Factors Associated with IgA and IgG Fluctuations at the Cervix during the Menstrual Cycle

Abstract

Objective

The objective of this analysis was to describe patterns and determinants of cervical immunoglobulins A (IgA) and G (IgG) during the menstrual cycle.

Methods

One hundred and fifty four women who attended three visits coinciding with follicular, peri-ovulatory, and luteal phases of menstrual cycle were studied. Cervical secretions were collected at each visit for total IgA and IgG determination. Questionnaires administered at each visit inquired about demographics and behavioral practices.

Results

Total IgA and IgG levels were higher among oral contraceptive (OC) users than naturally-cycling women. IgA and IgG levels declined at mid-cycle particularly among OC-users. After adjusting for time in cycle, specimen weight, and blood in the sample, reporting a recent illness was associated with lower IgA and IgG, and increasing age was associated with higher IgA and IgG among OC-users and non-OC users. Increased pregnancies were associated with higher IgA among non-OC users, and higher IgG among OC-users. Change in immunoglobulin levels between visits was associated with sample weight and presence of blood among both OC-users and non-users.

Conclusions

Time in cycle and OCs were significant determinants of cervical IgA and IgG levels. Role of endogenous and exogenous hormones on cervical immunoglobulins levels should be further investigated.

Introduction

Recent trials of prophylactic VLP-based HPV vaccines have shown high efficacy in preventing infections with the HPV types in the vaccine formulation (HPV types 16 and 18 that together cause ~70% of cervical cancers worldwide) (1;2). The prophylactic protection conferred by these vaccines is likely mediated through HPV-specific neutralizing antibodies at the cervix (3). While vaccination induces high levels of circulating anti-HPV neutralizing antibodies systemically (4), protection against HPV has to occur locally at the cervix. Systemic antibodies can transudate into the cervix, with variable levels of HPV16 VLP antibodies at the cervix following vaccination (0.5% to 50% of systemic levels)(5).

It is not known whether and how local immunity at the cervix affects long-term duration of HPV vaccine efficacy as antibody levels decline. Given their potential impact on vaccine efficacy, it is important to investigate fluctuations in immunoglobulins as markers of cervical immunity during the menstrual cycle. An earlier study conducted among a group of HPV16 VLP vaccinated women showed immunoglobulin A (IgA) and G (IgG) titers being highest at follicular phase, decreasing around ovulation, and increasing at luteal phase(6). We conducted this study to determine antibody patterns among non-vaccinated women, and investigate inter- and intra-women fluctuations in total IgA and IgG levels at the cervix during single menstrual cycle, and the role of demographic and behavioral factors on immunoglobulin patterns among naturally cycling and OC using women.

Materials & Methods

Study population

Data to address the aims of this study are from the NCI sponsored Proyecto Epidemiológico Guanacaste (PEG) study, described in detail elsewhere (7;8). Briefly, the primary aim of this population-based cohort was to study the natural history of HPV infection and cervical intraepithelial neoplasia. Between 1993–1994, 10,049 women from the Guanacaste province in Costa Rica enrolled in the study. Participation among eligible women was over 93%. At each visit, participants were interviewed to obtain information on demographic, behavioral, and sexual practices. All women signed an informed consent at enrollment. The study was approved by the IRBs of NCI and Costa Rica.

A subset of women who were 25–35 years old, with intact uterus, not pregnant and without evidence of cervical high grade disease were contacted to participate in this study.

Initial Contact

Women were visited at home where we explained the purpose and procedures of this sub-study. An appointment was made for clinic visits for women with regular menstrual cycles (cycle length of 25–35 days) and willing to participate in the sub-study.

Three clinic visits were scheduled to capture women at follicular, peri-ovulatory and luteal phase during a single month. At the first clinic visit, further information regarding the study procedures was provided, and informed consent specific to this study obtained. Non-OC users targeted their first visit (follicular phase) days 5–8 in cycle, second visit (peri-ovulatory) days 14–16 in cycle, and third visit (luteal phase) days 19–22 in cycle. Similar visits were scheduled for women using OCs, and for consistency, we will refer to these visits as follicular, peri-ovulatory, and luteal phase.

Data and Specimen Collection

At each visit a questionnaire was administered ascertaining information on sexual behavior, reproductive history, smoking, douching, recent illness, medication, and vitamin use. Additional interview data collected as part of the main cohort study were also available. For non-OC users, ovulation was confirmed at the second clinic visit using a urinary luteinizing hormone (LH) test kit (Ovuquick® Conception Technologies, San Diego, CA).

At each visit a pelvic exam was performed during which cervical secretions were collected using cellulose-based Ultra-cell sponges (Medical Technologies, Inc., North Stonington, CT). Sponges were gently placed in the cervical os for 30 seconds to passively absorb cervical secretions. Sponges were placed in cryovials, frozen immediately and shipped to the U.S. in liquid nitrogen where they were stored until extraction. Prior to extraction, the sponges were thawed on ice and weighed. The extraction process was previously described (9). Briefly, 300μl of extraction buffer (PBS (Invitrogen, Grand Island, NY), 256mM NaCl (Sigma-Aldrich, St. Louis, MO), 100μg/ml Aprotinin (Sigma-Aldrich, St. 5 Louis, MO) was slowly added to the top of the sponge. Following 30 minute incubation at 4°C, sponges were centrifuged at 13,000g for 15 minutes at 4°C. An additional 300μl of extraction buffer was added to the sponge and immediately centrifuged. Then, 5μl of extract was tested for blood contamination using Hemastix® (Bayer Corporation, Elkhart, IN) prior to the addition of 4μl of fetal calf serum. The extracts were aliquoted and frozen at −70°C until ELISA testing. At the last visit, cervical cells were also collected into PreservCyt solution (Cytyc Corporation, Marlborough, MA) to prepare liquid cytology and determine HPV, Chlamydia and gonorrhea status.

Determination of IG Levels

Cervical secretion extracts were diluted at 1:1000 and tested for total IgA and IgG levels using an enzyme-linked immunoassay (Bethyl Corporation, Montgomery, TX) in duplicate, per manufacturer’s instruction (10).

Statistical Methods

IgA and IgG levels were standardized to account for differences in volume of secretion collected. 5–10 unused sponges randomly picked from each lot were weighed; mean weight of these sponges was assumed to be the dry weight of all sponges from that lot. We standardized each IgA and IgG level using the following formula:

$$\begin{array}{c}(\text{Specimen}\text{weight}-\text{Average}\text{dry}\text{sponge}\text{weight}+0.6(\mathrm{i}.\mathrm{e}.\text{weight}\text{of}600\text{uL}\text{extraction}\text{buffer}))/(\text{Specimen}\\ \text{weight}-\text{Average}\text{dry}\text{sponge}\text{weight})\end{array}$$

Standardized IgA and IgG levels were log-transformed to achieve a normal distribution. Geometric mean concentrations are presented.

Factors associated with IgA and IgG between women

An objective of this analysis was to investigate the determinants of IgA and IgG levels. We used linear regression models to obtain mean immunoglobulin levels among categories of a given covariate. Because there were three visits per woman, and values for each woman were auto-correlated, we then applied multivariate models for correlated data to determine factors independently associated with IgA and IgG levels. Generalized estimating equations (GEE) models with an unstructured correlation structure were used to estimate standard errors and coefficients adjusted for multiple observations for each woman (11). Model coefficients were exponentiated and interpreted as the percent change in immunoglobulin level for a given level of an exposure variable relative to reference category. IgA and IgG results were adjusted for sample weight when standardized, however, analyses suggested residual effects of weight on the immunoglobulin measures; hence we additionally adjusted for sample weight in our models.

Factors associated with magnitude of change in IgA and IgG within women

A second objective of this analysis was to evaluate determinants of change in IgA and IgG levels between the follicular, peri-ovulatory and luteal phase. We assessed change in relation to immunoglobulin levels at the peri-ovulatory visit, where levels were lowest for both naturally cycling women and OC users. Our rationale for choosing the lowest value as the referent was to facilitate the interpretation of the models. Thus the outcome variable was constructed as the difference between log transformed immunoglobulin levels at peri-ovulatory and follicular, and similarly at peri-ovulatory and luteal visits. We used linear regression to examine the effect of covariates on the magnitude of change in immunoglobulin levels between visits. Again, coefficients were exponentiated and can be interpreted as percent change.

All analyses were performed using STATA 9.2 and stratified by OC use. Covariates with statistical significance of p≤0.10 in univariate models were entered in the multivariate models. The final models were developed by including all covariates that were statistically significant in any of the multivariate models.

Results

Population characteristics

Two hundred and two women were selected, of whom 199 women met the eligibility criteria and 196 completed all 3 clinic visits. Of those who met the eligibility criteria, we excluded 17 who were breast-feeding at study entry and, 28 whose menstrual cycle was either too short (<25 days; n=9) or too long (>36 days; n=15), or missing (n=4) leaving 154 women in the analytic sample; 69 OC-users and 85 non-OC users (Figure 1). An additional 38 records were excluded because of low specimen weight, an indication that not enough secretion was collected. Women excluded were similar to those in the analytic sample with respect to age, current smoking status, lifetime pregnancies, and HPV status (data not shown). For the 154 women included in the analysis, the median time since LMP for the follicular, peri-ovulatory and luteal visits were 7 days (range 5–8), 14 days (range 13–17), and 19 days (range 19–22), respectively.

Figure 1.

Consort Diagram of the Study Population:

OC users were slightly younger than naturally cycling women, median age 31 compared to 32, (p=0.008). As expected, women using OCs differed from non-OC users in choice of birth controls, with non-OC users more likely to use condoms, IUDs, and having had a tubal ligation compared to OC-users. No other differences were observed between OC-users and non-users with respect to the other covariates evaluated (data not shown).

Patterns of IgA and IgG across menstrual cycle

Figure 2 shows the log-transformed IgG and IgA measurements by visit for 69 OC-user and 85 normally cycling women. Among naturally cycling women, both IgA and IgG were higher in the follicular phase, with a sharp drop around ovulation, followed by increasing levels in the luteal phase. OC users had more stable levels over time; however, a decline in levels in the peri-ovulatory period was still evident. Median titers were higher among OC users compared to non-OC users for both IgA (0.38 and 0.23, respectively, p<0.001) and IgG (1.03 and 0.44, respectively, p<0.001). IgA and IgG levels were highly correlated (Figure 3, Spearman rho=0.76, p<0.001).

Figure 2.

Distribution of IgA and IgG levels at different phases of the menstrual cycle among OC users and normally cycling participants

*p-value adjusted for specimen weight.

Figure 3.

Scatterplot of log-transformed IgA and IgG levels in cervical secretions:

Although we targeted the second clinic visit to coincide with expected ovulation, we were successful for only 13 of 85 women based on testing for the LH surge. Ovulation typically occurs within 36 hours of an LH surge. The pattern observed for the LH positive women (Figure 4) was similar to that observed overall, although the magnitude of the decline at mid-cycle was slightly more pronounced for those positive for LH surge. However, even among these women, there was wide variability across women with some having higher levels at the start of the cycle with a more pronounced decline around ovulation, while other women had a less pronounced change.

Figure 4.

Patterns of IgA and IgG levels at different phases of the menstrual cycles among naturally cycling women (n = 13) whose LH surge occurred on the day of their peri-ovulatory visit

Factors associated with IgA and IgG between women

Univariate and multivariate effects of covariates on IgA and IgG during the menstrual cycle are presented in table 1. In the final multivariate model, after adjusting for time in cycle and specimen weight; increasing age, blood in the sample and ectopy were associated with higher IgA and IgG. Reporting a recent illness was significantly associated with lower IgA and IgG. Higher immunoglobulin levels were associated with more pregnancies, although significant only for IgA. IgG levels were significantly higher among women reporting current IUD use. While we observed higher IgA and IgG levels among carcinogenic HPV-positive women in univariate models, the association was no longer significant in multivariate models. Higher IgA and IgG levels were not associated with current smoking, years since menarche, vaginal douching, vaginal discharge, talcum powder use, recent medication, Chlamydia trachomatis, Gonorrhea, or vitamin use (data not shown). Similar patterns were observed when we examined IgA and IgG patterns by covariates among all visits for the 13 naturally cycling women with a documented LH surge (data not shown).

Table 1.

Crude and adjusted Determinants of IgA and IgG levels between women: naturally cycling women

		IgA		IgG
	N*	Percent Change (95%CI)	Adjusted Percent Change† (95%CI)	Percent Change (95%CI)	Adjusted Percent Change† (95%CI)
Age group (years)		1.06 (0.99–1.13) §	1.07 (1.01–1.13) §	1.04 (0.96–1.13) §	1.07 (1.00–1.15) §
Lifetime Pregnancy
0/1	64	1	1	1	1
2/3	122	1.26 (0.86–1.84)	1.34 (0.96–1.87)	0.96 (0.60–1.56)	0.81 (0.55–1.20)
4–6	52	2.01 (1.27–3.19)	1.81 (1.22–2.68)	1.23 (0.69–2.20)	1.00 (0.63–1.58)
Years since 1st Pregnancy
0–11	72	1		1
12+	130	1.63 (1.18–2.26)		1.23 (0.82–1.86)
Days since last sex
<= 2 days	81	1		1
3–10 days	86	1.48 (1.03–2.12)		1.74 (1.15–2.65)
11+ days	69	1.34 (0.90–1.99)		1.75 (1.09–2.80)
IUD Use
No	202	1	1	1	1
Yes	36	1.21 (0.75–1.95)	1.17 (0.78–1.75)	2.13 (1.23–3.69)	2.33 (1.45–3.75)
HPV & Cytology
Negative	185	1		1
Abnormal cytology	18	1.08 (0.51–2.29)		1.17 (0.47–2.92)
HPV+ only	22	2.00 (1.13–3.52)		2.36 (1.19–4.70)
HPV+ & abn. cyto	6	1.15 (0.40–3.28)		0.94 (0.26–3.67)
Sick in past 7 days
No	158	1	1	1	1
Yes	80	0.76 (0.55–1.04)	0.74 (0.61–0.90)	0.77 (0.53–1.12)	0.72 (0.56–0.92)
Ectopy
External OS	93	1	1	1	1
Tx zone in canal	76	1.16 (0.79–1.70)	1.45 (1.06–1.99)	1.38 (0.88–2.16)	2.10 (1.44–3.06)
Moderate/extensive	69	1.59 (1.07–2.38)	1.58 (1.14–2.19)	1.93 (1.19–3.12)	2.03 (1.39–2.98)
Blood in the sample
Absent	55	1	1	1	1
Trace	116	2.21 (1.57–3.12)	1.54 (1.22–1.94)	2.89 (1.95–4.30)	1.62 (1.22–2.14)
Visible	67	3.46 (2.40–4.99)	1.82 (1.36–2.43)	4.69 (3.06–7.19)	1.81 (1.27–2.60)
Specimen Weight
≤0.12	96	1		1
>0.12	142	0.30 (0.23–0.40)	0.13 (0.10–0.18) §	0.29 (0.21–0.40)	0.14 (0.10–0.21) §
Visit
Visit 1	77	1	1	1	1
Visit 2	82	0.21 (0.17–0.27)	0.45 (0.36–0.57)	0.17 (0.13–0.22)	0.35 (0.26–0.47)
Visit 3	79	0.53 (0.41–0.69)	0.64 (0.51–0.80)	0.47 (0.34–0.64)	0.57 (0.42–0.74)

^*Numbers over the 3 visits;

^†Adjusted for the other significant variables in the model;

^§Entered as a continuous variable in the multivariate model.

Table 2 presents the univariate and multivariate results of covariates of IgA and IgG levels among OC-users. In the final multivariate model, after adjusting for time in cycle and specimen weight; blood in the sample and ectopy were associated with higher IgA and IgG. Reporting a recent illness was associated with lower IgA and IgG. A significant association was observed between lifetime number of pregnancies and IgG levels. We did not observe any association with HPV or cytology, current smoking, years since menarche, vaginal douching, vaginal discharge, talcum powder use, recent medication, Chlamydia trachomatis, Gonorrhea, or vitamin use (data not shown).

Table 2.

Crude and adjusted Determinants of IgA and IgG levels between women: OC users

		IgA		IgG
	N*	Percent Change (95%CI)	Adjusted Percent Change† (95%CI)	Percent Change (95%CI)	Adjusted Percent Change† (95%CI)
Age (years)		1.06 (1.00–1.11) §	1.05 (1.00–1.10) §	1.05 (0.99–1.13) §	1.04 (0.98–1.10) §
Lifetime pregnancy
0/1	45	1	1	1	1
2/3	92	1.17 (0.83–1.66)	1.10 (0.79–1.52)	1.38 (0.92–2.07)	1.47 (1.02–2.12)
4–6	51	1.29 (0.88–1.90)	1.25 (0.89–1.77)	1.27 (0.81–1.99)	1.31 (0.89–1.91)
Years since 1st Pregnancy
0–11	73	1		1
12+	109	1.39 (1.05–1.83)		1.35 (0.97–1.88)
Days since last sex
<= 2 days	79	1		1
3–10 days	64	1.20 (0.98–1.46)		1.28 (0.98–1.68)
11+ days	45	1.24 (0.95–1.62)		1.12 (0.79–1.59)
IUD Use
Yes	0	NA	NA	NA	NA
No	188
HPV & Cytology
Negative	122	1		1
Abnormal cytology	23	0.82 (0.50–1.37)		1.20 (0.66–2.21)
HPV+ only	22	0.88 (0.57–1.36)		1.20 (0.71–2.03)
HPV+ & abn. cyto	17	0.76 (0.46–1.24)		1.04 (0.58–1.87)
Sick in past 7 days
No	135	1	1	1	1
Yes	53	0.89 (0.73–1.08)	0.85 (0.71–1.00)	0.69 (0.53–0.90)	0.67 (0.53–0.84)
Ectopy
External OS	45	1	1	1	1
Tx zone in canal	66	1.29 (0.91–1.85)	1.38 (1.01–1.90)	1.72 (1.13–2.60)	1.86 (1.30–2.65)
Moderate/extensive	77	1.20 (0.85–1.69)	1.51 (1.04–2.18)	1.15 (0.77–1.72)	1.27 (0.89–1.83)
Blood in sample
Absent	22	1	1	1	1
Trace	104	1.39 (1.06–1.81)	1.32 (1.00–1.89)	1.82 (1.25–2.65)	1.86 (1.35–2.58)
Visible	62	2.15 (1.62–2.85)	1.35 (0.97–1.88)	2.57 (1.71–3.85)	2.25 (1.55–3.28)
Specimen Weight
≤0.12	135	1	0.34 (0.23–0.52)§	1	0.19 (0.11–0.32) §
>0.12	53	0.74 (0.59–0.91)		0.55 (0.42–0.72)
Visit
Visit 1	62	1	1	1	1
Visit 2	64	0.69 (0.56–0.81)	0.72 (0.61–0.85)	0.68 (0.52–0.89)	0.68 (0.53–0.86)
Visit 3	62	0.68 (0.57–0.81)	0.70 (0.58–0.84)	0.76 (0.58–0.98)	0.73 (0.57–0.94)

^*Numbers over the 3 visits;

^†Adjusted for the other significant variables in the model;

^§Entered as a continuous variable in the multivariate model.

Factors associated with magnitude of change in IgA and IgG within women

We observed an average decline of 78% in IgA, and 83% in IgG between the follicular and peri-ovulatory visits among naturally cycling women. For OC users, the decline in levels between the follicular and peri-ovulatory visits was 33% for IgA and 34% for IgG. Conversely, between the peri-ovulatory and luteal visits, levels increased on average 58% for IgA and 62% for IgG for naturally cycling women, while for OC users, the respective increases in levels between the peri-ovulatory and luteal visits were 2% (IgA) and 15% (IgG) (Figure 2).

Next we evaluated factors associated with IgA and IgG change over the menstrual cycle. Although not statistically significant, among naturally cycling women (table 3), the percent change between follicular and peri-ovulatory visits was greater for older women for both IgA and IgG. Between the peri-ovulatory and luteal visits the opposite was seen, i.e. the percent change for older women was less than that seen for younger women. The percent change in both IgA and IgG was lower among those who were sick during the cycle. Women with a greater increase in specimen weight between follicular and peri-ovulatory phases, and between peri-ovulatory and luteal phases had less of a change in IgA and IgG levels. Lastly, higher immunoglobulin levels at visit 1 were associated with less change between follicular and peri-ovulatory visits but higher change between peri-ovulatory and luteal visits.

Table 3.

Factors associated with change in immunoglobulins within women: according to OC use

Naturally cycling Women
Change: Mean (SD)	Difference between visit 2 and 1 (absolute difference)		Difference between visit 2 and 3 (absolute difference)
	IgA 0.22 (3.17)	IgG 0.17 (3.46)	IgA 0.42 (3.52)	IgG 0.38 (3.57)
	Percent Change	Percent Change
Age
26–31	1	1	1	1
32–35	1.14 (0.80–1.62)	1.16 (0.75–1.79)	0.80 (0.52–1.24)	0.70 (0.43–1.12)
Recent illness
None at either visit	1	1	1	1
Sick at one visit only	0.80 (0.54–1.17)	0.59 (0.37–0.94)	1.12 (0.67–1.87)	0.96 (0.54–1.68)
Sick at both visits	0.36 (0.22–0.60)	0.36 (0.20–0.68)	0.53 (0.29–0.96)	0.38 (0.20–0.73)
Change in specimen weight
1 unit increase	0.15 (0.09–0.26)	0.17 (0.09–0.31)	0.13 (0.08–0.23)	0.20 (0.11–0.36)
Immunoglobluin level at visit 1
1 unit (log) increase	0.78 (0.64–0.95)	0.79 (0.64–0.97)	1.11 (0.89–1.39)	1.16 (0.93–1.45)

OC-Using Women
	Change between visit 2 and 1 (absolute difference)		Change between visit 2 and 3 (absolute difference)
Change: Mean (SD)	IgA 0.67 (1.87)	IgG 0.66 (3.48)	IgA 0.98 (2.08)	IgG 0.85 (2.50)
Age
26–31	1	1	1	1
32–35	0.86 (0.66–1.12)	1.48 (0.91–2.41)	0.82 (0.55–1.23)	1.21 (0.75–1.95)
Recent illness
None at either visit	1	1	1	1
Sick at one visit only	1.24 (0.94–1.63)	0.75 (0.44–1.26)	1.16 (0.77–1.76)	0.77 (0.46–1.28)
Sick at both visits	1.08 (0.69–1.68)	1.04 (0.47–2.29)	1.10 (0.46–2.63)	0.42 (0.15–1.21)
Change in specimen weight
1 unit increase	0.52 (0.29–0.92)	0.23 (0.07–0.71)	0.43 (0.20–0.92)	0.24 (0.10–0.60)
Immunoglobluin level at initial visit
1 unit increase	0.67 (0.56–0.81)	0.51 (0.39–0.67)	1.07 (0.82–1.39)	1.01 (0.80–1.28)

Similar patterns were observed among OC users (table 3). Women with a greater increase in specimen weight between follicular and peri-ovulatory visits and between peri-ovulatory and luteal visits had lower change in both IgA and IgG levels. Higher IgA and IgG levels at follicular visit were associated with significantly lower changes between visits 1 and 2 but greater changes between visit 2 and 3 (not significant).

We did not observe any association with current smoking, years since menarche, vaginal douching, vaginal discharge, talcum powder use, recent medication or vitamin use, parity, years since first pregnancy, days since last sex, IUD, blood in the sample, ectopy, and HPV or cytology status among both groups (data not shown).

Discussion

We were interested in understanding patterns and determinants of immunoglobulin fluctuations at the cervix during the menstrual cycle because of the potential impact of virus-specific neutralizing antibody fluctuations on the long-term efficacy of the newly licensed HPV vaccines. Levels of HPV antibodies following vaccination are typically 10-fold lower in the genital tract compared to systemic levels, and blood titers have been shown to decline with time since vaccination (6;12;13). While to date vaccine efficacy has been shown to remain high for up to 5 years following vaccination (14;15), the potential impact of these declining antibody levels on longer-term vaccine efficacy deserves further attention.

In our study, we demonstrated a sharp decline in total IgA and IgG levels around ovulation among naturally cycling women, consistent with a previous, smaller report (16). In addition, we observed OC-users typically had higher cervical immunoglobulins than naturally cycling women. However, even among OC-users, a pattern of declining immunoglobulins at ovulation was observed, albeit less pronounced than among naturally cycling women. The fact that systemic antibody levels decline in the years following vaccination, that local levels are considerably lower than observed systemically, and that these levels decline further at certain times during the menstrual cycle (6), underscores continued surveillance of HPV vaccinated women in the years following vaccination.

The patterns in immunoglobulin levels during the menstrual cycle suggest an important role for sex hormones in the regulation of mucosal immunity. Early in the follicular phase, estradiol levels are low; later in this phase, levels rise exponentially and peak just prior to the LH surge. At that time, estradiol levels plummet, only to rise again following ovulation, and plateau during the mid-luteal phase of the menstrual cycle. Previous animal and human studies have demonstrated a role of sex hormones, particularly estradiol, in the regulation of cervical immunity (16;17). In one study (16), immunoglobulin levels increased with increasing levels of estradiol. Since the LH surge signals a steep drop in estradiol pursuant to ovulation, the lower immunoglobulin response we observed may be associated with lower estradiol levels. The higher immunoglobulin levels observed during the luteal phase is consistent with a hormonal mechanism as estradiol levels rise again during this phase of the cycle. In women on OCs, the highest cervical immunoglobulin levels occurred during the early phase of the cycle, corresponding to increasing levels of norethindrone, a synthetic form of estradiol, but declined significantly in the last week when oral ingestion of the synthetic estrogen is stopped. Finally, our finding of a sharp decline in IgA and IgG levels at mid-cycle was particularly evident among the group of naturally cycling women with a confirmed LH surge.

A unique feature of the present study was the large number of women evaluated. This allowed us to examine not only the overall immunoglobulin levels and patterns in cervical secretions over the menstrual cycle among naturally cycling and OC-users, but to explore demographic, and lifestyle factors that might modulate these levels. In doing so, several correlates of immunoglobulin levels were observed.

Our finding that older age was associated with higher immunoglobulin levels, especially IgA levels among OC-users, is consistent with previous reports(18). While the population in the present study were all pre-menopausal, the age association is physiologically plausible, since older women have lower volumes of cervical secretions, hence less diluted and more concentrated immunoglobulin levels.

Higher number of pregnancies were associated with higher immunoglobulin levels among both OC-users and non OC-users. The strongest effect was for IgA among naturally cycling women, consistent with a previous report in which both local IgA and IgG levels were elevated among currently pregnant women compared to non-pregnant women (19). Our results showing an association with lifetime pregnancies suggest that the effect of pregnancy-related hormonal changes may be retained after pregnancy has ended. To examine this further, we restricted our final tables to women whose most recent pregnancy was >2 years before our study and observed the same trends as for the entire population (data not shown).

Reporting a recent illness was consistently associated with lower total IgA and IgG levels at the cervix among both OC-users and naturally cycling women. Most of the illnesses (68%) reported by the participants were of an acute infectious nature, mainly due to cold. These results were unexpected and suggest the need for additional studies to confirm and better understand these results.

As seen in other smaller studies (20–23), current IUD use was associated with higher IgG, and IgA levels, although significant only for IgG. IUD use can induce inflammation at the cervix and vagina, and potentially increased levels of immunoglobulins.

Consistent with previous reports, we observed that presence of hemoglobin in the cervical specimen was associated with higher IgA and IgG levels (18;24). This is expected, given that immunoglobulin levels are higher in serum than in the cervix. We collected cervical secretions prior to sampling for cytology, using a gentle sponge to passively collect samples so as to not abrade the cervix; hence blood detectable in specimens may reflect a biologically relevant source of immunoglobulin at the cervix rather than a collection-induced artifact. The association of higher immunoglobulines with cervical ectopy is also expected as with ectopy the glandular endocervical columnar epithelial cells are exposed to the ectocervix, making it more prone to bleeding.

We used specimen weight to approximate collection volume. While there are limitations for using this method for determining volume, it is interesting to note that weight was a strong predictor of immunoglobulin levels at the cervix. For both naturally cycling and OC-users, IgA and IgG levels were found to decrease with increasing specimen weight. This suggests that dilution effects of cervical secretions resultant from increased secretion levels during the peri-ovulatory period is an important determinant of immunoglobulin concentrations at the cervix. However, this dilution effect alone is unlikely to fully explain the patterns observed during the cycle, since other covariates (tables 1 and 2) were found to be independently associated with IgA and IgG levels in multivariate models that controlled for specimen weight.

Finally the degree of change in immunoglobulin levels during the menstrual cycle was not associated with any demographic or lifestyle variables that we evaluated, depending only on the initial immunoglobulin level, specimen weight and recent illness.

In summary, we observed a broad range of immunoglobulin levels across women during the menstrual cycle, being lowest around the time of ovulation. While dilution effects resultant from changes in the volume of cervical mucus during the cycle might partially explain the pattern observed, additional factors appear to exert independent effects on immunoglobulin levels at the cervix. Oral contraceptive use and lifetime pregnancies were important modifier of immunoglobulins detected at the cervix, further supporting a role of hormonal factors in defining mucosal immunological responses. Our study did not measure endogenous hormones, precluding a direct evaluation of their effect on cervical levels of immunoglobulins at this time. Implications of our findings include the need for surveillance of women who receive the new HPV vaccine in the years following vaccination to determine whether and when anti-HPV antibody levels at the cervix drop below protective levels, and what the effect of menstrual cycle changes and other factors are on virus-specific antibody titers and duration of vaccine-induced protection.