Skip to content

Overview: Innovative Study Designs vs. Traditional FIH

Innovative Study Designs vs. Traditional FIH

When are adaptive, umbrella, basket or hybrid trials optimal?

First‐in‐human (FIH) trials are the link between preclinical studies and human testing with the primary goal being to determine the safe dose range for further development of promising drug candidates.

Optimal FIH study design for dose-finding depends upon the agent under investigation. For a traditional cytotoxic drug, efficacy and toxicity are expected to rise in direct relation to dose, so the goal is to find the maximum tolerated dose (MTD). However, for a novel, molecularly targeted agent, monotonic dose-toxicity and dose-efficacy relationships cannot be assumed. These drugs work in myriad ways and the relevant study goal is to find the optimal biological dose.1 In this case, a traditional 3+3 design may not make sense.

Over the past 20 years, master protocol trials have matured through ongoing advances in design and execution. The development of immunotherapies such as pembrolizumab and nivolumab changed the drug development paradigm, shifting it further toward efficiency and collaboration. The result is increasing adoption of multiple design types seen today: umbrella, basket, hybrid, and adaptive. But when are these innovative study types truly beneficial?

What follows is an overview of several novel trial designs and traditional designs for early phase clinical trials in a variety of scenarios.


Master protocol and adaptive clinical trial designs

Master protocol and adaptive designs are intended to answer multiple questions with a single study.

Umbrella trials

This method enables multiple drugs to be tested against a single disease and may incorporate an adaptive element. In oncology, umbrella trials enroll patients with the same cancer type but with different genetic mutations. An excellent example is testing multiple drugs, with cohorts sorted based on pre-screening biomarker test results.


Basket trials

Basket trials test a single drug across multiple diseases. Study patients across different disease groups or subgroups — such as those with different types of cancer — share a common factor, such as a genetic mutation. These patients form a cohort and are assigned a drug that is expected to work in patients with that specific factor.


Hybrid Retrospective-Prospective Trials

In these studies, prospective and retrospective data are integrated to assess investigational or marketed drugs or devices. Existing, decentralized clinical trial (DCT) data is collected and linked with traditional clinical data for the same pool of subjects, such as ongoing electronic health record (EHR) entries.

Adaptive trials

This category of design allows pre-specified modifications to a trial after initiation, without protocol amendment. It enables key, prospectively planned study design changes in response to emerging study outcome data. The multiple types are sometimes combined:

  • Adaptive randomization, where the probability of treatment assignment changes according to assigned treatment of patients already enrolled in the trial
  • Drop-the-loser, which allows for dropping inferior treatment groups or adding additional arms
  • Group sequential, where the sample size is not fixed in advance
  • Sample size re-estimation, which allows for increases in sample size based on interim analyses
  • Adaptive dose-ranging, which shifts treatment allocation toward more promising doses
  • Biomarker-adaptive, which incorporates information from and may adapt based on biomarkers
  • Adaptive treatment-switching, where the investigator can switch a patient’s treatment from an initial assignment to an alternative treatment based on evidence of lack of safety or efficacy
  • Model-based designs, such Bayesian optimal interval (BOIN) and modified toxicity probability interval-2 (mTPI-2). BOIN is designed to derive the optimal decision rules for dose escalation or de-escalation. mTPI-2 partitions the probability of toxicity into a set of intervals to inform dose selection decisions.
  • Seamless phase II/III, which combines dose selection and confirmatory studies into a single trial


Traditional 3+3 design vs. BOIN vs. mTPI-2

A traditional FIH trial to determine the safe dose range for further clinical development evaluates one drug for a single disease, utilizing prospective data. The mTPI-2 study design is a variant of the traditional 3+3 design that uses a modified toxicity probability interval (TPI) to guide dose escalation and the BOIN design uses a Bayesian optimal interval to guide the dose escalation process.

Across all methods the key goals are to identify the new drug’s recommended phase 2 dose (RP2D) and MTD. The 3+3 design is understandably prevalent as it is easy to execute without any software or knowledge of basic statistical concepts. Of note, if toxicology data support sufficient safety of the starting dose, single person cohorts may be proposed for the initial dose cohorts in 3+3 designs.

Pre-specified starting doses are based on toxicological data from animal trials or previous clinical trial data. If no more than one in the initial trio of patients develops a toxic response, a second trio of patients are dosed. If more show toxicity, the MTD is the previous (lower) dose. If not, the test is repeated at the next, higher dose level. See the schema below of a 3+3 dose escalation design.


Key takeaways

Whether advanced or traditional, FIH study design and execution require deep experience and expertise to maximize insights and efficiency. Precision is positioned to support all aspects of early phase studies with efficient clinical development strategies, advanced trial designs based on statistical rigor and biomarker expertise, and real-time insights via our computational, informatics, and virtual sample inventory management platform.

Explore innovative trial designs vs. traditional FIH >


  1. Zang, Yong, et al. “Adaptive Designs for Identifying Optimal Biological Dose for Molecularly Targeted Agents.” Clinical Trials. 2014;11(3):319–327.