​Drug improvement stays some of the capital-intensive actions in life sciences. A 2021 peer-reviewed research revealed in Clinical and Translational Science found that the success fee of a drug candidate from the beginning of scientific trials to advertising approval sits at roughly 10–20% and has not meaningfully modified in a long time.

A separate evaluation revealed in JAMA and listed by the National Institutes of Health found that Phase III trials account for the biggest share of scientific improvement prices, pushed by bigger affected person enrollment and longer trial durations than in earlier phases.

Many of those prices stem from selections made with fragmented proof — dose choice primarily based on restricted early-phase knowledge, security indicators assessed in isolation, and affected person variability understood solely after massive trials are already underway.

In a dialog between Emerj’s Matthew DeMello and Shefali Kakar, Global Head of PK Sciences and Oncology at Novartis, the dialogue defines how AI is reworking drug improvement by giving enterprises earlier readability on program viability and a extra exact understanding of actual affected person impression — two levers that straight affect velocity, danger, and capital allocation throughout the R&D portfolio.

This article explores two of Shefali Kakar’s core insights for enterprise groups navigating AI‑enabled drug improvement:

• Accelerating improvement selections to enhance capital focus: Earlier readability on program viability allows quicker go/no‑go selections and tighter alignment of assets to the alternatives with the strongest proof.
• Deepening affected person perception to strengthen program design: A clearer view of how particular person affected person traits have an effect on drug response helps safer, extra focused trials and builds a stronger proof base throughout improvement phases.

Episode:  Rethinking Clinical Trials with Faster AI-Driven Decision Making – with Shefali Kakar of Novartis​

Listen to the complete episode beneath:

Guest: Shefali Kakar, Global Head of PK Sciences and Oncology at Novartis

Expertise: Clinical Pharmacology, Oncology Drug Development, Pharmacokinetics/Pharmacodynamics (PK/PD), Biologics & Immuno-oncology

Brief Recognition: Shefali Kakar leads international PK Sciences and Oncology at Novartis, the place she drives scientific pharmacology and dose optimization technique throughout oncology applications, together with biologics and immunotherapies. Over almost 20 years at Novartis, she has held progressive management roles from Fellow to Executive Director, shaping PK/PD modeling and proof integration approaches throughout respiratory and oncology portfolios. Prior to Novartis, she served as a Senior Principal Scientist at Pfizer, contributing to drug metabolism and scientific pharmacology applications. She holds a Ph.D. in Clinical Pharmacology from the University of Michigan.

Accelerating Development Decisions to Improve Capital Focus

Kakar opens the dialogue by contrasting the sluggish, sequential choice patterns which have lengthy formed drug improvement with the extra built-in, evidence-driven strategy AI now makes attainable.

Historically, groups superior dose and program selections step-by-step — Phase I, then Phase II, then Phase III — every stage ready on the final and every counting on slender slices of information. As Kakar makes clear, this construction typically forces organizations to commit capital earlier than they’ve an entire image of viability.

From there, she describes how AI-enabled modeling adjustments the economics of improvement. Instead of treating every section as a discrete gate, groups can now consider proof throughout research and determine viable dosing methods earlier within the course of — typically surfacing solutions that giant, devoted trials had been by no means designed to provide. Kakar factors to a case the place built-in modeling throughout a modest dataset revealed one thing a full Phase III comparability had not:

“We had a Phase III research the place a selected dose was investigated, however once we checked out the entire knowledge collectively and modeled it out, it turned clear that whether or not the drug was given as soon as a day or twice a day with the identical whole dose, we ended up with the identical efficacy. This was not one thing that was investigated in a really massive Phase III trial — it got here from modeling throughout a extra modest dataset. That is the type of reply you may solely get whenever you cease treating dose as a set selection at every section and begin it as a continuum throughout the complete proof base.”

— Shefali Kakar, Global Head of PK Sciences, Oncology at Novartis

This means to revisit assumptions utilizing the complete physique of proof — even after Phase III — represents a direct departure from the historic “decide one dose and commit” paradigm. It reduces the necessity to take a look at each permutation in massive trials and provides groups earlier readability on whether or not a program is heading in the right direction.

For enterprise groups, the sample Kakar describes — earlier readability on dose viability, fewer massive confirmatory trials, and the power to revisit assumptions throughout phases — naturally results in quicker go/no‑go selections and tighter alignment of capital to the alternatives with the strongest proof.​

Deepening Patient Insight to Strengthen Program Design

One of the themes Kakar returns to is how little visibility groups have historically had into the affected person‑degree elements that form drug publicity. Instead of a unified view throughout research, organizations typically labored from small, standalone impairment cohorts that had been by no means powered to disclose how traits like kidney or liver perform meaningfully altered response. The end result wasn’t simply sluggish studying — it was fragmented studying, with every research providing solely a partial image of how actual sufferers would fare.

AI‑enabled modeling provides groups a manner out of that fragmentation. By pooling affected person‑degree knowledge from massive Phase III trials, groups can study how particular covariates affect publicity or antagonistic occasions with out operating a separate sub‑research for every query. The perception comes from the proof already in hand, not from a brand new, narrowly scoped experiment.

Kakar describes how this shift has modified dose selections in follow:

“In the previous, we’d run a separate research simply to grasp how kidney impairment affected drug publicity. Now we are able to embed that query straight into the Phase III trial — have a look at the sufferers already enrolled, study how their kidney perform correlates with publicity or antagonistic occasions, and use that to find out whether or not a dose adjustment is required. What used to require its personal cohort now comes from the info we have already got.”

— Shefali Kakar, Global Head of PK Sciences and Oncology at Novartis

For Kakar, that is the true worth of built-in modeling: earlier readability on how completely different affected person teams reply, and the power to regulate dose or monitoring expectations earlier than these variations turn out to be late‑stage surprises.

​