Sex and Hormonal Effects on Drug Cue Reactivity and Its Regulation in Human Addiction.

Forty-nine men with HUD (age=41.96±9.71) and 32 age-matched women with HUD or CUD [iHUD: n=16; individuals with CUD (iCUD): n=16; age=38.85±9.84] from the greater New York City area were recruited. Participants underwent a comprehensive clinical diagnostic interview conducted by trained research staff under a clinical psychologist’s supervision. All iHUD and iCUD met DSM-5 criteria for opioid or stimulant use disorder (with heroin or cocaine as the primary drug of choice or reason for treatment, respectively). See supplement for details of interview procedures, exclusionary criteria, comorbidities, and drug use/treatment related information. All participants were scanned with MRI once to examine sex differences. A subgroup of sixteen women (iCUD=13, iHUD=3) underwent an additional MRI procedure to examine menstrual cycle/hormone effects (days between scans=47.80±61.08). Menstrual phase information for both MRI days (follicular vs. luteal) was tracked via self-report and validated by objective blood ovarian hormone testing using chemiluminescence microparticle (Abbott Architect) or electrochemiluminescence (ECLIA, Roche) immunoassays (see Supplement ). All 16 women had a regular menstrual cycle (past 3 months average days=30.52±2.21) and none used hormonal contraceptives (self-reported). They were randomized to have their first scan at the follicular (n=6) or luteal (n=10) phase, with no significant order effect between women who started their first scan in the follicular vs. those started in the luteal phase (p=0.289). The study was approved by the IRB of the Icahn School of Medicine at Mount Sinai. All participants provided written informed consent. Throughout the manuscript, the terms “men” and “women” are used to describe self-reported biological sex. The task paradigm was reported in detail in our previous study.( 29 ) In short, participants were instructed to passively look at drug (images of drug preparation, use, and paraphernalia; heroin for iHUD, and cocaine for iCUD), food, and neutral images, actively downregulate their emotional reactivity to the drug images, and actively upregulate their emotional reactivity to food images, during the “look”, “reappraise”, and “savor” conditions, respectively. Immediately before and after the task, participants provided self-reported drug and food craving and task motivation ratings. Self-evaluations of the difficulty and effectiveness of reappraisal and savoring were also collected after the task. After the MRI session, each participant provided ratings on valence, arousal, and craving (cue-induced craving; for food and drug cues only) on half of the drug, food, and neutral images viewed during the task (see Supplement ). Individual parameter estimates were generated using the general linear model (GLM) via FMRIB Software Library (FSL)’s FEAT (version 6.0).( 36 ) See Supplement and previous study( 29 ) for GLM details. A fixed-effects model was used for subject-level statistical maps of each task event (look drug/food/neutral, reappraise drug, savor food) and their contrasts to yield estimates of drug cue-reactivity (look drug>look neutral or look food) and its regulation via reappraisal (reappraise drug>look drug) and savoring (savor food>look drug). For completeness, we also inspected the contrasts of reappraise drug>savor food and savor food>look food; the savoring results are reported in the Supplement . Group-level estimates were calculated using FSL FLAME 1+2 mixed-effects model to improve group-level variance estimation and population inferences via Markov chain Monte Carlo simulations.( 37 ) Given the nonuniformity in number of cue types per task condition, instead of calculating condition-related main effects and interactions, guided by previous study( 29 ), we preselected the above-mentioned four contrasts of interest to test our hypotheses. For within-women (delta measures between menstrual phases) correlations, the delta directions are follicular minus luteal phase for correlations with estradiol and drug cravings and luteal minus follicular phase for correlations with progesterone and scaled progesterone/estradiol ratio for better interpretation. These same delta directions were used for the BOLD fMRI data. These correlations were conducted in a within-subject’s approach (comparing the menstrual phases) to reduces between-subject variance, making it particularly suitable for the small sample size in this project. All analyses were conducted voxel-wise using a cluster-defining threshold of Z>3.1.( 38 ) For multiple comparisons correction, analyses of the two drug cue-reactivity contrasts used cluster-extent thresholds corrected to p<0.05/2=0.025 for sex and menstrual phase differences separately, and to p<0.05/(2×2)=0.0125 for correlations with cue- and task-induced (post- minus pre-task) drug craving. Analyses pertinent to the reappraisal (and savoring) contrasts and correlations with ovarian hormones were more exploratory and therefore corrected to a cluster-extent threshold of p<0.05. In addition to whole-brain analyses, we used an independent anatomical mask (including the entire frontal cortex and striatum) for a restricted search in our regions of interest guided by our previous studies( 15 , 29 ) (related results are labeled with “anatomically masked”; see Supplement for additional details of the anatomical mask, neuroimaging data acquisition and preprocessing).

Table 1 illustrates significant sex differences (women>men) in past 30-day use of heroin/cocaine (p=0.005); this effect was not correlated with any of the neuroimaging outcomes of interest and thus was not included as a covariate. No significant sex differences were observed for any of the other drug use or alcohol and smoking variables, demographics, and neuropsychological measures. Group differences were also not observed in any drug use variables between iHUD and iCUD in women. Menstrual phase differences within the subgroup of women who underwent two scans were also not observed for dynamic drug use variables (i.e., past 30 days use of drug and days since last drug use). Therefore, no variables reported in Table 1 were included as covariates in our neuroimaging analysis. See supplement for a general breakdown of other drug co-morbidities history. There were no significant interaction effects with sex or menstrual phase in any of the ratings. See Supplement for details ( Figure S1 – S4 ). Whole-brain analyses revealed greater drug cue-reactivity (>look neutral or food) in the medial PFC (mPFC) in women compared to men ( Figure 1A ). Compared to men, women also showed lower drug reappraisal (>look drug) in the left frontal eye field (FEF)/dlPFC (anatomically masked; Figure 1B ). The higher the left dlPFC drug reappraisal, the lower the cue-induced drug craving only in men (anatomically masked; non-significant in women; significant sex differences in slopes via extracted parameter estimates [p=0.042]; Figure 1C ). See Table 2 for details. For completeness, the same analyses were conducted separately within iHUD (men vs. women) and women (iHUD vs. iCUD). Similar sex differences in cortico-striatal activity were observed at a trend level within the smaller sample of iHUD only. There was also no significant group (SUD type within women only) differences that overlapped with these results (see Supplement also for menstrual phase locked sex differences analyses). Compared to the late-follicular phase, the mid-luteal phase was characterized by a significantly higher progesterone level (p=0.008), demonstrating the expected progesterone domination (higher scaled progesterone/estradiol ratio: p=0.004). While higher averaged estradiol was observed in the late-follicular phase, there were no significant differences when directly compared to the mid-luteal phase (p=0.56, Figure S5 ) attributed to the second peak of estradiol level in the menstrual cycle.( 17 , 39 ) Exploratory correlations showed a trend for Δ (changes between the menstrual phases) estradiol level to be positively correlated with Δtask-induced drug craving, such that the higher the follicular>luteal estradiol, the more the respective drug craving (r=0.48, p=0.08). This finding aligns with other studies showing that estradiol enhances drug-seeking behavior.( 2 , 9 ) See supplements for the exploratory hormonal correlations with food cravings. Compared to the luteal phase, there was higher FEF/dlPFC drug cue-reactivity (>look food) in the follicular phase (anatomically masked; Figure 2A ). These Δ (follicular>luteal contrast) drug cue-reactivity in the IFG and vmPFC was positively correlated with the respective Δcue-induced and Δtask-induced drug craving ( Figure 2B & 2C ). As compared to the follicular phase, the luteal phase was instead characterized by higher drug reappraisal (>look drug) in the anterior PFC/orbitofrontal cortex (aPFC; Figure 3A ) in addition to other areas ( Table 3 ; see Supplement ). Fluctuations in ovarian hormones were directly correlated with these changes in cortico-striatal drug cue-reactivity and drug reappraisal. Specifically, for the follicular>luteal contrast, Δestradiol was correlated positively with Δdrug cue-reactivity (>look food) in the vmPFC (anatomically masked; Figure 2D ) and negatively with Δdrug reappraisal (>look drug), again, primarily in default mode network regions inclusive of the vmPFC but also in the FEF/aPFC (anatomically masked; Figure 3B ) and striatum (including the caudate and putamen; Figure S6B ) and other areas ( Table 3 ). In contrast, for the luteal>follicular contrast, the Δprogesterone/estradiol ratio was positively correlated with Δdrug reappraisal (>look drug) in the dlPFC ( Figure 3C ) and other regions ( Table 3 ; see Supplement ). Consistent with these results, correlations with absolute hormonal values in the follicular phase showed a positive correlation between estradiol and drug cue reactivity (look drug>look food) in the dlPFC and a negative correlation with drug reappraisal (>look drug) in the dlPFC, striatum, vmPFC and other areas (anatomically masked). See supplement ( Table S1 and S2 for savoring contrasts; also Table 2 and 3 for contrasts reported in the main text) for results that are outside of our main goals and/or regions of interest (i.e., frontal cortico-striatal areas) for completeness. Also see Figures S7 – S9 for summarized (whole-brain and anatomically masked) cortical BOLD fMRI activation for menstrual cycle and hormonal effects.

We first demonstrated elevated mPFC drug cue-reactivity in women and higher FEF/dlPFC drug reappraisal as correlated with lower cue-induced drug craving in men. Additionally, significant menstrual cycle effects were found, where drug cue-reactivity was higher in the FEF/dlPFC during the follicular phase and drug reappraisal was higher in the aPFC/orbitofrontal cortex during the luteal phase. Within-subject correlations on changes between menstrual phases (follicular>luteal) showed that ΔvmPFC drug cue-reactivity correlated with both higher respective Δestradiol and Δtask-induced drug craving; a similar correlation was observed for the IFG. Conversely, Δdrug reappraisal (luteal>follicular) correlated with lower respective Δestradiol in the vmPFC, FEF/aPFC, and striatum and higher Δprogesterone/estradiol ratio in the dlPFC. Consistent with findings in people with problem alcohol use( 21 , 22 ) but opposite to a study in iCUD( 40 , 41 ), we found higher mPFC drug cue-reactivity in women with HUD or CUD compared to men with HUD; differences from the previous studies could potentially be attributed to different task stimuli used (imaginary scripts vs. pictures) and the regulation conditions in our task. Consistent with studies in healthy individuals ( 30 , 31 ), we observed higher drug reappraisal in the FEF/dlPFC, further linking this activity to lower cue-induced drug craving in the HUD men. This result aligns with the general role of the dlPFC in top-down cognitive control [e.g., during emotion regulation( 42 , 43 )] and its specific role in craving regulation [e.g., in current smokers( 26 )]. The absence of a similar correlation in the SUD women further suggests a potentially sexually dimorphic mechanism, where men may better mobilize top-down control resources during cognitive reappraisal to reduce drug cue-reactivity and the associated experience of craving, while in women absence of these effects may suggest more susceptibility/vulnerability specifically during drug cue exposure. These observed sex differences effects may be influenced by gonadal hormones. Indeed, enhanced drug cue-reactivity was observed during the follicular>luteal phase (in FEF/dlPFC) as associated with increased respective Δdrug craving (in IFG and vmPFC) and Δestradiol (in vmPFC). The luteal>follicular phase instead showed enhanced drug reappraisal (in aPFC) as associated with reduced respective Δestradiol [in FEF/aPFC and vmPFC and striatum (encompassing putamen and caudate), potentially attributed to estrogen’s positive modulation effects on the dopamine system, specific to females( 44 , 45 )] and increased Δprogesterone/estradiol ratio (in dlPFC). These menstrual phase differences in both drug cue-reactivity (follicular>luteal) and reappraisal (luteal>follicular) as associated with the respective estradiol and progesterone/estradioal changes (and with craving changes for the former), are notable for the following. They suggest a unique role for the vmPFC where this default mode network region may be a menstrual/hormone-modulated neural marker both for addiction vulnerability/risk (cue reactivity, craving, and estradiol) and resilience (reappraisal and progesterone) in women. Therefore, the vmPFC’s role in subjective valuation( 46 ) and the generation of affective meaning( 47 ) during drug cue exposure and its regulation could be targeted (for its attenuation in the former and enhancement in the latter) in future intervention efforts. Results also highlight executive and cognitive control regions( 48 ) (e.g., the dlPFC, a region within the frontoparietal control network) in reappraisal as associated with progesterone (luteal>follicular), consistent with progesterone’s protective effects in the rodent model of addiction ( 49 , 50 ). These anterior lateral PFC regions’ role in emotion regulation( 42 , 43 , 51 ) could therefore be targeted for enhancement [e.g., with transcranial direct current stimulation( 52 )]. Abnormalities within and between the default mode network and attention and executive control networks have also been associated with facilitating craving and relapse in SUD( 53 ), and hence the dynamic co-activation of these networks could be a further treatment target. Some limitations need to be acknowledged and future directions discussed: 1) The current study lacked a group of men with CUD, preventing a direct comparison between substances and potentially contributing to the opposite drug cue-reactivity sex differences results when compared to previous studies in CUD. However, using an fMRI task we adapted to be relevant to both iCUD and iHUD, our results suggest some generalizability across substances (note also stability of results when excluding participants who are comorbid for both HUD and CUD, see Supplement ). Nevertheless, future efforts with a larger sample size should fill this gap. A larger sample size would also allow for inspection of the brain-hormonal correlations within a menstrual phase (as absolute and not delta values) and of the potential effects on results of various confounds (e.g., other substance use and psychiatric comorbidities). Adding a group of healthy control women (and/or a second scan in men) for reference of the hormonal effects would also increase the validity of the current findings; 2) Our study targeted the mid-follicular and mid-luteal phases (when estradiol is relatively high, Figure S5 ), potentially contributing to some negative results. Future studies should therefore assess the additional early-follicular phase (low estradiol and progesterone), especially as the strongest ventral striatal drug cue-reactivity was observed when comparing this phase to the late-follicular and mid-luteal phases in nicotine-dependent women( 17 ); 3) Stress-induced drug cue-reactivity and fluctuations in the stress and reward brain circuits’ function during the menstrual cycle were reported in women with CUD, linked to fluctuations in cortisol in addition to ovarian hormones.( 40 , 54 ) Therefore, measuring the cortisol hormone could be informative for future research; 4) Long-term exposure to opioids or opioid replacement therapy (e.g., methadone and buprenorphine) could lead to hypogonadism (e.g., deficiencies in estradiol and progesterone).( 55 ) Indeed, some of our participants with urine positive for methadone or buprenorphine showed very low levels (below sensitivity level) of progesterone or estradiol, suggesting that this factor is important for future study. We could also not equate the women subgroups on this factor, although we note lack of differences between the CUD and HUD women; 5) For feasibility reasons, the current study relied on self-reported menstrual cycle history to schedule the first MRI. Serum hormone measurements confirmed that most participants were accurately scanned during the targeted phases (late follicular and mid-luteal). Moreover, our delta hormone correlation analyses focused on within-subject hormonal changes in relation to the respective changes in brain activation, thereby reducing between-subject variability. Nonetheless, if feasible, future studies should consider supplementing this approach with objective measures from ovulation testing kits to enhance scheduling precision vis-à-vis the targeted phase; and 6) While no participants in our analyses reported current or past gynecological problems that require medical attention nor recent alterations in menstrual cycle patterns, future studies will benefit from a more extensive medical examination to rule out potential conditions (e.g., endometriosis or polycystic ovary syndrome) that could alter hormonal profiles. To the best of our knowledge, this is the first clinical neuroimaging study simultaneously examining sex differences, menstrual phase, and hormonal effects in cortico-striatal drug cue-reactivity and its regulation in SUD. Although the sex differences reported here remain to be reconciled with the higher overdose mortality rates in men at the epidemiological level,( 10 ) our findings offer valuable and timely insights into the hormone-modulated vulnerability and resilience factors in women with SUD. Specifically, our results could inform translational efforts to develop precisely timed, hormonally informed treatments. For example, we recommend targeting the follicular phase to reduce drug cue-reactivity by enhancing executive control functions (e.g., cognitive reappraisal and/or savoring of an alternative reward, see Supplement ), which may reduce craving and ultimately improve treatment outcomes in women with SUD.

In 2023, opioid- and cocaine-related overdose deaths exceeded 79,000 and 29,000, respectively, taking a substantial toll on both individuals and society.( 1 ) Women have been reported to exhibit greater severity of drug abuse, a faster progression to problematic drug use, and higher relapse risk than men( 2 – 6 ) although these sex differences are not always robust, with opposite results also reported.( 7 – 9 ) For example, even after controlling for overall levels of drug exposure, we recently reported that men have higher opioid, cocaine, and psychostimulant overdose mortality rates than women.( 10 ) And while the age-adjusted rate of drug overdose deaths declined from 2022 to 2023—by 2.9% in men and 5.7% in women—they remain alarmingly high, with 44.3 deaths per 100,000 among men and 18.3 per 100,000 among women.( 1 ) Overall, these discrepancies in sex differences may be partially driven by fluctuations in women in the two primary ovarian hormones,( 2 – 6 ) estradiol and progesterone, which have been suggested to exert opposing effects on addiction-related behaviors: estradiol facilitates the initiation of drug taking( 2 ) and reinstatement of drug-seeking behavior,( 2 , 9 ) while progesterone reduces craving,( 11 ) positive subjective drug effects,( 12 ) and amount( 13 ) of substance use. Yet, the complex interplay of sex- and hormonal-based patterns of risk and resilience in addiction is understudied partly because women remain significantly underrepresented in basic neuroscience (including neuroimaging) studies.( 14 , 15 ) Of particular relevance, merely five neuroimaging studies examined the neural correlates of sex differences in cue-induced reactivity and craving in cocaine use disorder (CUD) and none in heroin use disorder (HUD) as previously reviewed.( 9 ) Furthermore, studies of hormonal effects in women are scarce, particularly in substance use disorder (SUD).( 16 , 17 ) Although these notions have been increasingly challenged by researchers who emphasize the importance of investigating sex differences, this underrepresentation can be attributed to challenges such as care responsibilities, strict inclusion criteria, and concerns about increased variability. ( 18 , 19 ) Of particular relevance, merely five neuroimaging studies examined the neural correlates of sex differences in cue-induced reactivity and craving in cocaine use disorder (CUD) and none in heroin use disorder (HUD) as previously reviewed.( 9 ) Furthermore, studies of hormonal effects in women are scarce, particularly in substance use disorder (SUD).( 16 , 17 ) Enhanced salience of drug cues at the expense of non-drug reinforcers(e.g., food, social, sexual cues), associated with hyperactivity across reward, salience, and executive control networks (among others), is a hallmark of addiction.( 15 , 20 ) While limited in number, and although a mixed direction of effects across substances including cocaine and alcohol has been observed, clinical neuroimaging studies indicate consistent sex differences in such ventromedial prefrontal cortex (vmPFC) drug cue-reactivity in people with SUD.( 21 ) For example, drug cue-reactivity in the vmPFC to alcohol pictures was found to be higher in women than men ( 22 ), correlate with alcohol craving, and predict heavy drinking days, with the latter exclusively in women.( 23 ) Importantly, ovarian hormones can influence drug cue-reactivity, whereby cigarette-dependent women display striatal hyperreactivity to smoking cues during the late-follicular phase (increased estradiol) as compared to the mid-luteal phase (increased progesterone) and early-follicular phase (low estradiol and progesterone).( 17 ) Given its association with prospective drug use and relapse,( 24 ) reducing drug cue-reactivity has been a main goal in addiction treatment.( 25 ) A commonly used behavioral intervention for this purpose is emotion downregulation via cognitive reappraisal of drug cues; emotion upregulation via savoring of non-drug alternative (reward) cues has also been used, albeit much less frequently.( 26 – 28 ) In addition to the expected higher drug cue-reactivity in the vmPFC and striatum (and in the orbitofrontal cortex and inferior frontal gyrus [IFG]), we previously identified elevated cortico-striatal activity during drug reappraisal as directly compared to food savoring in individuals with HUD (iHUD). This activity in the dorsolateral PFC (dlPFC) correlated with higher methadone dose, while its direct contrast with drug cue-reactivity correlated with (lower) craving, suggesting that cognitive reappraisal is an effective, yet resource-demanding, regulation strategy.( 29 ) In the general population, studies found that men, compared to women, have higher dlPFC activation during emotion downregulation (>passive viewing) of negative images.( 30 , 31 ) Additionally, although an opposite effect was reported in postmenopausal women with depression,( 32 ) in naturally cycling women, estradiol administration lowers IFG/middle frontal gyrus activity during emotion downregulation.( 33 ) Progesterone has the opposite effect as suggested by a study where intrauterine release of a synthetic progestin (levonorgestrel) promoted higher frontal N2 amplitude during emotion upregulation of negative images, suggesting the recruitment of more attention and cognitive control resources.( 34 , 35 ) These findings collectively suggest potential sex and hormonal effects also on the regulation of drug cue-reactivity with implications for women’s mental health. Therefore, our study aimed to explore sex and hormonal effects on cortico-striatal mechanisms underlying drug cue-reactivity (vs. processing of non-drug reward and neutral cues) and its regulation in individuals with SUD. Given the limited and inconsistent findings, and especially absence of relevant studies in HUD, our first hypothesis posited significant sex differences in drug cue-reactivity, without specifying a direction for this effect. Reflecting risk and protective effects, respectively, we then hypothesized that women would exhibit higher drug cue-reactivity and lower drug reappraisal during the follicular phase, with the opposite pattern in the luteal phase, as directly correlated with fluctuations in estrogen, progesterone and craving. Due to the scarcity of studies on emotion regulation with alternative reward savoring, and space limitations, these results are reported in the Supplement .

Supplement Description: Supplement Methods , Results , Figures S1 – S11 , Tables S1 – S2