of systemically-administered HRT (patches and tablets) containing E₂ suppressed the level of FSH (Table 2, Figure 3). There were statistically significant differences in P₄ and SHBG levels across categories of HRT use (Table 2), although not in a dose-dependent manner in the case of P₄. The assumption of homogeneity of variance was violated in the analysis of E2 and FSH (Levene's test p < 0.01). The ratio between highest and lowest variance was 7.6 for E2 and 8.7 for FSH.

The post hoc comparisons showed that there was no difference in plasma E₂ level between vaginal E₂ application and no E₂ use (i.e. no HRT use and HRT without E₂). The main difference was between no/vaginal E₂ use and systemically administered HRT (p ≤ 0.01). The same tendency was seen for FSH levels, although not as conclusive as for E₂ levels. The post hoc comparison did not reveal any systematic pattern of differences for P₄ and SHBG levels across HRT categories. There was, however, a statistically significant difference (p = 0.02) in SHBG level between use of oral HRT (54.5 nmol/L, 95% CI: 45.8–64.9) and the other HRT users (39.8 nmol/L, 95% CI: 32.6–48.6). There was a borderline significant difference in E₂ level (p = 0.05) between those who had taken their last tablet on the day of blood sampling (n = 17) and those who had taken their last tablet the day before (n = 11).

There was no significant difference in BMI between premenopausal and postmenopausal women (Table 1). Among postmenopausal women there was no significant difference in BMI between HRT users (25.1 kg/m², 95% CI: 24.0–26.2) and non-users (25.7 kg/m², 95% CI: 25.2–26.2) (p = 0.30).

Among non-users, there was a significant difference in FSH and SHBG levels across the three categories of BMI, but not in E₂ or P₄ levels (Table 3). The regression coefficients (β) showed that one unit increase in BMI was significantly associated with a -3.5% decreased FSH level (95% CI: -4.5% – -2.4%), a -5.4% decreased SHBG level (95% CI: -6.6% – -4.3%), and a 1.6% increased E₂ level (95% CI: 0.5% – 2.7%), after adjusting for age. Among HRT users (data not shown), there was significant association between BMI and FSH (p < 0.01), but not between BMI and the other hormones analysed (adjusted for age and estrogen dosage category).

There was negative correlation between time since menopause and plasma E₂ levels (r = -0.16, p = 0.03) and P₄ (r = -0.29, p < 0.01), and positive correlation between SHBG levels and time since menopause (r = 0.15 and p = 0.04). FSH levels were not significantly correlated with time since menopause.

There was no statistically significant difference in hormone levels between the samples received within 24 hours and those transported over 2, 3 or 4 or more days among either pre-/perimenopausal or postmenopausal women (data not shown). Similarly, there was no statistically significant difference in plasma level between fasting and non-fasting subjects for any of the hormones measured, and lipaemia (n = 13) and haemolysis (n = 30), encountered by visual examination, did not influence the hormone levels (data not shown).

Figure 4 shows plasma levels of E₂, P₄ and FSH, according to days since menstruation among premenopausal women who reported having regular periods and who filled in the date of the first day of their most recent menstruation (n = 62). Although there were few women in each two-day period (n = 1–8), the pattern of hormonal variation throughout the menstrual cycle was recognizable both for the gonadal hormones and FSH. Progesterone levels > 20 nmol/L were only found among women in their luteal phase (≥15. day). There was no recognizable cyclic hormone pattern among the 20 perimenopausal women who had reported their menstruation date. One woman used an oral contraceptive, a progestagen-only pill.

Sensitivity and specificity for the variable "menopausal status" defined by the two questionnaires used is shown in Table 4. The 66 HRT users were excluded from this analysis. Sensitivity was higher in the two-page questionnaire accompanying the blood sample (p < 0.05). The eight-page questionnaire scored higher on specificity (p < 0.01).

The NOWAC study has the advantage of being population-based and prospective. The opportunity of random sampling from the complete Central Population Register, together with high response rates, provides a representative sample of the Norwegian female population aged 48–62 years 43. The two-page questionnaire accompanying the blood sample provides detailed and updated information on central variables such as menstruation status, weight and HRT use, which are complementary to the preceding eight-page questionnaire. Results show that it is especially valuable to be able to differentiate between different types of HRT. The effect of HRT exposure will vary between populations using different HRT regimens, particularly with respect to estrogenic carcinogenicity.

Compliance with HRT is an important determinant for the accurate measurement of HRT use. With 86–95% of the women showing compliance with HRT use, we do not consider lack of compliance to be a significant source of bias in our study.

Probably due to small sample sizes, the assumption of equal variances was violated in the ANCOVA analysis of differences between HRT categories. This causes some concern and the results must be interpreted accordingly. Even so, the very low p-values suggest that the results for E2 and FSH are valid.

Reports on the validity of self-reported body-size generally show that particularly obese people tend to underestimate their weight/waist circumference 4445. In the present study, there is an increase in weight (mean: +0.6 kg) between the eight-page questionnaire and the two-page questionnaire six months later and the increase is higher among the women who had their weight measured at their general physician's office (n = 49). This should not influence our analyses to the degree that the associations found are artefacts. However, if the BMI is under-estimated, the change in hormone levels per increased unit of BMI could be over-estimated.

Choosing plasma or serum is a trade-off between rational collection logistics and the broadest assortment of feasible analyses. Not all general physicians' offices have equipment for blood centrifugation, and at the beginning of the NOWAC blood specimen collection in 2002 the range of future analyses had not yet been determined. With the aim of collecting blood from as many participants as possible, it was decided to build a plasma biobank. Citrate was chosen as the anticoagulant, due to the collaboration of NOWAC with EPIC and their collection of citrate plasma. Citrate plasma is not the optimum matrix for immunometry analyses. Although the verification analysis performed by the laboratory did not show alarming inconsistencies, the results must be interpreted accordingly, especially in view of the general limitations of direct immunoassays mentioned above.

A transport delay of over two days could potentially interfere with our measurements. Several studies have concluded that sex hormones FSH and SHBG are fairly stable with regard to transport conditions, temperature variations and delayed processing 4647484950. These reports are not based on an analysis of citrate plasma. However, in our sample there were no statistically significant differences due to transport delay, and we did not exclude any blood samples on these grounds. The analytical methods used are fairly robust with regard to interference by haemoglobin and triglycerides, and since we found no differences due to lipaemia or haemolysis these samples were not excluded from the analysis. Due to the study design, blood samples were not drawn at the same time of day, nor were the women requested to be fasting. However, we found no differences in hormone levels due to fasting and have not adjusted our results according to time since last meal.

Because of the cross-sectional nature of this study, we cannot draw any conclusions regarding causal association based on our results. However, imminent gene-expression analyses and future follow-up of the women in our population sample will contribute to our knowledge on causal relationships.