Sexual Intercourse Frequency During Pregnancy: Weekly Surveys Among 237 Young Women From A Random Population-Based Sample

Abstract

Background:

Significant differences in sexual frequency during pregnancy have been documented in cross-sectional and once-per-trimester longitudinal studies, with the highest sexual frequency in the first trimester and the lowest in the third trimester. However, changes in sexual frequency may be more complex than these comparisons suggest; patterns of sexual frequency have not been assessed using frequent (e.g., weekly) assessments throughout a woman’s pregnancy.

Aim:

To assess changes in the frequency of sexual intercourse across all weeks of pregnancy.

Methods:

We used data from 237 women (54% White; 43% Black) who reported a pregnancy during the Relationship Dynamics and Social Life (RDSL) study. RDSL was based on a random population-based sample of 992 women in the United States, aged 18 or 19, who completed a baseline interview and brief weekly follow-up surveys for 2.5 years. We used generalized multilevel modeling to fit and compare linear, quadratic, and piecewise (via b-splines) models.

Outcome:

Weekly probability of sexual intercourse.

Results:

Sexual intercourse frequency clearly declined across pregnancy, yet the pattern followed the course of common pregnancy symptomology (i.e., nausea, fatigue) more closely than trimester cutoffs. The best fitting model demonstrated that the probability of sexual intercourse declined sharply (~17% each week) between conception and 11 weeks, subsequently increased by ~3% each week between weeks 11 and 21, and then declined steadily (~6% each week) through the end of pregnancy.

Clinical Translation:

Documenting more precise patterns of change in sexual frequency during pregnancy provides important information to many who wish to maintain sexual intimacy while pregnant, or to those who would otherwise find the sexual disruptions particularly challenging.

Strengths & Limitations:

This study is the first to document changes in sexual intercourse frequency throughout all weeks of pregnancy as they naturally occurred among a representative sample of young women. The focus on sexual intercourse limits the findings to only one aspect of human sexuality. The narrow age range of the sample precludes generalization to all pregnant women.

Conclusion:

Changes in sexual frequency are more complex than the general declines suggested by other studies; within-trimester patterns reveal the shortcomings in understanding sexual behavior changes when aggregated by trimester, such as severely underestimating the degree of fluctuation in the first trimester. Pregnancy symptomology may be most favorable to intercourse towards the end of the first and beginning of second trimesters, and least favorable near the end of the pregnancy.

INTRODUCTION

Pregnancy is a period of intense physiological changes that may affect women’s¹ sexual functioning and disrupt sexual connections with intimate partners throughout this potentially sensitive time.^1,2 Understanding the trajectory of sexual intercourse frequency throughout pregnancy could offer valuable information to women, their intimate partners, and their medical care teams as they navigate this period of intense physiological change. For nulliparous women in stable intimate partnerships, pregnancy is also a time when the couple is preparing for parenthood, an intense shift in their relationship from a dyad to a triad, during which sexual connections with a partner could be of heightened importance.² However, sexual disruptions such as declines in frequency as well as desire and satisfaction are common during pregnancy.^1,2 For pregnant women who may struggle to maintain their preferred sexual connections with a partner, awareness of typical changes in sexual behavior are just as important as other bodily changes during pregnancy. However, the exact nature of changes in sexual activity across pregnancy is largely unknown, as most of the data has been summarized into trimesters—very broad time spans with high variability, which mask any patterns of change that may occur over periods shorter than three months. Documenting more precise changes in sexual frequency during pregnancy could provide women with critical information in terms of managing expectations for sex and giving them time to prepare for any changes they may experience.

Physiological Changes throughout Pregnancy

Pregnancy is typically characterized by three distinct trimesters (weeks 0-13, 14-24, and 25-40), which are often used to broadly distinguish between distinct episodes of physiological symptoms (e.g., the second trimester characterized as the “honeymoon” period due to reduced symptoms; the third trimester as the most physically uncomfortable), though the symptoms do not follow these trimesters exactly.³ During the first several weeks, progesterone levels and the body’s metabolism increase rapidly to support the pregnancy, and these are a primary source of many early pregnancy symptoms. Common symptoms in early pregnancy include nausea and vomiting (which typically begins between weeks 4 and 9), fatigue, breast tenderness, frequent urination, constipation, and heartburn.

After the first two or three months (around weeks 9-12), nausea and some other early pregnancy symptoms may begin to dissipate, while some symptoms may persist, such as constipation and heartburn, and additional symptoms may arise, such as abdominal cramps, dental changes (e.g., bleeding gums), nasal congestion, and back pain. The last few months are typically the most physically uncomfortable as the body continues to adjust and accommodate the growing fetus and typically experiences weight gain of about 30 pounds.² Backaches, constipation, and heartburn are common, as is frequent urination and disrupted sleep. As the pregnancy nears full term, other physical discomforts are also likely to arise, such as hemorrhoids, leg cramps and swelling, and pelvic pressure. Nausea may return for some women as well.

Changes in Sexual Functioning and Behavior throughout Pregnancy

Pregnancy symptoms—e.g., fatigue, physical discomfort, nausea, vomiting—are commonly cited reasons for avoiding sexual intercourse during pregnancy,¹ as are issues with sexual functioning—when women experience declines in sexual functioning, such as arousal or desire difficulties or increased pain during sex, they are less likely to engage in sexual intercourse during their pregnancy.¹ Given the link between sexual functioning and sexual behavior, documented changes in sexual functioning could also shed light on how the frequency of sexual intercourse may change during pregnancy, Cross-sectional and/or once-per-trimester longitudinal studies suggest changes in sexual functioning and behavior across pregnancy that generally correspond to the intensification of physiological changes as a pregnancy progresses.

Previous studies and systematic reviews have consistently reported an overall decline in sexual functioning and frequency over the course of pregnancy for most women, beginning fairly early in the first trimester and reaching the lowest point towards the end of the third trimester.^1,4,5 However, there have been few prospective longitudinal studies in the U.S. that have surveyed pregnant women about their sexual frequency more than a few times during their pregnancy. Instead, most studies have relied on long-term retrospective accounts of sexual frequency, cross-sectional comparisons of women currently in different trimesters of pregnancy, or women’s overall retrospective perceptions of whether their sexual frequency declined during pregnancy. To our knowledge, only one U.S. study assessed women more than three times during pregnancy—Elliott, and Watson⁶ interviewed 128 women from a London pregnancy center at 13, 16, 20, 28, 36, 38, and 39 weeks gestation. However, they aggregated these into four (undefined) periods to reduce missing data. Their rank-based indicated “clearly significant decreases (P< .0001)” (P. 543) in sexual frequency, satisfaction, and interest, though it is unclear which periods differed from each other, or what weeks gestation was included in each period.

A close look at the results from other studies suggests that sexual functioning and frequency may not monotonically decline throughout the entire pregnancy, yet these potential nuances were often not a focus of those studies. An early review (from 1999) of mainly cross-sectional or retrospective studies reported a general pattern of declining sexual functioning and frequency during the first and third trimesters and greater variability during the second trimester.² In a cross-sectional study of 589 pregnant women across all three trimesters, sexual function was similar among women in the first two trimesters, with the lowest sexual functioning scores reported by women in their third trimester.⁷ Notably, vaginal discomfort was significantly lower among women in the second trimester compared to women in the first or third trimesters, which may suggest a slight increase in sexual functioning during the second trimester. The authors’ ultimate conclusion was that sexual functioning was similar in the first two trimesters and lowest in the third trimester. In another cross-sectional study, of 271 healthy pregnant Brazilian women in committed romantic relationships, the researchers concluded that sexual functioning was similar among those in the first two trimesters and substantially lower among those in the third trimester.⁸ The results indicated that sexual functioning was slightly higher in the second trimester relative to the first. Yet, differences between the first and second trimesters were not statistically significant and the percent of those meeting criteria for sexual dysfunction was lower among those in the second trimester relative to those in the first trimester, which the authors explained as resulting from the second- trimester being a time of relative emotional stability, feminine reaffirmation, high pelvic blood flow, and reduced nausea. Their ultimate conclusion was that sexual functioning was lowest in the third trimester.

In two small, longitudinal studies—59 women in Turkey ⁹ and 30 women in Egypt ¹⁰—researchers assessed sexual functioning once per trimester and their results documented that sexual desire, arousal, lubrication, orgasm, and satisfaction increased slightly during the second trimesters. However, in both studies, the authors only remark that the lowest sexual function scores were found during the third trimester—the second-trimester increases in sexual function were not even mentioned (presumably because they were not an important part of either study’s purpose).

These slight increases in sexual functioning during the second trimester could translate to increases in sexual behavior. Accordingly, a longitudinal study of 663 women in Taiwan ¹¹ directly documented a small increase in sexual frequency (and sexual functioning) in the second trimester compared to the first, which was then followed by a large decline in the third, although second versus first trimester differences were not statistically significant. Overall, the main conclusion drawn by these previous studies is that sexual functioning and frequency generally decrease over the course of pregnancy, yet a careful examination of the results suggests that changes in frequency may be more complex than a simple linear decline.

Shortcomings in Prior Studies

When it comes to determining changes in sexual frequency throughout a pregnancy, these studies are limited in key ways. The cross-sectional studies assessed different women at different stages in pregnancy, inferring—but not demonstrating—longitudinal change, and with little to say about potentially more nuanced trajectories of sexual frequency. The longitudinal studies followed women across their pregnancies, assessing sexual frequency and/or function once per trimester, which could potentially reveal within-subject trajectories. However, these were not assessed; only proportions or means aggregated by trimester were presented. Moreover, collecting data only once per trimester could mask any within-trimester changes or other more complicated trajectories. It is unclear when, exactly, any increases or decreases in sexual frequency occur, or whether the large variability within each trimester could be explained by clear patterns of substantial change within trimester. Methodological limitations are of note as well, as prospective studies are rare (though not unheard of ³) and recruitment typically relies on the women receiving services at specific care centers, which reduces their generalizability.

The Current Study

The goal of the current study was to examine changes in the frequency of sexual intercourse across pregnancy, drawing from an intensive longitudinal design consisting of weekly reports. Using multilevel modeling techniques, which can account for the interdependence of repeated measures and assess within-subject trajectories, ^12,13 we compared 5 models of change, including linear, quadratic, and spline models. A linear spline model (aka, “piecewise,” “broken stick,” or “segmented” model), which models different rates of change across different periods, can document more complex patterns. Splines have been used in this way to model other trajectories during pregnancy, such as blood pressure and fetal growth.^12,13 Thus, spline models are ideally suited for our aims as they allow for potential changes in slope that correspond to, for instance, trimesters or to different stages of physiological adaptation throughout a woman’s pregnancy.

This study advances current knowledge in three key ways. First, the intensive longitudinal design (i.e., weekly surveys throughout pregnancy) allows for a more precise understanding of how sexual intercourse frequencies change over the course of pregnancy than has been documented before. Second, the sample and data collection were not dependent upon pregnancy-related factors— women were recruited via a simple random probability sample of a geographic area (not pregnancy care centers, for instance) and data was collected by assessing women’s weekly experiences over the course of 2.5 years, regardless of pregnancy status, and therefore represent more naturalistic data collection. Third, we examine young women (ages 18 through 22 at study end), who are less likely to have issues with sexual dysfunction,¹⁶ and thus any changes in sexual frequency are less likely to be related to aging-related sexual issues or pregnancy complications.

MATERIALS AND METHODS

Study Population and Recruitment

The Relationship Dynamics and Social Life (RDSL) study was an intensive longitudinal study designed to assess young women’s relationship, sexual, and pregnancy experiences,¹⁷ and included a baseline interview and weekly follow-up surveys. The University of Michigan’s Survey Research Center (SRC) drew the sample and collected the data. Respondents were enrolled across 16 months during 2008 and 2009. The population-based sample of women aged 18-19 years old was drawn from Genesee County in Michigan via driver’s license and Personal Identification Card databases, which (at the time) were provided to researchers at no charge by Michigan’s Secretary of State. The SRC estimated the two lists provided coverage of 95% of the county’s population of 18 and 19-year-old women. Potential respondents were invited (via a letter sent on University of Michigan letterhead) to participate in a study called the “Michigan Study of Young Women,” a research study about social life. Of the population of 18-19-year-old women at the time, 1,208 (about 1 in 6) were contacted to participate. Of the 1,208 women contacted, 1,003 completed a baseline interview, resulting in a response rate of 83%.¹⁸ Of these 1,003 women, 992 (99%) participated in the follow-up surveys, and 744 (75%) remained in the study for at least 18 months.¹⁹ A total of 58,594 weekly follow-up surveys were submitted, which was approximately 45.4% of the total possible weekly surveys. However, the RDSL was designed with the expectation that not every possible weekly survey would be submitted (as is typically true in many intensive longitudinal studies, e.g., ²⁰) and thus the retrospective reporting period (for sexual intercourse and other behaviors) was adjusted to match the time since the prior survey. If more than 14 days had passed, respondents were asked to report on the previous 7 days. Detailed information about RDSL design features, along with detailed information about sample selection, recruitment procedures, retention strategies, and response rates for the RDSL study can be found in.^18,19

In all, 239 of the 992 women reported 287 pregnancies during the study period. The characteristics of the RDSL sample corresponded closely to those of the 18-19-year-old women in the National Survey of Family Growth sample in terms of previous pregnancies and births, high religiosity, the mother with the first birth as a teen, two-parent household during childhood, receipt of public assistance at age 18/19, experiencing penile-vaginal intercourse prior to age 17.²¹ However, 26% of the RDSL dataset had ever been pregnant at the time of the baseline interview (age 18 or 19), which is higher than the 19% in the NSFG. This is at least in part because a larger portion of the RDSL respondents self-identified as African American or Black (34%), which matched the population of the county, but is higher than in the NSFG’s national sample (16%). Because African American women on average have their first pregnancies and births younger than white women, this accounts for these differences in pregnancy rates between the RDSL sample and the U.S. population. Therefore, we are confident that our sample of 239 women who experienced pregnancy is an unbiased sample of women who experienced pregnancies among the population of 18-22-year-old women in that county during those years, and thus the characteristics of their pregnancies are a reasonable representation of that population of pregnancies. Further, the 239 women largely do not differ from women in U.S. population, with the exception of a somewhat larger fraction of African-American women.

To maintain statistical independence at the pregnancy level (critical for nested data such as these, with some women reporting multiple pregnancies;¹³), and because the majority of women only reported a single pregnancy, the current study includes one pregnancy per woman. If a woman reported more than one pregnancy, we used the pregnancy that provided the most information (i.e., more weekly surveys were submitted by the woman during the pregnancy). If the exact same number of weeks were reported for multiple pregnancies, we used the earliest pregnancy. During these 239 pregnancies, two women did not provide any data on sexual intercourse. Thus, the final sample was 237 pregnancies, or one pregnancy each for 237 women. During these pregnancies, the women submitted 2,897 weekly surveys. Figure S1 presents the distribution of surveys across pregnancy weeks, with elevated frequencies during the beginning few weeks of pregnancy, but a fairly uniform distribution overall. Elevated frequencies early during pregnancy is due at least in part to some pregnancies ending early via abortion or miscarriage.

Procedure

The study began with a face-to-face baseline interview (conducted from 2008-2009) and the weekly follow-up surveys spanned approximately 2.5 years following the baseline interview. The baseline interview was conducted by a professional interviewer from the SRC at a time and location of the woman’s choosing, allowing her to decide where she felt most comfortable (the interviewer met her wherever she chose; most interviews took place at the respondent’s home; 16% were at another location like a library). Sensitive questions (e.g., pregnancy, adolescent sexual behaviors, receipt of public assistance) were administered via computer aided self-interview (CASI), during which the respondent silently answered the questions on the interviewer’s laptop outside of the interviewer’s view. The weekly surveys were completed either by website (via username and password) or phone call to the University of Michigan’s Survey Research Center’s phone survey unit, per the respondent’s choice each week. Detailed information about these methods is presented in.¹⁹ Compensation included $5 sent in the letter in advance of the baseline interview and $30 for completing the baseline interview. Incentives for the weekly surveys included $5 per survey for the first four weeks and $1 per survey after that, and $5 bonuses were given for on-time completion of five sequential surveys. Participants provided written informed consent for the baseline interview and assent for the online weekly surveys; the study was approved by the Institutional Review Board at the University of Michigan.

Measures

Sexual intercourse frequency.

The weekly surveys asked respondents if they had penile-vaginal sexual intercourse during the past week (or since their last survey, if their previous survey was submitted 8-14 days prior). The survey provided a definition of sexual intercourse each time: “By sexual intercourse, we mean when a man puts his penis into a woman’s vagina.” (Hetero)sexual intercourse frequency was coded as a binary variable, with 0 = no intercourse that week and 1 = at least one instance of intercourse that week. The survey did not ask how often they had sex during the week. Thus, the weekly surveys assessed a specific measure of sexual frequency—a week-level time-varying indicator of whether the woman engaged in sexual intercourse at least once each week.

Week of Pregnancy.

In each weekly survey, respondents were asked, “Do you think there might be a chance that you are pregnant right now?” Respondents who answered yes were asked, “Has a pregnancy test indicated that you are pregnant?” Respondents who answered yes to the question about the pregnancy test were coded as pregnant. Week of conception was estimated based on a straightforward algorithm that drew from several pieces of information: when the pregnancy was reported, the due date (which was updated during the weekly interviews), the weeks during which the woman had sex with the father, and the birth date (if it occurred during the study period). For the small number of pregnancies (n=12; 5%) that did not include all pieces of information or where one or more pieces of information conflicted with the others (usually because no due date was provided and the birth did not occur during the study), week of conception was hand-coded based on all available information spanning several months before and after the pregnancy was reported.

Consistent with the professional medical and obstetric community,² the estimated day of conception was considered week 2 of pregnancy.² Weeks beyond the 40th were not included; very few women reported pregnancies that lasted more than 40 weeks (<3%).

Woman characteristics.

The models accounted for multiple individual characteristics that are associated with sexual behavior. Age at the beginning of the pregnancy is coded in years. Women’s sexual and reproductive history included young age at first intercourse (if prior to age 17, coded as 1; otherwise, 0) and number of previous pregnancies (reported value). Women’s family background is represented by an index of childhood disadvantage that is a sum of four indicators: mother did not graduate from high school (yes=1; otherwise, 0), mother’s first birth was as a teen (if prior to age 20, coded as 1; otherwise, 0), grew up in non-two-parent household (yes=1; otherwise, 0); and received public assistance during childhood (yes=1; otherwise, 0). High religiosity was coded as religious faith very important or more important than anything else=1; not at all or somewhat important=0. Relationship status was also included (married/engaged/cohabiting=1; otherwise, 0). African American racial identity (coded yes=1) was also included as it was the largest racial/ethnic minority group, and other groups did not comprise a large enough subsample for testing group differences.

Modeling Change Across Pregnancy

To understand how sexual intercourse frequency changes over time, statistical models representing different types of change (e.g., linear, quadratic) can be compared to determine which fits the data best. Previous studies and reviews suggested an overall decline in sexual frequency, with the highest sexual frequency during the first trimester and the lowest sexual frequency during the third trimester, indicating a potentially consistent, linear decline. Several studies also suggested a steeper decline in sexual functioning or frequency towards the end of the third trimester, suggesting a quadratic term may be important for correctly modeling the trajectory.

The studies suggesting an increase in sexual functioning and frequency during the second trimester (relative to the first) indicate a potentially more complex trajectory. The distinct patterns of physiological symptoms typically experienced during pregnancy, which are often closely tied to the experiences of sexual intercourse (e.g., dyspareunia, physical discomfort ¹), also suggest that any decline may not be uniform across an entire pregnancy. We, therefore, incorporate linear spline models to document this more complex pattern. In these models, trajectories are split into connected phases, and each phase has a different slope. The phases are connected at breakpoints, or “knots.” Our first spline model included breakpoints at the beginning of each trimester, allowing for different linear slopes from conception to week 14 (first trimester), from week 14 to week 26 (second trimester), and from week 26 onward (third trimester). Our second spline model included breakpoints

For the first step in the model selection process, we created a descriptive line plot of the proportion of women who indicated they had sexual intercourse that week during each week of pregnancy (i.e., raw data; see Figure 1 panel a). Next, we fit a linear random intercept model, with pregnancy week included as a time-varying predictor and woman characteristics included as control variables. The second model was a random slope model, which allowed the slope for pregnancy week to vary across women, and the third model included a quadratic term for pregnancy week. The fourth model used b-splines to insert break points at 14 and 25 weeks, i.e., the weeks officially designated as trimester cutoffs in the medical and obstetrical community.² The fifth model also used b-splines, with the break points based on evident points of change in the raw data. Specifically, based on the graph of the raw data (Figure 1, panel a), the authors agreed that clear shifts in the pattern were seen around weeks 11 and 21. Break points were thus set to weeks 11 and 21.

Figure 1.

Raw data (a) of the proportions of women who reported sexual intercourse during each week of pregnancy, and results from the generalized multilevel models overlaying the raw data (b-f).

Note. Model results depict predicted probabilities of sexual intercourse by pregnancy week. Week 2 of pregnancy designates week of conception, consistent with the professional medical and obstetric community.² The trimester break points were at weeks 14 and 25.² Visual inspection breakpoints were at weeks 11 and 21. Gray shaded area represents 95% confidence interval. Note that some lines are not strictly linear because the transformation from log-odds to predicted probabilities is curved. N = 2,872 weekly surveys during 237 pregnancies among 237 women. Week of conception was not included in the models due to it being an outlier with 100% of women reporting sex that week.

Data Analysis

The repeated measures data is nested at two levels, with the weekly surveys nested within women. Multilevel modeling is an ideal analytical approach as it can simultaneously examine effects across weeks (within-subject effects) and across women (between-subject effects), while accounting for the inherent interdependence within the levels.²² We used two-level generalized multilevel models (GMLM), with Bernoulli distribution, logit link function, and full maximum likelihood estimation to estimate the probability of sexual intercourse as a function of pregnancy week. Because the proportion of sex during the week of conception was unusually high (100%), models were run without this outlier.³ Data were analyzed using R 4.1.0 ²³ via RStudio 1.4.1106. Multilevel models were run using the ‘glmer’ function, part of the ‘lmer’ package.²⁴ R syntax available in supplemental materials.

We assessed model fit using several criteria, including the Akaike information criterion (AIC) and the Bayesian (or Schwarz) information criterion (BIC). Vrieze ²⁵ provides a thorough discussion on the distinct theoretical principles and purposes of each. The BIC incorporates a heavier penalty for more complex models. Thus, the BIC tends to favor less complex models in comparison to the AIC, which prioritizes predictive utility over parsimony. Because of the multilevel data structure, the BIC was calculated using N = number of women (versus number of observations, which is often the default).²⁵ Lower values indicate better model performance. Estimators for explained variance that are specific to generalized multilevel models include the marginal R², which is an estimate of variance explained by the fixed effects, and the conditional R², which can be interpreted as the variance explained by both fixed and random effects (i.e., the entire model).²⁶ Higher R² values indicate greater predictive utility of the model. Models were also compared using the likelihood ratio chi-squared test when applicable (via the ‘anova’ function from the ‘stats’ R package ²³). We also present the log-likelihood and deviance, in which smaller absolute values indicate better fit.

Lastly, because the physiological symptoms that accompany ectopic pregnancies or miscarriages may differ from other pregnancies, we conducted a sensitivity analysis with only the pregnancies that ended in a birth or were ongoing at the end of the study. Pregnancies ending in abortion were also not included as it was not clear why they were terminated early. This resulted in a subsample of n=199 for the sensitivity analysis.

RESULTS

Participant Characteristics

Among the 237 women, the average age at the beginning of pregnancy was M = 20.1 years (SD = 0.9, range 18-22). The majority identified as White (128; 54.0%), with a large portion (101; 42.6%) identifying as African American, and <4% identifying as either multiple races, American Indian, Asian, or Pacific Islander, or refusing to answer; most (55%) were highly religious. Women were married, engaged, or cohabiting during 2,311 (79.8%) of the weeks during their pregnancies; 145(61.2%) were married, engaged, or cohabiting (shared address) during the entirety of the pregnancy, 31(13.1%) were never married, engaged, or cohabiting during their pregnancy, and the remaining were married, engaged, or cohabiting during an average M= 60.0% (SD = 27.1%, range 10-96%) of their submitted reports during pregnancy. Among the 143 (60.3%) women in the current study who completed a supplemental survey administered April-May 2010, one woman self-identified as a lesbian, 14 (9.8%) as bisexual, 112 (78.3%) as straight, and 16 (11.2%) as another identity.

As would be expected for a sample of young pregnancies, there are elevated rates of previous pregnancies and young penile-vaginal intercourse experiences among these women compared to the general population of 18-22-year-old women. Of the 237 women, 142 (59.9%) had never experienced a pregnancy prior to the study, 62 (26.2%) had experienced a single pregnancy prior to the study, and 33 (13.9%) had experienced 2 pregnancies prior to the study. Mean number of previous pregnancies was M= 0.54 (SD=0.73, range 0-2) and 60 (25.3%) women reported a young age at first intercourse (<17).

Sexual Intercourse During Pregnancy

Table 1 presents descriptive statistics for pregnancies and sexual intercourse weeks. Overall, the women had sexual intercourse in 63% of weeks during their pregnancies. Percent of intercourse weeks by trimester were 67% during the first, 65% during the second, and 57% during the third. Of the 237 pregnancies, most (77%) ended in a live birth during the study period; 9% ended in a miscarriage or ectopic pregnancy, 6% ended in abortion, and 8% of the women were still pregnant at the end of the study.

Table 1.

Descriptive statistics of pregnancies and intercourse weeks

Characteristic	N (%)
Weeks from each trimester
First	767 (26.5%)
Second	972 (33.6%)
Third	1,158 (40.0%)
Weeks with sexual intercourse	1,812 (62.5%)
Weeks with sexual intercourse, by trimester
First	521 (67.9%)
Second	635 (65.3%)
Third	656 (56.6%)
Pregnancy outcome
Live birth	181 (76.4%)
Miscarriage/ectopic	22 (9.3%)
Abortion	14 (5.9%)
Still pregnant at end of study	18 (7.6%)
Unknown	2 (0.01%)
Pregnancies that began prior to study onset	60 (25.3%)
Pregnancies ending in a live birth that began and ended during study	125 (52.7%)

Note: 2,872 weekly surveys during N = 237 pregnancies reported by 237 women in the Relationship Dynamics and Social Life (RDSL) study.

Table 2 presents the pregnancy week coefficients, model fit statistics from the five generalized multilevel models, and results from the likelihood ratio tests when applicable (full model output presented in Table S2). Overall, week of pregnancy was a significant and negative predictor of sexual frequency in all models, suggesting a declining probability of sexual intercourse throughout pregnancy. However, the nature of the decline differed in each model. Figure 1 illustrates the models graphically (panels b-f), with the predicted probability of sexual intercourse on the y-axis and the week of pregnancy on the x-axis.

Table 2.

Week of pregnancy predicting week-level sexual intercourse: coefficients and fit indices from generalized multilevel models

	Model 1: Random intercept		Model 2: Random slope		Model 3: Quadratic		Model 4: B-splines (Trimester)		Model 5: B-splines (Visual)
	B (SE)	AOR	B (SE)	AOR	B (SE)	AOR	B (SE)	AOR	B (SE)	AOR
Key Variables
Intercept	−1.96 (0.39)	0.14**	−2.28 (0.41)	0.10**	−2.09 (0.46)	0.12**	−1.82 (0.46)	0.16**	−1.46 (0.49)	0.23*
Week of pregnancy	−0.04 (0.01)	0.96**	−0.04 (0.01)	0.96**	−0.06 (0.03)	0.94*
Week of pregnancy (squared)					0.001 (0.001)	1.00
Week of pregnancy splines a
First segment							−1.40 (0.38)	0.25**	−1.74 (0.43)	0.18**
Second segment							−1.16 (0.36)	0.31*	−1.47 (0.39)	0.23**
Third segment							−1.93 (0.45)	0.15**	−2.28 (0.48)	0.10**
Model Characteristics
Fixed parameters	9		9		10		11		11
Random parameters	1		3		3		3		3
AIC	2193.0		2157.5		2157.9		2151.9		2148.1
BICb	2227.6		2199.1		2203.0		2200.5		2196.6
Log-likelihood	−1086.5		−1066.7		−1065.9		−1062.0		−1060.0
Deviance	1749.1		1622.2		1613.5		1598.2		1593.7
Marginal R2	0.453		0.440		0.442		0.479		0.482
Conditional R2	0.711		0.736		0.740		0.725		0.728
Likelihood Ratio Test	X2dif	p	X2dif	p	X2dif	p	X2dif	p	X2dif	p
vs. Model 1	–	–	39.49	< .001	41.07	< .001	49.05	< .001	52.88	< .001
vs. Model 2	–	–	–	–	1.58	.209	9.56	.008	13.39	.001
vs. Model 3	–	–	–	–	–	–	7.98	–c	11.81	–c
vs. Model 4	–	–	–	–	–	–	–	–	3.83	–c

Note. B = Log-odds. SE = Standard error. AOR=Adjusted odds ratio. AIC=Akaike’s information criterion. BIC=Bayesian information criterion (calculated with N = number of pregnancies [versus observations]).

X²_dif = Difference in X² between the two models. The trimester break points were at weeks 14 and 25 [2]. Visual inspection breakpoints were at weeks 11 and 21. Full model output presented in Table S2.

^aSpline coefficients are comparative, and do not represent the individual slopes of each segment. For calculations of individual slopes, see Table S1.

^bNumber of previous pregnancies reported by participants (range: 0-2).

^cLikelihood ratio test not possible because models are not nested; only X²_dif reported.

^*p < .05.

^**p < .001.

Model 1 included a random intercept term, which means that the model allowed the intercepts to differ across women. It did not include a random slope term, and therefore assumed that every woman’s probability of sexual intercourse changed in the same way across their pregnancies. The significant negative coefficient indicates a decline in the probability of sexual intercourse, with a 4% lower likelihood (AOR = 0.96, P< .001) of sexual intercourse for each successive week during pregnancy. According to this model, the predicted probability of sexual intercourse began at .82 and declined each week to a low of .44. However, all fit statistics indicated that this model was the poorest fit for the data.

Model 2 added a random slope term, which assumes that each woman’s probability of sexual intercourse throughout pregnancy changes at a different rate. The significant negative coefficient indicates that the average change across all women was a 4% decline in likelihood of sexual intercourse each week (AOR = 0.96, P< .001). The results of Model 2 are very similar to Model 1; the probability of sexual intercourse in this model ranged from .79 to .48. Adding the random slope significantly improved the fit compared to Model 1 (X²₍₂₎ = 38.4, P< .001).

Model 3 added a squared term to Model 2. The coefficient for weeks squared was not statistically significant (P = .32). Adding the quadratic term did not significantly improve the fit relative to Model 2 (X²₍₁₎ = 3.56, P = .059).

Model 4 estimated the trajectories of the probability of sexual intercourse separately for each trimester (defined by breaks at weeks 14 and 25). Note that the spline coefficients in Table 2 are comparative, and do not represent the individual slopes of each segment; calculations of individual slopes are presented in Table S1. This model showed that the probability of sexual intercourse decreased rapidly during the first trimester, with 11% decreases each week (AOR=0.89). The probability of intercourse also declined in the third trimester, but less rapidly than in the first trimester (~5% decrease in probability of sexual intercourse each week; AOR=0.95). Notably, the probability of sexual intercourse actually increased during the second trimester by about 2% each week (AOR=1.02). Using splines improved the model fit compared to Models 2 and 3, according to fit indices; likelihood ratio tests indicated significant improvement over Model 2 (X²₍₂₎ = 15.1, P<.001) (Model 3 not nested).

Model 5 was very similar to model 4, but used apparent breaks seen in the raw data at weeks 11 and 21 rather than trimesters to define the periods. The same overall pattern was found as in Model 4, but with an even steeper decline in the first segment—from weeks 2-11, the probability of sexual intercourse decreased by 18% each week (AOR = 0.82). This was followed by a slight increase in the second segment, 3% each week, from weeks 11-21 (AOR = 1.03). In the third segment, the probability of sexual intercourse decreased by 6% each week AOR = 0.94). Fit indices indicated an improvement to model fit compared to Model 4. The likelihood ratio test is not applicable for comparing these two spline models because, though the breakpoints are at different places, they have the same set of parameters representing the three segments (i.e., they are not nested). However likelihood ratio tests can be used to compare this model to Model 2, and the results again indicate a clear improvement over Model 2 (X²=11.8, P< .001). Overall, the visual spline model, based on changes in slopes seen in the raw data at 11 and 21 weeks, fit the data better than all other models. The visual spline model also appears to fit the data best compared to all the other models (Figure 1).

Based on the several fit indices and likelihood ratio tests, the superiority of the spline models indicates that the additional predictive utility outweighed the additional complexity. Additionally, because the two different spline models were equivalent in their complexity, the more favorable fit indices of the visual-based spline model suggest it provided the best description of the data, even though the indices were not drastically better than the trimester-based spline model.

Estimates for the control variables were consistent in each model (Table S2). No significant associations were found for age at beginning of pregnancy, childhood disadvantage, high religiosity, early first intercourse, or number of previous pregnancies. African American racial identity was associated with less frequent sexual intercourse, and marriage, engagement, and cohabitation were associated with more frequent sexual intercourse compared to less serious relationship types.

The sensitivity analysis (i.e., models run only with pregnancies ending in birth or ongoing at the end of the study) indicated similar results. The spline model with breaks at weeks 11 and 21 fit the data best and indicated a steep decline sexual intercourse frequency, followed by a slight increase and then a decrease. See Table S3 in the supplementary materials for model characteristics and fit indices from each model.

DISCUSSION

The current study documented the trajectory of sexual intercourse frequency over the course of pregnancy, drawing from a naturalistic, intensive longitudinal study of sexual intercourse among a random sample of young women from age 18 to 22. The frequency of sexual intercourse declined during pregnancy, but the decline was not uniform throughout the entire pregnancy. Rather, the best fitting model indicated that the pattern of changes in sexual intercourse during pregnancy were more complex than previously documented, with periods of both decreases and increases. A sharp decrease during the first two months (weeks 2-11) was followed by a period of small increase, and then a less sharp decrease starting in the middle of the second trimester and continuing through the third trimester (weeks 24-38). To our knowledge, this is the first time these nuanced within- and across-trimester patterns have been documented. Knowing this complex pattern of changes in sexual frequency throughout pregnancy can assist clinicians in providing accurate information to women and couples about average fluctuations during pregnancy in terms of sexual behavior. Awareness about potential significant and sudden declines in sexual frequency will likely prove beneficial to women who may otherwise find such sexual disruptions challenging and stressful. Moreover, the second trimester is often referred to as the “honeymoon period” of pregnancy due to reduced symptoms—yet our results suggest, at least when it comes to penetrative sexual activity with a partner, that this period may actually begin earlier, and that only about half the second trimester may really be experienced as a “honeymoon” period.

The results echo average patterns of sexual functioning documented across pregnancy ²—yet the inflection points for the slopes did not follow standard trimester cutoffs. The increase in sexual frequency began around week 11, earlier than the beginning of the second trimester, and ended around week 21, earlier than the beginning of the third trimester. This is notable because a large majority of previous research on sexuality throughout pregnancy reported outcomes by trimester, averaging results within the trimester. Due to the dynamic changes within pregnancy trimesters documented here, aggregating sexual behavior by trimester likely masks important sexuality changes. When aggregated by trimester, changes in sexual intercourse frequency across the trimesters may appear uniformly negative. The average predicted probability of intercourse was .68 in the first trimester, .65 in the second trimester, and .57 in the third trimester—a clear, decreasing pattern that is consistent with cross-sectional studies and those with one observation per trimester. However, the probability ebbed and flowed within the first two trimesters, with the most dramatic decline documented only within the first two months. In the first trimester, the predicted probability of intercourse ranged from .63 (week 11) to .91 (week 3; not including week of conception), and in the second trimester, ranged from .65 (weeks 14 and 25) to .70 (week 21); yet when averaged within trimester, as is done in most previous studies, these fluctuations are completely masked. Moreover, the frequency of sexual intercourse during the first trimester may be artificially low in other studies compared to ours because we were able to include the earliest weeks of pregnancy, during which many women may not even know they are pregnant (and therefore are unlikely to be visiting pregnancy care centers where much recruitment occurs).

Notably, nausea and vomiting, common pregnancy symptoms, tend to begin fairly early in pregnancy, but begin to largely subside between weeks 9 and 12, ³ which matches the trajectory of sexual intercourse frequency in the current study. This supports previous findings from retrospective reports ¹ that nausea may play an important role in women’s desire for and engagement in sexual intercourse. Assessing pregnancy symptomology was beyond the scope of the RDSL study–future research should confirm these suspected links and further elucidate how specific symptoms differentially influence sexual desires and behaviors, which could help health care professionals meet the information needs of their patients.

It is also notable that the predicted probability of intercourse was never below .40, indicating that many women did not stop having intercourse entirely, but continued engaging in sexual intercourse throughout their entire pregnancy, at least to some degree. It is well-documented that pregnancy symptoms vary widely for individual women, with some experiencing mild symptoms with minimal effect on their daily functioning, while others experience more severe symptoms that can cause substantial disruptions to their lives.³ Understanding this variability is critical for furthering our understanding of changes in sexual functioning across pregnancy. In accordance with this, the random slope models (Models 2-5), which allowed for variation in trajectories across individual women, fit the data better than the model that did not allow for this variability (Model 1) across all fit indices and likelihood ratio tests (when applicable). Indeed, these analytical models are based on averages, and our results are therefore not intended to prescribe exactly when pregnant people should expect changes in their sexual experiences. Rather, the findings provide evidence that changes may occur within smaller timeframes and to larger degrees than has been documented before.

In addition to pregnancy symptomology, sociocultural factors such as religion and superstitions are other aspects of the social context that could affect the probability of intercourse during pregnancy. For example, many women and couples report hesitance to engage in sexual intercourse due to fears of harming the baby’s well-being, and this fear may grow as the due date nears (for review, see ⁴). Assessing these factors was beyond the scope of the current study. Future research should address how pregnancy symptomology, cultural factors, and the social context jointly predict the probability of sexual intercourse during pregnancy. Additionally, it is not known whether women with pregnancy outcomes such as miscarriages or ectopic pregnancies experienced symptoms that would differentially affect their sexual intercourse behavior. Our sensitivity analysis indicates these did not alter the results; however, future research should investigate these potential links more directly. The study was also not able to account for the women’s intentions to conceive, which may play a relevant role. However, using a detailed assessment of the women’s subsequent sexual behavior and ratings of pregnancy desires, we were able to rule out the possibility that the steep initial decline in sexual intercourse frequency was due to some women engaging in sexual intercourse solely to conceive, and therefore completely discontinued sexual activity once pregnant.

Sexual desire and frequency fluctuate over time with age and within intimate relationships as they endure.^27–29 The RDSL dataset only included women during the transition to adulthood, ages 18 through 22, who tend to have shorter and less serious intimate relationships than more mature women. In addition, the younger age range may correspond to more frequent sexual intercourse. Thus, our results may not be generalizable to older women in very long-term relationships. Relatedly, relationship status was a strong predictor of sexual frequency, consistent with previous studies, ³⁰ and indicating that stage of pregnancy is an important predictor of sexual frequency above and beyond the woman’s relationship status. Moreover, while the random slope models and woman-level control variables accounted for many relevant individual differences such as African American identity, directly examining individual differences in the patterns was beyond the scope of the current study; this is likely an important path for future research. Sexual orientation is also a characteristic relevant to the study of partnered sexual behavior. The majority of the women identified as heterosexual or “straight”. However, a large minority did not provide this information, some identified as non-exclusively heterosexual, and, perhaps more importantly, sexual orientation is a dynamic and evolving aspect of many women’s identities.³¹ Future research should include more detailed analysis of sexual identity as it relates to sexual behavior during pregnancy.

There are numerous ways for partners to connect sexually; the current study assessed only penile-vaginal intercourse. Previous studies of specific sexual behaviors throughout pregnancy indicated that the frequency of other activities such as oral sex and masturbation may not decline as rapidly as frequency of sexual intercourse,^1,32,33 potentially due to the particular discomfort and/or physical demands of sexual intercourse. Oral sex and manual stimulation, for instance, offer partners more configural flexibility, making it easier for the pregnant partner to find a comfortable position. Documenting the course of a fuller range of sexual behaviors over pregnancy, and directly connecting these behaviors to pregnancy symptoms as well as to individual and relational well-being, is an important next step in understanding how sexuality is affected by pregnancy, and offering potential solutions for how couples can cope with likely impending sexual disruptions during pregnancy.

CONCLUSION

This intensive longitudinal study is the first to document trajectories of women’s sexual intercourse frequency across all weeks of pregnancy—as they naturally occurred among a representative sample of young women. We found more complex patterns than those reported in previous studies, particularly steep declines and subsequent increases early in pregnancy, suggesting first-trimester fluctuations in sexual frequency may be more extreme than previously documented. Changes in weekly intercourse frequency generally followed patterns of common pregnancy symptomology such as nausea and fatigue, suggesting pregnancy symptoms may be most favorable to intercourse towards the end of the first and beginning of the second trimesters, and least favorable near the end of the pregnancy. For those trying to maintain sexual intimacy during pregnancy, this detailed assessment provides critical information that could help prepare for such potential sexual disruptions; women struggling with declines in sexual intercourse while pregnant or their healthcare providers could consider targeting pregnancy symptoms such as nausea, suggesting alternative forms of sexual and intimate connection, or by managing expectations around sexual intimacy during pregnancy.