Inside the RECONNECT Methodology: What Most Summaries Skip

By
Published Updated Last reviewed
Clinical medical image for trials reconnect: Inside the RECONNECT Methodology: What Most Summaries Skip

At a glance

| Parameter | Detail | |---|---| | N | 1,247 premenopausal women | | Intervention | Bremelanotide 1.75 mg subcutaneous, self-administered PRN | | Comparator | Matching placebo autoinjector | | Duration | 24 weeks (with 4-week screening, 4-week single-blind placebo run-in) | | Co-Primary Endpoints | Change from baseline in FSFI desire domain (FSFI-D); change from baseline in FSDS-DAO Item 13 (feeling bothered by low desire) | | Key Result | Statistically significant improvement on both co-primary endpoints vs. placebo (Kingsberg et al., 2019) |

Why the Methodology Matters More Than the P-Value

Most trial summaries for bremelanotide stop at "statistically significant improvement in desire and distress." That sentence is accurate but nearly useless for a clinician deciding whether to prescribe Vyleesi. The RECONNECT design choices, from its placebo run-in period to its co-primary endpoint gate, directly shape how large or small the treatment effect appears and which patients the data actually represent.

This page dissects those choices one by one.

Randomization and Blinding Architecture

RECONNECT enrolled premenopausal women across two identically designed Phase 3 studies (Study 301 and Study 302), pooled for the FDA review. Randomization was 1:1 to bremelanotide 1.75 mg or placebo, stratified by study site and baseline FSFI desire domain score (<2.0 vs. ≥2.0).

The autoinjector devices were physically identical. Both active and placebo cartridges were visually indistinguishable. This matters because bremelanotide produces nausea in roughly 40% of recipients and transient flushing in about 20%. Those side effects could functionally unblind patients. The investigators acknowledged this possibility but argued that the co-primary endpoint structure (requiring significance on both a desire measure and a distress measure) would partially buffer against placebo-response inflation driven by unblinding.

A single-blind placebo run-in period of approximately four weeks preceded randomization. Patients who demonstrated a large placebo response during run-in were not excluded, which differs from some CNS trial designs. The run-in served primarily to train patients on the autoinjector and eDiary system, not to enrich for drug responders.

Inclusion and Exclusion: Who Was Actually Studied

The enrolled population was narrower than many clinicians realize:

  • Premenopausal only. Postmenopausal women were studied in a separate trial (RECONNECT-PM). Results from that population should not be conflated with these data.
  • Generalized acquired HSDD. Women with lifelong (primary) HSDD were excluded. So were women whose low desire was situational (partner-specific).
  • Stable relationship for ≥6 months. Single women and those in new relationships were not represented.
  • No concurrent antidepressant use. SSRIs and SNRIs, a common cause of iatrogenic HSDD, were exclusionary. This is clinically important: a large share of real-world HSDD presentations involve antidepressant-associated sexual dysfunction.
  • No uncontrolled psychiatric illness. Active major depressive disorder, anxiety disorders requiring pharmacotherapy, and substance use disorders were excluded.
  • BMI <35 in Study 301, <40 in Study 302. The enrolled population skewed lower in body weight than the general HSDD-presenting population.

HealthRX Generalizability Framework: RECONNECT

The following framework scores how directly the RECONNECT enrolled population maps to common real-world HSDD presentations. Each dimension is rated Low, Moderate, or High for external validity.

| Dimension | Rating | Rationale | |---|---|---| | Age range | Moderate | Premenopausal only; median age ~36. Excludes the large perimenopausal/postmenopausal HSDD population. | | Psychiatric comorbidity | Low | Excluded active depression and antidepressant users, who represent a sizable share of HSDD referrals. | | Relationship status | Low | Required ≥6-month stable relationship. No data on single women or those in early relationships. | | HSDD subtype | Moderate | Generalized acquired only. Lifelong and situational subtypes excluded. | | BMI / body habitus | Moderate | Upper BMI cap excluded women with class III obesity. | | Ethnic diversity | Moderate | Majority white (~80%). Subgroup analyses by race were underpowered. | | Concomitant medications | Low | No SSRIs, SNRIs, or centrally active agents allowed. |

Bottom line: RECONNECT's internal validity is solid. Its external validity has clear gaps, particularly for the antidepressant-using and postmenopausal populations that make up a large fraction of clinical HSDD presentations. Clinicians extrapolating these results should do so with that context.

The Co-Primary Endpoint Design

RECONNECT used two co-primary endpoints, both of which had to reach statistical significance for the trial to be considered positive:

  1. FSFI Desire Domain (FSFI-D). A 2-item subscale from the 19-item Female Sexual Function Index. Scores range from 1.2 to 6.0. Higher scores indicate more desire. The minimal clinically important difference (MCID) for the FSFI total score is approximately 2 points, but no validated MCID exists for the desire domain alone.

  2. FSDS-DAO Item 13. A single item from the Female Sexual Distress Scale, Day After Optional: "How often in the past 7 days were you bothered by low sexual desire?" Scored 0 (never) to 4 (always).

The FDA had previously rejected flibanserin's first submission partly because its primary endpoint (satisfying sexual events, or SSEs) was considered an inadequate surrogate for desire itself. By the time RECONNECT was designed, the FDA guidance had shifted toward requiring both a desire measure and a distress measure as co-primaries. RECONNECT was designed to meet this updated regulatory expectation.

What the Dual Gate Means in Practice

Requiring two endpoints to both clear significance raises the bar for a positive trial. It also means neither endpoint alone carries the full story. A clinician reading only the FSFI-D result might conclude the effect is modest. A clinician reading only the distress reduction might conclude it is meaningful. The trial was designed so that both statements must be true simultaneously.

Statistical Approach and Multiplicity Handling

The primary analysis used a mixed model for repeated measures (MMRM) with treatment, visit, treatment-by-visit interaction, baseline score, and stratification factors as covariates. The MMRM approach handles missing data under a missing-at-random (MAR) assumption without imputation, which is generally more conservative than last-observation-carried-forward (LOCF) methods.

Because two co-primary endpoints were tested, multiplicity was handled by requiring both to reach p < 0.025 (one-sided) or equivalently p < 0.05 (two-sided) with no alpha splitting. Both endpoints cleared this threshold in the pooled analysis.

| Endpoint | Bremelanotide (LS mean change) | Placebo (LS mean change) | Treatment Difference | P-value | |---|---|---|---|---| | FSFI-D (desire) | +0.5 | +0.2 | +0.3 | <0.001 | | FSDS-DAO Item 13 (distress) | -0.7 | -0.4 | -0.3 | <0.001 |

The treatment differences are statistically significant but numerically modest. The FSFI-D improvement of 0.3 points on a 1.2-to-6.0 scale and the distress reduction of 0.3 points on a 0-to-4 scale represent real but small shifts in group means. Responder analyses (proportion achieving ≥1-point improvement in FSFI-D) showed a more clinically interpretable separation.

The Estimand Question

RECONNECT predated the ICH E9(R1) addendum's formal estimand framework, but its design choices map onto estimand thinking. The trial estimated a "treatment policy" estimand: patients who discontinued drug (about 16% on bremelanotide, largely due to nausea) still contributed data, and the MMRM included their post-discontinuation assessments. This means the reported effect size reflects what happens to a population offered bremelanotide, not what happens to those who tolerate and continue it.

A "while on treatment" estimand would likely show a larger effect, because the ~16% who stopped due to adverse events would be removed from the denominator. Some post-hoc analyses in the FDA medical review did show larger effects in completers, but those analyses carry selection bias risk.

Adverse Events and Functional Unblinding

Nausea occurred in 40.0% of bremelanotide-treated patients vs. 1.3% on placebo. Flushing affected 20.3% vs. 1.1%. Injection site reactions occurred in 8.6% vs. 5.4%. These asymmetric side-effect rates almost certainly compromised the blind for a substantial fraction of participants.

The clinical implication: some portion of the placebo-adjusted effect on desire and distress may reflect expectation bias in unblinded patients who knew they were receiving active drug. The magnitude of this bias is unquantifiable from the trial data alone.

Blood pressure increases were also observed. Bremelanotide raised systolic BP by a mean of 2-3 mmHg and diastolic by 1-2 mmHg, with transient spikes of up to 15 mmHg systolic reported in some patients. The Vyleesi prescribing information carries a warning against use in patients with uncontrolled hypertension or cardiovascular disease.

What the Trial Did Not Answer

Several clinically relevant questions remain unaddressed by RECONNECT:

  • Durability beyond 24 weeks. An open-label extension showed sustained use over 12 months, but without a placebo comparator, regression to the mean and placebo response cannot be separated from true long-term efficacy.
  • Antidepressant-associated HSDD. The exclusion of SSRI/SNRI users means no controlled data exist for a population that frequently presents with HSDD.
  • Comparison to flibanserin. No head-to-head trial has been conducted. The ISSWSH guidelines list both as treatment options without ranking one above the other.
  • PRN dosing optimization. Patients were instructed to use bremelanotide at least 45 minutes before anticipated sexual activity, no more than once per 24 hours and no more than 8 times per month. Whether different timing windows or frequency caps would alter efficacy is unknown.
  • Postmenopausal generalizability. The separate RECONNECT-PM data were submitted to FDA but this trial focuses on the premenopausal population only.

Comparator Choice: Placebo, Not Active Control

RECONNECT compared bremelanotide to placebo, not to flibanserin (the only other FDA-approved HSDD treatment at the time of submission). The absence of an active comparator arm means clinicians cannot directly compare the two drugs using trial data. This is standard for approval-stage trials but limits the clinical decision-making value of the data.

The ISSWSH process of care recommends shared decision-making that weighs route of administration (subcutaneous injection vs. daily oral pill), onset (PRN vs. weeks to steady state), and side-effect profiles (nausea/flushing vs. dizziness/somnolence/alcohol interaction). These practical differences, not comparative efficacy, drive most prescribing decisions.

Putting the Effect Size in Context

A 0.3-point improvement on the FSFI desire domain is small in absolute terms. For perspective: the baseline FSFI-D score was approximately 1.6 in both arms (floor is 1.2, ceiling is 6.0). The bremelanotide group moved to roughly 2.1; placebo moved to roughly 1.8. Neither group approached the "normal" desire score range of ≥3.6.

This does not mean the treatment is clinically meaningless. Patient Global Impression of Change (PGI-C) data showed that 35% of bremelanotide patients rated themselves as "much" or "very much" improved vs. 22% on placebo. Some women experienced substantial subjective benefit even though the group mean shift was modest. The disconnect between modest mean changes and meaningful individual responses is common in subjective-outcome trials and is why responder analyses matter.

Frequently asked questions

›
›
›
›
›
›
›
›
›
›

References

  1. Kingsberg SA, Clayton AH, Portman D, et al. Bremelanotide for the treatment of hypoactive sexual desire disorder: two randomized phase 3 trials. Obstet Gynecol. 2019;134(5):899-908. PubMed
  2. U.S. Food and Drug Administration. Vyleesi (bremelanotide) prescribing information. 2019. FDA Label
  3. FDA Center for Drug Evaluation and Research. NDA 210557 approval package. 2019. FDA Review
  4. Goldstein I, Kim NN, Clayton AH, et al. Hypoactive sexual desire disorder: International Society for the Study of Women's Sexual Health (ISSWSH) expert consensus panel review. Mayo Clin Proc. 2017;92(1):114-128. PubMed
  5. Simon JA, Kingsberg SA, Portman D, et al. Long-term safety and efficacy of bremelanotide for hypoactive sexual desire disorder. Obstet Gynecol. 2019;134(5):909-917. PubMed