Bremelanotide RECONNECT Phase 3 Trials

RECONNECT comprised two randomized, double-blind, placebo-controlled phase 3 trials in premenopausal women with acquired, generalized hypoactive sexual desire disorder (HSDD). Across NCT02333071 and NCT02338960, 1,267 participants were randomized. Both trials reported statistically significant differences favoring bremelanotide on their final co-primary patient-reported outcomes: sexual desire and distress related to low desire. The key secondary satisfying-sexual-event outcome was not statistically significant.[1–3][6][7]

That mixed result needs the analysis history alongside it. The endpoint hierarchy changed by amendment, more than two-fifths of active-group participants did not complete the controlled phase, and the primary end-of-study assessment could occur before week 24. The findings supported evaluation of a specific finished drug product; they do not establish benefit for every outcome or equivalence of another bremelanotide-labeled material.

Program at a glance

The population was predominantly White (85.6%) and U.S.-based (96.6%). Entry criteria included stable relationships and a defined HSDD diagnosis; several psychiatric conditions and medication exposures were excluded. These limits matter when considering generalizability.[2][3][7]

Endpoint definitions and amendment history

The final co-primary endpoints were change from baseline to end of study in the Female Sexual Function Index desire domain (FSFI-D) and item 13 of the Female Sexual Distress Scale–Desire/Arousal/Orgasm (FSDS-DAO). Higher desire scores and lower distress scores favored active treatment. End of study meant the participant's last double-blind visit, not necessarily a completed week-24 assessment.[6–8]

The FDA review records an October 10, 2016 protocol amendment that moved satisfying sexual events from an original co-primary endpoint to a secondary endpoint and made distress co-primary. This followed a September 8 meeting; FDA indicated that the change was acceptable before database lock. The reviewer acknowledged that blinded event data might have suggested the event-count result would be null, but considered the change consistent with workshop/guidance input and not apparently a significant threat to trial integrity.[6]

This is the final amended hierarchy, not a claim that the endpoints remained unchanged from trial inception. The standalone protocol versions, full statistical analysis plan, and database-lock record were not obtained for this editorial review. The chronology comes from FDA's assessment.

The FDA review describes testing both co-primary endpoints first at a two-sided 0.05 level, requiring both to succeed before proceeding through ranked secondary endpoints. The paper's sample-size discussion instead mentions a 0.025 inference threshold. These source descriptions are not silently merged; the reported co-primary p-values were at most 0.005, so this difference does not change their reported significance.[6][7]

Co-primary results

The paper reports the following between-group differences in mean change from baseline to end of study:

These are scale-score differences, not percentages cured or a guarantee of individual benefit. FDA's primary-analysis tables identify the study-specific mean differences and Wilcoxon rank-sum tests. They should not be relabeled as adjusted contrasts from the paper's separate repeated-measures plots.[6][7]

The paper's primary method carried the last observation forward. Its Figure 1 presents a different analysis of repeated visits using a mixed model. A graph of modeled changes over time is not interchangeable with the primary end-of-study comparison.[7]

The paper also reports anchored responder analyses. Those depend on the chosen anchor and definition: FDA's outcome-assessment review did not accept the sponsor's proposed desire-score threshold unchanged. Statistical significance, an average score difference, and an individual response threshold answer different questions.[6][7] See confidence intervals and p-values for broader interpretation principles.

Satisfying sexual events and the testing sequence

The key secondary endpoint counted satisfying events associated with study drug within a defined window: events within 16 hours, reported within 72 hours. These are measurement rules, not instructions for use. The paper reports no significant group difference: p=0.764 in Study 301, p=0.702 in Study 302, and p=0.630 in the integrated analysis.[6][7]

Under the final testing hierarchy, that null result stopped confirmatory testing of the remaining ranked secondary endpoints. Later results were supportive or exploratory, even if an individual p-value was small. The paper's post hoc analysis of the proportion of encounters judged satisfying used a different denominator; it does not replace the null result for the number of satisfying events.[6][7]

The event-count result does not erase the desire/distress findings, and the positive co-primary findings do not make the event-count result positive. A nonsignificant comparison also does not prove that the groups were equivalent. The negative and null study-results guide explains these distinctions.

Attrition, populations, and missing observations

The controlled-phase flow is consistent across the paper's supplement, the randomized registry milestones, and FDA's disposition table:

Overall, 856 of 1,267 randomized participants completed and 411 did not. Active-group noncompletion was 41.9% versus 16.0% with placebo in Study 301, and 43.8% versus 28.4% in Study 302. Adverse events were an important reason, but these are all-cause noncompletion rates, not adverse-event-only discontinuation rates.[2][3][6–8]

Randomized, treated, modified intention-to-treat, and endpoint-evaluable populations are different. The modified intention-to-treat set required receipt of treatment and at least one double-blind follow-up visit. Further missing baseline or endpoint values reduced the evaluable totals to 1,198 for desire and 1,194 for distress. These should not be presented as 1,267 complete week-24 observations.[5–7]

Registry headline enrollment includes an earlier study stage and is not the randomized denominator. Study 302 also has a source inconsistency: its “received intervention” flow reverses the by-arm 301/303 counts relative to its adverse-event module and FDA's disposition table. The safety figures here follow the paper, supplement, and FDA table: 627 active and 620 placebo overall. The conflicting registry row is not silently used to recalculate them.[2][3][6–8]

Last-observation-carried-forward does not recover what would have happened at a missing later visit. FDA additionally reports that absent encounter-diary data were treated as zero encounters/events and that participants missing a co-primary endpoint were excluded from that endpoint's primary analysis. These assumptions are especially relevant with unequal withdrawal.[6][7]

FDA examined alternative imputations and completer/dropout analyses and considered plausible sensitivity analyses supportive overall; not every extreme or trimmed analysis was statistically significant. Some extreme assumptions reversed the Study 302 mean direction, although FDA regarded those worst-case assumptions as implausible. This does not justify either ignoring missingness or dismissing all efficacy findings. See trial estimands and missing data for the questions such methods address.

Harms and safety-reporting scope

The paper's integrated safety table reports:

The integrated adverse-event table includes events after the core phase but before open-label extension treatment began. Consequently, its pooled counts are not always the sum of the separately displayed within-study counts. These windows should remain explicit.[7][8]

Seven active participants experienced ten serious events, compared with four events in three placebo participants. One serious episode in Study 301, involving vomiting and headache, was considered related to study drug. No serious event in Study 302 was considered related, and no deaths were reported in either core study. Those are the paper's attribution statements, not proof that every other event was unrelated.[7][8]

The paper reports that nausea led 8.1% of active participants to discontinue study drug. Many events were mild or moderate, but that does not make them unimportant. Recognizable adverse effects such as nausea may also affect masking when outcomes are self-reported; the randomized design alone does not establish that participants remained unaware of assignment.[7]

The current finished-product label separately reports adverse-reaction discontinuation of 18% versus 2% and includes contraindications for uncontrolled hypertension or known cardiovascular disease, plus warnings about blood-pressure/heart-rate changes, focal hyperpigmentation, and nausea. Label summaries and paper-specific analyses should retain their own scope; this is not individualized treatment advice.[5]

Sponsor role and conflicts

Palatin Technologies sponsored the trials and had roles in design, execution, analysis, reporting, and funding. The authors reported access to relevant individual and aggregated data and took responsibility for the publication. Editorial support from Phase Five Communications was funded by AMAG Pharmaceuticals, the licensee at that time.[7]

The paper discloses sponsor/licensee employees and stockholders among the authors, alongside other industry relationships. It also describes an independent anchor-assessment committee for responder thresholds. That trial committee is not an independent review of this webpage, and this summary is not a reanalysis of participant-level data.[7]

FDA approval and present-day boundaries

Drugs@FDA records the finished prescription product Vyleesi under NDA 210557, originally approved June 21, 2019, with Cosette currently shown as the company. The DailyMed page checked for this review was updated November 13, 2025; its prescribing information is revised March 2024. The approval-history page's linked 2019/2020 labels and the newer DailyMed document are distinct dated records.[4][5]

The indication is restricted to premenopausal women with acquired, generalized HSDD meeting the label's conditions, not low desire caused by another medical/psychiatric condition, relationship problems, or a medication/substance. It is not an indication for men, postmenopausal women, or sexual-performance enhancement.[5]

FDA's review explains that increased satisfying-event counts were not required for this HSDD approval decision; improvement in low desire and associated distress was central. That regulatory judgment does not turn the trial's null event-count result into a positive result.[6]

Approval applies to the specific finished product and labeling, not every material using the name bremelanotide or PT-141. Neither this trial nor a shared ingredient name authenticates a lot or establishes formulation, quality, or clinical equivalence. See approved drugs versus research materials, checking trial and FDA status, and the melanocortin research hub.

Frequently asked questions

Why are there two trial identifiers?

RECONNECT included two separate controlled studies, NCT02333071 and NCT02338960. Their integrated publication is not a third randomized trial.

Were all 1,267 randomized participants included in every result?

No. There were 1,247 treated safety participants, 1,202 in the modified intention-to-treat set, and smaller endpoint-evaluable populations.[6–8]

Did both co-primary outcomes succeed?

Both trials reported statistically significant desire and distress results under the final amended hierarchy. The change in that hierarchy and the analysis methods remain part of the evidence record.[6][7]

Did satisfying sexual events significantly increase?

No. The key event-count comparison was null in both trials. A later post hoc proportion-of-encounters analysis is a different outcome.[7]

Does end of study always mean week 24?

No. It meant the participant's last double-blind visit; early discontinuation could make it earlier. The primary method and repeated-measures figures should not be conflated.[6][7]

Why are the withdrawal rates important?

Withdrawal was higher with active treatment. Earlier observations and statistical assumptions cannot establish every missing later outcome, and adverse-event discontinuation is not the same as all-cause noncompletion.

Does Vyleesi approval apply to any PT-141 research material?

No. It concerns the approved finished product and its indication, warnings, and labeling, not another material identified by a familiar name.

Primary and official sources

1. Published paper record: RECONNECT, PMID 31599840
2. ClinicalTrials.gov: Study 301, NCT02333071
3. ClinicalTrials.gov: Study 302, NCT02338960
4. FDA Drugs@FDA: NDA 210557 approval history
5. DailyMed: Vyleesi prescribing information, revised March 2024
6. FDA multidisciplinary review for NDA 210557
7. Open-access primary paper, PMC6819021 and official ACOG-hosted paper PDF
8. Original journal supplement: trial design, endpoints, disposition, nausea, and serious events

Review and corrections

Sources checked September 2, 2026 (America/Edmonton) / September 3, 2026 (UTC). The review used the complete primary paper and its six-page journal supplement, fresh official registry records, relevant FDA review sections, approval history, and dated current labeling. Standalone protocols, the complete SAP, database-lock documentation, and individual participant data were not obtained; no claim is made to have audited the entire 539-page FDA bundle. PubMed returned an accompanying commentary relationship, not a correction notice; that retrieval is not a guarantee about notices elsewhere.

Serum Research is responsible for this editorial summary. No independent scientific, clinical, statistical, analytical, or regulatory review of this page is claimed. It is not medical advice, a use recommendation, or a quality/equivalence conclusion. Read how we review evidence, browse the study index and research library, or report a correction.