When the Owner Is Your Instrument: Measuring Whether a Treatment Actually Works | Animal Health Insights Town Hall

When the Owner Is Your Instrument: Six Lessons on Measuring Whether an Animal Drug Actually Works

Insights from Lesley Rausch-Derra, DVM, MS, a drug developer with a rare record of getting new therapeutics through the FDA’s Center for Veterinary Medicine. Shared during Prelude’s Animal Health Insights webinar series.

In a companion animal study, your patient cannot tell you whether the treatment worked. There is no smiley-face pain scale a dog can point to, no way to ask a cat if the nausea has passed. The read on whether a drug is doing its job comes from a person watching the animal at home. That makes the thing you measure, your endpoint, the most fragile part of the entire program. Get it wrong, and a drug that genuinely works can still fail to show it.

Few people have navigated that problem more times than Lesley Rausch-Derra. She led drug development and regulatory affairs at Aratana Therapeutics, where she is first author on the pivotal effectiveness study for a novel osteoarthritis pain therapy built on an owner-completed questionnaire. Her owner-assessment work runs well beyond pain, into appetite and nausea, where no validated instrument existed and she built one from scratch. She later served as Chief Development Officer at a companion animal gene therapy company, founded and led Verté Therapeutics, and now advises sponsors through Right Direction Animal Health Consulting. On a recent Animal Health Insights webinar with Prelude CEO Tommy Jackson, she walked through how to build an owner-reported endpoint that holds up. Six lessons stood out.

1. When there is no objective measure, the owner is your instrument

Some questions can only be answered from the couch, not the clinic. “Some things, you only know if the dog is getting better by the home assessment,” Rausch-Derra said. “Pain, nausea, seizures. You have to track that at home.” For an entire class of indications, the owner is not a supplementary data source. The owner is the measuring device. That reframes the design problem: you are not just choosing a molecule and a dose, you are choosing a human instrument, and everything that makes a human unreliable becomes part of your study risk.

2. You do not always have to start from scratch

A decade ago there was little to borrow. Today a large body of validated owner instruments exists, and Rausch-Derra pointed the audience to the RCVS Knowledge quality-of-life and pain assessment library, which catalogs instruments by condition and notes what has been tested. Pain is especially well served, with tools like the Canine Brief Pain Inventory (CBPI), the Helsinki index, LOAD, and client-specific outcome measures. Her advice is to do the literature work first: “It’s a lot cheaper to sit and review a bunch of literature than to run multiple studies.” Sometimes you lift a validated instrument wholesale; sometimes you borrow only the questions that proved they move.

3. When nothing exists, build it, then pilot it

For the appetite drug Entyce, there was no validated instrument, so they made one. “There was nothing. We were on a whiteboard,” she recalled, “writing which questions made sense and which ones would be easy for the owners to do.” The questions were deliberately concrete: willingness to eat, begging behavior, anticipating mealtime, how much the dog ate when food was put down. Where a question was too hard to answer reliably, they proposed a lab measure to CVM for that part of the effectiveness case, and the agency agreed.

The deeper lesson is how you run the study that de-risks a homemade endpoint. A pilot is where you find out whether your questions capture the effect, and the common mistake is treating the pilot like a mini-pivotal. “In your pilot, you do not need a primary variable. You don’t need a secondary variable,” she said. “Leave yourself room to look at the data. Don’t back yourself into a corner.” Go broad before you go narrow, then commit. “When you get to CVM for your pivotal, they will make you put a stake in the ground. You don’t for a pilot. So your best chance of success is running a good pilot first.” One technical warning: power the pilot honestly, because the field is messier than the lab and an underpowered study can miss a real effect.

4. Bias is unavoidable, so engineer around it

“You are never going to get rid of bias,” Rausch-Derra said. The owner is untrained, the placebo response is real, and animals are watched subjects: “The minute you start watching your cat, it behaves differently. They may not let you know that they know, but they know.” You do not solve bias, you dilute and contain it, with adequate power, owner training, tight inclusion and exclusion, and simple response scales. Two specifics: the person who trains the owner cannot be the unblinded treatment administrator, which is why app-delivered training beats a coach in the room; and on response options, keep it to three, better, same, or worse, because extra gradations just invite fatigue and hesitation.

5. Treat CVM as a partner, and know that Europe reads it differently

The most reassuring theme was that the regulator is not the adversary teams fear. “Always go to the regulator first,” she said. “Don’t be afraid to ask those questions to CVM. If you give them a reason why, they’ve always been fairly reasonable.” Bring pilot data, explain the rationale, and align on the definition of success before the pivotal. Geography matters: in her experience the US is comfortable with owner-reported endpoints, while Europe leans on the veterinarian’s assessment. She also flagged a live wildcard, recent turnover at CVM, which may change how flexible the familiar reviewers once were. Her verdict on the US regulator was still affectionate: “I’m actually thankful for CVM, because at least they’ll give you an answer, even if you don’t like it.”

6. Pair the subjective with the objective, but do not lean on wearables yet

When an objective gold standard exists that fits your indication, correlate the owner measure against it. For the appetite program, a bigger appetite should show up as weight gain, and the lab data backed the owner reports. That corroboration strengthens the case everywhere and is often what Europe wants to see. The obvious modern candidate is the wearable, and here she was direct: not yet. Activity monitors lose context, because a dog can move plenty and still be in pain, and she has not seen the validation. “If a monitor isn’t validated, I don’t want to touch it. I want to see the data.” She expects the field to get there, but the published evidence is not there today.

The Questions the Audience Wanted Answered

With several audience questions answered live during the session, a few threads added depth beyond the main takeaways.

How do you reduce owner bias when a long instrument like CBPI wears owners down?

Too many questions kills attention. Match the instrument tightly to what you are actually measuring, and consider client-specific outcome measures, where the owner scores the specific activities that already bother them. Placebo bias never fully disappears, so pair the instrument with adequate power and tight inclusion and exclusion criteria.

Should you train owners once up front or at every visit?

Retraining is fine, but keep it blinded. The unblinded treatment administrator should not be coaching owners, or a regulator can question whether the coaching influenced the answers. App-delivered training, with consistent instructions and pictures, sidesteps the problem.

For an owner overall assessment, are three choices better than four or five?

Three: better, same, or worse. It is simpler for the owner and cuts down on hem-and-haw answers. CVM has accepted a simple better-or-worse read as a primary effectiveness measure.

Does the time of day of the owner assessment affect the answers?

It depends on the indication. Appetite is best captured at feeding, nausea several times a day, and seizures whenever they happen. A practical setup is two sets of questions: scheduled prompts at set intervals, plus an always-available set the owner can complete the moment something happens.

Can a multi-pet household bias the owner’s assessment?

It can, depending on the indication and how the animals relate to each other. If it is a plausible risk, write a protocol parameter asking owners not to change the home environment during the study.

What This Means for Your Next Study

The thread running through every answer is that the measurement decision is made long before the data comes in, and it is the decision most likely to make or break the program. Choose the instrument deliberately, validate or pilot it before you build on it, keep the owner’s job simple, align with the regulator early, and back the subjective signal with an objective one wherever you can.

It also raises a quieter point about infrastructure. When your instrument is a pet owner at the kitchen table, the quality of your evidence depends on how cleanly and consistently that data comes in: the same questions, delivered the same way, at the right moments, captured without breaking the blind. Teams that treat owner-reported data as a first-class design problem, not an afterthought, are the ones whose studies hold up.

---

Watch the Full Conversation

This article is based on Prelude’s Animal Health Insights webinar featuring Lesley Rausch-Derra, DVM, MS, of Right Direction Animal Health Consulting. Watch the full replay here.

Animal Health Insights is a webinar series hosted by Prelude CEO Tommy Jackson, bringing together experts from across the animal health R&D community to discuss the topics that matter most to clinical development teams. Subscribe to our newsletter to get notified about upcoming episodes.