Peer Referral Pathways and 6-Month Outcomes in a Tirzepatide-Supported Digital Weight Loss Service: Retrospective Cohort Study

Introduction

The real-world effectiveness of modern weight loss medications, such as tirzepatide and semaglutide, depends heavily on medication adherence and sustained behavioral modification [1]. Recent studies have demonstrated that 6- and 12-month adherence rates in real-world tirzepatide-supported obesity programs are generally low [2,3]. While digital weight-loss services (DWLSs) deploy algorithmic nudges and multidisciplinary coaching to drive retention, external sociopsychological factors, such as preexisting peer support networks, are frequently overlooked [4]. Behavioral medicine frameworks suggest that patients entering interventions via social referrals benefit from built-in accountability and shared social proof, which can mitigate the attrition commonly observed in remote care [5,6]. However, the specific impact of peer-to-peer referral pathways on adherence and objective weight loss metrics within a medication-supported DWLS remains unquantified. This study evaluated the programmatic persistence and 6-month percentage weight loss of referred vs nonreferred patients receiving tirzepatide treatment via an Australian DWLS.

Methods

Background

This retrospective cohort study examined deidentified data from the Juniper (Eucalyptus Health) DWLS data repository on Google BigQuery. Patients were included if they initiated a tirzepatide treatment pathway between May 20 and December 2, 2025. Juniper patient eligibility is determined by doctors and nurse practitioners in accordance with Therapeutic Goods Administration guidelines for obesity treatment. Once accepted, patients communicate with prescribing clinicians, health coaches, and medical support staff via the Juniper app. All patient data are stored in the Juniper data repository, including weight data, which are submitted by patients via Bluetooth smart scales or manual input on the Juniper app.

The study cohort was stratified into referred patients and nonreferred patients. Referral was implemented as a tracked marketing mechanism using invitation links that existing patients could share with others. When a new patient enrolled via this link, the referrer and referee received a discount of AUD $75 (AUD $1=US $0.65 as of December 1, 2025) on 2 of their upcoming tirzepatide orders. The referral program did not include an in-app chat function or a peer-support group. Order costs ranged from AUD $449 to $855 (depending on tirzepatide dose) over the study period.

Adherence Criteria

Program adherence at 6 months was operationalized using 2 measures:

• Medication adherence: fulfillment of 5 or more orders within 183 days of program initiation.
• Per-protocol persistence: submission of at least 1 verified body weight entry within a strict window of day 173 to day 193 postinitiation.

Covariates

To isolate the independent effect of peer referral on mean 6-month percentage weight loss, propensity score matching (PSM) was implemented. Referred patients with a minimum 6-month program tenure were matched 1:1 with nonreferred controls using a nearest-neighbor algorithm based on age, sex at birth, ethnicity, baseline BMI, comorbidity count, previous tirzepatide/semaglutide use, and first order price. Differences in adherence rates were evaluated using Pearson χ² tests, and matched weight loss outcomes were compared via independent t tests. For differences in ordinal variables such as comorbidity count, we used a Wilcoxon rank sum test. To account for multiple statistical testing across baseline characteristics and outcome metrics, P values were adjusted using the Benjamini-Hochberg false discovery rate procedure. Statistical analyses were performed in RStudio (version 2023.06.1; Posit PBC), with significance set at P=.05.

Ethical Considerations

The Stanford University Institutional Review Board determined that this study did not constitute human subjects research (protocol 82970, October 22, 2025).

Results

The full cohort comprised 2576 referred and 31,873 nonreferred patients. Age, BMI, and ethnicity distributions were similar between groups (Table 1). However, the referred cohort organically exhibited a significantly higher concentration of male patients (592/2576, 23%) than the nonreferred cohort (2486/31,873, 7.8%) (P<.001) and a lower prevalence of prior tirzepatide/semaglutide use (438/3576, 17%) than the nonreferred cohort (7331/31,873, 23%) (P<.001).

A total of 9508 (27.6%) patients met the study’s 6-month adherence criteria (Table 2). Of the remaining 24,941 patients, 10,318 (30%) failed to receive at least 5 tirzepatide orders, while 14,623 (42.4%) met this criterion but failed to submit data within the specified window. Referred patients achieved significantly higher 6-month medication adherence (≥5 orders) than nonreferred patients (1236/2576 [48%] vs 13,387/31,873 [42%]; P<.01). This adherence disparity was maintained under the per-protocol persistence criteria, with 35% of referred patients sustaining medication compliance and submitting data within the 6-month weight window versus 27% of the nonreferred cohort (P<.01). Following 1:1 PSM to control for baseline imbalances (n=804 matched pairs), all mean SDs had values <0.10, indicating good covariate balance. Referred patients achieved a mean weight loss of 16.2% (SD 6.2%) compared to 14.1% (SD 6.5%) in the matched nonreferred control group (P<.001).

Discussion

In this retrospective analysis of a tirzepatide-supported DWLS, patients entering via peer referral were more likely to remain engaged in the program and, in PSM analyses, had higher observed 6-month percentage weight loss than nonreferred patients (16.2% vs 14.1%). These findings are preliminary but suggest that peer referral networks may have an important role in supporting retention and effectiveness in unsubsidized, medication-supported DWLSs.

Notably, the referral pathway was associated with a shift in participant demographics, particularly in the higher male representation (23% vs 7.8%). As men are traditionally underrepresented in weight management services [7], peer referral mechanisms may offer a promising approach to engage this group. However, further work is needed to understand underlying motivations and barriers.

This work has several limitations. Its retrospective, observational design limits causal inference, and although PSM was used to reduce demographic variability, residual confounding by unmeasured variables (eg, intrinsic motivation, socioeconomic status, or social network structure) is likely [8,9]. The analysis of weight outcomes in a per-protocol framework introduced survivorship bias and may overestimate effects relative to a broader DWLS population. Additionally, all weight data were self-reported and were therefore susceptible to error and social desirability bias. Finally, this study reflects a single commercial, unsubsidized Australian DWLS, which may limit generalizability to other health systems or populations. The observed associations cannot disentangle tirzepatide’s direct pharmacological effects from digital support components or social factors related to referral.

Despite these constraints, the data suggest that peer referral pathways may improve program persistence and weight loss within a medicated DWLS. Prospective studies are needed to clarify mechanisms, confirm reproducibility, and determine how best to operationalize peer networks to enhance retention and outcomes in obesity care.