The gap between preclinical efficacy and clinical success in pain drug development is well-documented and widely lamented. Fewer than 2% of analgesic candidates survive the journey from bench to bedside. What is less commonly discussed are the specific instances where preclinical models got it right: where the animal data not only predicted the clinical outcome but revealed the underlying mechanism that explained it.
The development of ZYNRELEF (HTX-011) is one such case, and the translational lessons from it are worth examining in detail.
Postoperative pain management has been dominated by opioids for decades, despite their well-known liabilities: respiratory depression, nausea, constipation, delayed recovery, and the risk of persistent use. Long-acting local anesthetics have offered partial solutions, but duration of action remains a fundamental limitation. Standard bupivacaine provides 6 to 12 hours of analgesia. Most surgical patients need 48 to 72 hours.
Heron Therapeutics developed HTX-011, a fixed-dose combination of bupivacaine and the NSAID meloxicam in an extended-release polymer matrix. The hypothesis was that co-delivery of an anti-inflammatory agent would enhance and prolong bupivacaine's analgesic effect. But the question was: through what mechanism?
The answer came from a pig postoperative pain study using a flank incision model with von Frey mechanical allodynia, distress behavior scoring, human approach testing, pharmacokinetics, and histological endpoints.
The pig data revealed a specific, previously unappreciated mechanism. Surgical tissue inflammation lowers local pH. At the acidic pH of inflamed tissue, only 0.6% of bupivacaine exists in the unionized form capable of crossing neuronal membranes. Meloxicam, by reducing inflammation, restores tissue pH toward physiological levels, increasing the unionized bupivacaine fraction to 7.5%. This 12-fold increase in the pharmacologically active form of bupivacaine explained the synergistic, extended analgesia observed in the pig model.
This was not a finding that could have emerged from a standard rodent reflexive withdrawal assay. It required a model system with human-like skin architecture, relevant wound biology, and endpoints sensitive enough to capture both the magnitude and duration of the analgesic effect.
The mechanism identified in the pig model was subsequently confirmed in a Phase II human bunionectomy trial involving 237 subjects (Ottoboni et al., 2019, Reg Anesth Pain Med). The clinical data showed the same pH-dependent synergy between meloxicam and bupivacaine, the same extended duration of analgesia, and a clinically meaningful reduction in opioid consumption.
ZYNRELEF received FDA approval, a direct connection from pig model mechanistic data to a marketed drug.
Around the same period, aprepitant, an NK1 receptor antagonist with a strong preclinical rationale in rodent pain models, was tested in a pig peripheral neuropathic pain (PNT) model. It showed no analgesic effect. The same compound subsequently failed in human clinical trials for neuropathic pain.
Castel et al., 2016, The Journal of Pain, 17(1): 36-49
This is a critical and underappreciated aspect of model validation. A model that only confirms positive results is useful but incomplete. A model that also predicts clinical failure provides a genuine decision-making tool, one that can save sponsors years and millions of dollars by identifying compounds that won't translate before they enter Phase II.
The predictive value extends to comparative efficacy. In a head-to-head evaluation of three local anesthetics in pig POP (Exparel, Marcaine, and Naropin), the AUC analgesic ranking in pigs matched the human postoperative clinical ranking exactly (Castel et al., 2017, J Pain Res).
When a preclinical model reproduces the relative efficacy of approved drugs in the correct order, it provides a calibrated reference frame for evaluating novel candidates. This is the difference between a model that generates data and a model that generates decisions.
The ZYNRELEF story is not an isolated anecdote. It is a case study in what translational pain research looks like when the preclinical model captures the relevant biology (skin architecture, wound inflammation, peripheral nerve pharmacology, behavioral endpoints) at sufficient fidelity to reveal mechanisms that survive the transition to humans.
For sponsors designing preclinical pain programs, the question is straightforward: does your model have a track record of clinical prediction? If the answer is "we don't know," that is itself a risk factor worth addressing before IND-enabling studies.
MD Biosciences' pig pain models, including POP, PNT, and neuroma platforms, have generated preclinical data supporting FDA-approved drugs. To discuss translational study design for your pain program, contact neuro@mdbiosciences.com.