Hormonal Contraception and Sexuality: Causal Effects, Unobserved Selection, or Reverse Causality?Laura J. Botzet, Tanja M. Gerlach, Julie C. Driebe, Lars Penke, Ruben C. Arslan. Collabra: Psychology (2021) 7 (1): 29039. Oct 21 2021. https://doi.org/10.1525/collabra.29039
Many of the women who take hormonal contraceptives discontinue because of unwanted side effects, including negative psychological effects. Yet scientific evidence of psychological effects is mixed, partly because causal claims are often based on correlational data. In correlational studies, possible causal effects can be difficult to separate from selection effects, attrition effects, and reverse causality. Contraceptive use and, according to the congruency hypothesis, congruent contraceptive use (whether a woman’s current use/non-use of a hormonal contraceptive is congruent with her use/non-use at the time of meeting her partner) have both been thought to influence relationship quality and sexual functioning. In order to address potential issues of observed and unobserved selection effects in correlational data, we studied a sample of up to 1,179 women to investigate potential effects of contraceptive use and congruent contraceptive use on several measures of relationship quality and sexual functioning: perceived partner attractiveness, relationship satisfaction, sexual satisfaction, and diary measurements including libido, frequency of vaginal intercourse, and frequency of masturbation. No evidence for substantial effects was found except for a positive effect of hormonal contraceptives on frequency of vaginal intercourse and a negative effect of hormonal contraceptives on frequency of masturbation. These effects were robust to the inclusion of observed confounders, and their sensitivity to unobserved confounders was estimated. No support for the congruency hypothesis was found. Our correlational study was able to disentangle, to some extent, causal effects of hormonal contraceptives from selection effects by estimating the sensitivity of reported effects. To reconcile experimental and observational evidence on hormonal contraceptives, future research should scrutinize the role of unobserved selection effects, attrition effects, and reverse causality.
Keywords: sexual frequency, sexuality, relationship quality, sex hormones, contraception
Discussion
Our study aimed to disentangle selection effects from causal effects of contraceptive use. It showed that additional selection effects (including information about demography and personality) did not describe the choice of contraceptive method and congruent contraceptive use substantially better than did selection effects of age, income, and relationship duration. Furthermore, there was no evidence for substantial effects of contraceptive method, congruent contraceptive use, and their interaction on perceived partner attractiveness, relationship satisfaction, sexual satisfaction, and libido. While congruent contraceptive use and its interaction with contraceptive use had no substantial effects on frequency of vaginal intercourse and frequency of masturbation, we found a positive effect of current use of hormonal contraceptives on frequency of vaginal intercourse and a negative effect of current use of hormonal contraceptives on frequency of masturbation. These links were robust to the inclusion of observed confounders and sensitivity analyses suggested that unobserved confounders would need to strongly influence outcomes (about as strong as all observed confounders taken together) in order to substantially alter conclusions.
Selection Effects
Including additional selection variables pertaining to demography (education) and personality (openness, conscientiousness, extraversion, agreeableness, neuroticism, and religiosity) did not substantially improve models predicting contraceptive method or congruent contraceptive use compared to models based only on age, relationship duration, and income. Therefore, Hypothesis 1—that the complex model explains more variance compared to the simple model in (1) the choice of contraception and (2) congruent contraceptive use—was rejected.
Of the predictors included in the simpler models, age and relationship duration had a significant effect on choice of contraceptive method and congruent contraceptive use; income was no significant predictor. Overall, age had a negative effect on the use of hormonal contraceptives, i.e. the percentage of women using hormonal contraceptives decreased with increasing age. In addition, age had a negative effect on congruency in women who were using hormonal contraceptives when they met their partner (i.e., older women were more likely to switch to no/nonhormonal contraceptives) but a positive effect on congruency in women who were using no/nonhormonal contraceptives when they met their partner (i.e., younger women were more likely to switch to hormonal contraceptives). Overall, women in a romantic relationship were more likely to use hormonal contraceptives. Relationship length played no significant role in choice of contraceptive use, but partnered women who were in longer relationships were more likely to switch contraceptive methods, independent of whether they had been using hormonal contraceptives or no/nonhormonal contraceptives when they met their partner. Even though the complex models showed no improvement in model fit over the simple models, three predictors in the complex models stood out: First, conscientiousness had a positive effect on hormonal contraceptive use and a positive effect on congruent contraceptive use in women who had been using hormonal contraceptives when they met their partner (i.e., they were more likely to continue using hormonal contraceptives). Second, openness had a negative effect on hormonal contraceptive use and a negative effect on congruent contraceptive use in women who had been using hormonal contraceptives when they met their partner (i.e., they were more likely to switch to no/nonhormonal contraceptives). Third, agreeableness had a positive effect on congruent contraceptive use in women who had been using hormonal contraceptives when they met their partner (i.e., they were more likely to continue using hormonal contraceptives) and a negative effect on congruent contraceptive use in women who had been using no/nonhormonal contraceptives when they met their partner (i.e., they were more likely to switch to hormonal contraceptives). Future research concerning selection effects on contraceptive use and congruent contraceptive use could consider excluding measures of income (where appropriate6) and including measures of conscientiousness, openness, and agreeableness in addition to age and relationship duration.
Effects of Hormonal Contraceptives
The evidence for effects of hormonal contraceptives is inconclusive on perceived partner attractiveness, relationship satisfaction, sexual satisfaction, and libido, therefore neither accepting nor rejecting Hypothesis 2.1(1–4)—hormonal contraceptives lead to decreased (1) perceived partner attractiveness, (2) relationship satisfaction, (3) sexual satisfaction, and (4) libido—and Hypothesis 2.2(1–4)—after controlling for all selection variables, hormonal contraceptives lead to decreased (1) perceived partner attractiveness, (2) relationship satisfaction, (3) sexual satisfaction, and (4) libido. The estimates were insufficiently precise; future research with even larger samples is needed to reach a conclusion. Nevertheless, given the rather small effect sizes, it appears unlikely that use of hormonal contraceptives has a strong association with these outcomes.
Hormonal contraceptives had a positive effect on frequency of vaginal intercourse, even after controlling for observed confounders—thereby rejecting Hypotheses 2.1(5) and 2.2(5). Contrary to the RCTs by Graham et al. (1995) and Zethraeus et al. (2016) that provided evidence for negative effects of hormonal contraceptives on sexual desire, sexual arousal, and sexual pleasure, the results of our study are in line with studies based on correlational data that found a positive relationship between hormonal contraceptives and sexual frequency (Alexander et al., 1990; Caruso et al., 2005; McCoy & Matyas, 1996).
Hormonal contraceptives had a negative effect on frequency of masturbation, even after controlling for observed confounders—thereby accepting Hypotheses 2.1(6) and 2.2(6). Most studies show no difference in frequency of masturbation between HC users and non-HC users (Alexander et al., 1990; Bancroft et al., 1991), but a recent study by Mark et al. (2016) provided evidence of a positive association between hormonal contraceptives and women’s dyadic libido and a negative association between hormonal contraceptives and women’s solitary libido. The libido item in our study included dyadic and solitary libido (“I experienced increased libido [desire to have sexual intercourse/to masturbate/to be sexually active].”) and did not distinguish between them as proposed by Spector et al. (1996). Thus, it seems possible that the divergent relationships described by Mark et al. (2016) resulted in the overall null relationship between hormonal contraceptives and libido that we observed. Our study could therefore provide evidence for behavioral consequences (measured as frequency of vaginal intercourse and frequency of masturbation) of the divergent relationships between hormonal contraceptives and dyadic and solitary libido described by Mark et al. (2016).
Effects of Congruent Contraceptive Use
Evidence was inconclusive on effects of congruent contraceptive use on perceived partner attractiveness, relationship satisfaction, sexual satisfaction, libido, frequency of vaginal intercourse, and frequency of masturbation after considering observed confounders. We could therefore neither accept nor reject Hypotheses 3.1—congruent contraceptive use leads to increased (1) perceived partner attractiveness, (2) relationship satisfaction, (3) sexual satisfaction, (4) libido, (5) frequency of vaginal intercourse, and (6) frequency of masturbation—and 3.2—after controlling for all selection variables, congruent contraceptive use leads to increased (1) perceived partner attractiveness, (2) relationship satisfaction, (3) sexual satisfaction, (4) libido, (5) frequency of vaginal intercourse, and (6) frequency of masturbation.
While these findings do not support most of the literature based on the congruency hypothesis (Birnbaum et al., 2019; Cobey et al., 2013; French & Meltzer, 2020; Roberts, Cobey, et al., 2014; Roberts et al., 2012; Roberts, Little, et al., 2014; Russell et al., 2014), they are in line with a recent large-scale replication attempt by Jern et al. (2018). Marcinkowska et al. (2019) provided additional evidence that questions the congruency hypothesis: In a large-scale study (n = 6,482), they found no evidence that women using the pill had weaker preferences for male facial masculinity than did women not using the pill. Differences in partner preferences have been suggested to be the driving mechanism behind the congruency hypothesis: Incongruent contraceptive methods are thought to lead to a shift in partner preferences, resulting in less satisfaction with the current romantic partner.
Jern et al. (2018) pointed out an important difference between their study and earlier studies on the congruency hypothesis: Earlier studies often had unequal distributions of congruent and incongruent users such that one group was almost entirely based on HC users or non-HC users. For instance, in the study by Cobey et al. (2013) the group of incongruent current HC users consisted of only four participants (3% of the final sample), while the group of congruent HC users consisted of 71 participants (59% of the final sample). This is especially problematic because most studies featured relatively small incongruent HC user groups and relatively large congruent HC user groups. Considering the small expected effect sizes based on the congruency hypothesis, main effects of current contraceptive use could have led to a spurious effect of congruency based on unequal distributions. Although the sample sizes of congruent and incongruent use differed in our study, the subgroups were relatively balanced (congruent non-HC users: 32%; congruent HC users: 31%; incongruent non-HC users: 19%; incongruent HC users: 17%) and the models always accounted for current contraceptive use and its interaction with contraceptive congruency.
Another possibility is that the reported effects based on the congruency hypothesis were false positives (Simmons et al., 2011). None of the earlier studies used preregistered hypotheses, and sample sizes were relatively small (ranging between n = 48 and n = 365), apart from the study by Roberts et al. (2012), which found a positive effect of congruent use on perceived partner attractiveness and sexual satisfaction (n = 993). Indeed, two failed large-scale replication attempts (Jern et al., 2018 and the current study) and a range of recent evidence question the theory underlying the congruency hypothesis (Arslan et al., 2018; C. R. Harris et al., 2013; Jones, Hahn, & DeBruine, 2018; Jünger, Kordsmeyer, et al., 2018; Jünger, Motta-Mena, et al., 2018; Stern et al., 2020; Stern & Penke, in press; Wood et al., 2014; for a current discussion on evidence for psychological cycle shifts see Gangestad et al., 2019a, 2019b; Higham, 2019; Jones, Hahn, & DeBruine, 2018; Roney, 2019; Stern et al., 2019).
Overall, recent work has cast doubt on the evidence for both the assumed mechanism and the interaction effect underlying the congruency hypothesis. Our study could not accept the null hypotheses that there are no effects of congruent contraceptive use on perceived partner attractiveness, relationship satisfaction, sexual satisfaction, libido, frequency of vaginal intercourse, and frequency of masturbation because the sample size was too small and because we applied rigorous decision criteria for accepting the null hypotheses. Future research on congruent contraceptive use should be preregistered, be adequately powered to detect small effects, and appropriately account for current and past contraceptive use.
Sensitivity to Unobserved Confounders
We estimated the robustness of the effects of hormonal contraceptives and congruent contraceptive use in light of potential unobserved confounders. Sensitivity analysis suggested that the influence of unobserved confounders would need to be nearly 1.5 times as strong as the influence of observed confounders to fully account for the effect of hormonal contraceptives on frequency of vaginal intercourse, and nearly as strong as the influence of observed confounders to fully account for the effect of hormonal contraceptives on frequency of masturbation. Even when taking into account the broad range of included observed confounders (demography, personality, and romantic relationship information) it seems plausible that unobserved confounders might exist that would fully explain the reported effects of hormonal contraceptives on frequency of vaginal intercourse and frequency of masturbation. Besides potential unobserved selection effects, we now consider three additional possible challenges: reverse causality, attrition effects, and further unobserved confounders.
Selection Effects and Reverse Causality
There was a positive effect of hormonal contraceptive use on frequency of vaginal intercourse. Although frequency of vaginal intercourse was measured after contraception in the diary, it is somewhat habitual and thus stable. Reverse causality might therefore be at play, even after excluding women who were not sexually active and therefore not using hormonal contraception. Women who have sex more frequently might place a larger premium on safeness and convenience for contraception. Higher frequency of vaginal intercourse is associated with a higher risk of (unwanted) pregnancy, and therefore safe contraception is even more important, especially if a woman does not want to forego sexual intercourse or use additional contraceptive methods. In addition, higher frequency of vaginal intercourse affects economic considerations: At higher sexual frequencies, the pill can be cheaper than condoms. Reverse causation would explain why there are effects on behavior (frequency of vaginal intercourse) but not on the psychological outcomes that might be expected to precede the behavior in the causal chain (libido and sexual satisfaction). A similar, if slightly more speculative, explanation could be plausible for frequency of masturbation. If women who have sexual intercourse only infrequently eschew the pill and its cost and side effects, they might instead opt to use condoms. If these women have the same level of libido as women who have sex more frequently, they might masturbate more. Because the stable component of some of these outcomes could be quite large, these are plausible unobserved confounders, and repeated longitudinal data would be needed to adjust for them.
Attrition Effects
Unlike our study, RCTs reported negative effects of hormonal contraceptives on libido, sexual arousal, and sexual pleasure (Graham et al., 1995; Zethraeus et al., 2016) as well as on general well-being (Zethraeus et al., 2017). One potential reason for the positive effect of hormonal contraceptives on frequency of vaginal intercourse reported in our study and in earlier correlational studies (Alexander et al., 1990; Caruso et al., 2005; McCoy & Matyas, 1996) are attrition effects. As Graham (2019) noted, there is great variability in women’s experiences with hormonal contraceptives, with reports of negative, positive, and no effects. Women with negative experiences were more likely to stop using hormonal contraceptives (Bancroft & Sartorius, 1990; Sanders et al., 2001), and discontinuation rates are high. For instance, 11.6% of Swedish women who took hormonal contraceptives for the first time stopped using them within six months (Josefsson et al., 2013). Predictors of discontinuation include emotional side effects, worsening of the premenstrual syndrome, decreased frequency of sexual thoughts, and decreased psychosexual arousability (Sanders et al., 2001). Women with depressive and premenstrual complaints tend to discontinue hormonal contraceptive use, leaving the remaining users with greater reported well-being (Bancroft & Sartorius, 1990). Therefore, it is likely that empirical, correlative evidence suggesting positive effects of hormonal contraceptives on sexual functioning stems at least in part from the fact that women with negative experiences of hormonal contraceptives switch to other contraceptive methods. Much of the current evidence on positive effects of hormonal contraceptives might thus rest on samples skewed toward women who have already tailored their contraception regimen to their experiences with hormonal contraceptives; reported correlations could even be the reverse of the average causal effect. This may also be the case in our study: The estimated effect of hormonal contraceptives may not equal their average effect because women with negative experiences of hormonal contraceptives had already stopped taking them. Women who continued using hormonal contraceptives would be more likely to have had positive experiences with them, which would result in an overall positive relationship between hormonal contraceptive use and frequency of vaginal intercourse based on correlational data masking negative causal effects on average. Analyses on the congruency of contraceptive use at the time of meeting the current partner can only partly address this, as participants were generally old enough to have been able to try out different methods of contraception before meeting their current partner. This limitation implies that estimated effects should not be expected to generalize to the experiences of women using hormonal contraceptives for the first time. Attrition effects could be studied using longitudinal data. In addition, research on women using hormonal contraceptives for the first time could provide more information on how preferences for contraceptives form.
Further Unobserved Confounders
Besides the already included selection and outcome variables, frequency of vaginal intercourse has been found to relate to less restricted sociosexuality (Grøntvedt et al., 2020), increased satisfaction with own body image (Ackard et al., 2000), and increased satisfaction with life (Muise et al., 2016) in women. Frequency of masturbation has been found to be positively associated with less restricted sociosexuality (Velten & Margraf, 2017), body acceptance and orgasm frequency (Burri & Carvalheira, 2019), and greater importance of sex and higher levels of general anxiety and depression (Rowland et al., 2020) in women. Regnerus et al. (2017) reported a negative relationship between frequency of masturbation and contentment with sexual frequency.
While some of these potential unobserved confounders were not measured in the available dataset (in particular orgasm frequency, importance of sex, anxiety, depression, and contentment with sexual frequency), others would have been available (in particular sociosexuality, general life satisfaction, and satisfaction with own body image) but we decided not to include them in the current study to prevent controlling for potential colliders or mediators (Rohrer, 2018). Nevertheless, they could be strong unobserved confounders that could explain the observed links between hormonal contraceptive use and frequency of vaginal intercourse and frequency of masturbation. For example, higher desire for penetrative intercourse could lead to higher frequency of vaginal intercourse. At the same time, it could lead to the decision to use hormonal contraceptives because they are among the safest contraceptive methods available. Body acceptance could lead to higher frequency of masturbation and, at the same time, to the decision to use no/nonhormonal contraceptives in order to avoid artificial hormones. Therefore, even though our study quantitatively estimated the needed strength of unobserved confounders, it is unable to definitively rule out the possibility that the observed relationships are due to the influence of potential unobserved confounders. In future research on larger samples, more pointed comparisons of contraceptives with similar Pearl indices indicating the effectiveness of this birth control (e.g., the pill and intrauterine devices) might answer some of these questions, and again, examining within-subject changes in sexuality in longitudinal data would reduce some of the concerns about potential unobserved confounders.
Causal Effects of Hormonal Contraceptives on Frequency of Vaginal Intercourse and Frequency of Masturbation?
Our study provides evidence for a positive effect of hormonal contraceptives on frequency of vaginal intercourse and a negative effect of hormonal contraceptives on frequency of masturbation. Both effects were somewhat attenuated when adjusting for observed confounders. Fairly strong unobserved confounders would be necessary to nullify or reverse the remaining effects but some plausible candidates exist. The questions of reverse causality, selection effects, and attrition effects regarding the reported effects persist.
Limitations and Strengths
Our study was not without limitations: First, while the total sample size of our study was relatively large, the sample sizes used for the analyses—especially those investigating effects of congruent contraceptive use—were too small to reach a definite conclusion about potential effects. Second, even though our study provides valuable insights into the links between contraceptive use, relationship quality, and sexual functioning, the conclusions based on correlational data remain inconclusive about any putative causal effect of hormonal contraceptives.
Nevertheless, our study had several key strengths: First, even though conclusions remain cautious because of the available sample size, the size of the sample still exceeds domain standards. Second, measures for libido, frequency of vaginal intercourse, and frequency of masturbation were based on diary reports, which have been shown to be more reliable than retrospective behavior measurements (McAuliffe et al., 2007) and described as the gold standard for measuring sexual frequency by Graham et al. (2003). Third, by providing a directed acyclic graph, controlling for observed confounding, and estimating the sensitivity to unobserved confounding, this study is better positioned than previous correlational work to disentangle selection effects from causal effects.
Constraints on Generality
Following the guidelines on constraints on generality (Simons et al., 2017) the following four factors reduce the broad generalizability of the current results: First, the sample studied consisted of heterosexual WEIRD (Henrich et al., 2010) women with a high proportion of undergraduate psychology students. Although our main explanation assumes a universal biological mechanism for the results, different absolute hormone levels in less prosperous and well-nourished populations (Vitzthum, 2009) might affect the frequency of ovulation and hence the observable effect size. Moreover, some of our alternative explanations (such as reverse causality) may be much more dependent on circumstances. Second, even though the gold standard for measuring sexual frequency and masturbation frequency by using diary reports was applied, all results are purely based on self-reports and generalizability to other measures might be limited. Third, from a temporal perspective the composition of hormonal contraceptives (e.g., the dosage of estrogen and progestin) has changed over the decades and will change further in the future. If the reported effects are due to certain estrogen or progestin dosages, results might not be replicable based on samples from different times.
[...]
Posts
No comments:
Post a Comment