The Spreadsheet Was Never Built for a Basket Trial

By Bhavish Lekh, Co-founder & CEO

The hardest timeline I ever built was not the biggest one.

It was a Phase I combined SAD and MAD study. Single ascending dose cohorts running up one track, multiple ascending dose cohorts starting part-way through on another, and the rule that each MAD cohort could only open once the matching SAD cohort had cleared its safety review. Add site activation that did not line up with the cohorts, sentinel dosing, review periods that stopped one track while the other carried on, and a sponsor who wanted the whole thing compressed. On paper it was a small study. In the spreadsheet it was dozens of dependent dates, every one of them hanging off a start or a stop somewhere else.

Then the sponsor asked the obvious question. What if the MAD track opens one cohort earlier? One change. Dozens of dependent cells. I spent the rest of the afternoon tracing date arithmetic by hand, and I still was not certain I had caught everything before the answer went back.

Trials got more complex. The tools did not.

Over ten years in feasibility I watched trial designs change shape. Dose-escalation studies with sequential cohorts and safety reviews. Basket trials running several indications in parallel. Umbrella and master protocols. Adaptive designs that drop an arm at interim. Crossover studies with washout periods in between.

What did not change was the tool. Feasibility teams still model these designs in spreadsheets, and a spreadsheet is built for one thing: a line of numbers over time. A multi-cohort design is not a line. It is a structure. Some things run in parallel, some things wait for others, and some things only start once a review has finished. Forcing that structure into rows of formulas is slow to build, easy to break and almost impossible to compare. Want to see a second scenario? Duplicate the tab and hope nothing in the first one was hard-coded.

I have lost count of the number of times a version of that spreadsheet went to a budget or project management colleague with a date that no longer matched the one in the specification. Not through carelessness. Simply because two outputs built from the same chain of formulas drift apart the moment one of them is edited.

Two building blocks

When we sat down to design Simulator-C, the question we kept coming back to was how little we could get away with. A trial design tool that needs twenty component types is just a different kind of spreadsheet.

The answer turned out to be two. A patient cohort, which carries the enrollment parameters you would expect: target patients, sites, enrollment rate, screen failure, treatment cycles and activation lag. And a pause or review period, which represents a DLT review, an interim analysis, a safety committee or a washout window, and which either blocks the next step or runs alongside it.

Those two blocks are placed on a grid. Rows are parallel tracks. Columns are sequential stages. Where a component sits is the design. A cohort one column to the right starts later. A cohort on its own row runs in parallel. A blocking review between two cohorts gates the second on the first. There are no arrows to draw and no formulas to maintain. Staggering, gating and parallel arms are all expressed by position, which is how most of us already sketch these studies on a whiteboard.

The dates take care of themselves

Every time anything on the canvas changes, the engine recomputes the full timeline. First Patient In, Last Patient In, First Patient Out and Last Patient Out for every cohort, and the study-level landmarks on top. Move the expansion cohort one column to the left and the study-level Last Patient Out updates immediately, with the trade-off visible rather than buried in a chain of cell references.

The same design is shown in three coordinated views. The canvas is where you compose. The Gantt view lays cohort enrollment bars and review windows over calendar months. The specification view is a printable summary of the cohort schedule, pause periods and chronological timeline events. Because all three come from one source, the numbers match. The date in the spec is the date in the Gantt is the date in the export.

That export matters more than it sounds. The output of a feasibility design is not a picture, it is a handover. Simulator-C builds the deliverable in-app for budget, project management and biostatistics teams, and scenarios save and reload through Aurora Intel project storage so the version everyone signed off is the version everyone can find.

For teams who would rather not start from a blank grid, eleven templates ship with the module: SAD, MAD, combined SAD and MAD, 3+3 dose escalation, Simon 2-stage, basket, umbrella, adaptive 2-arm, 3-arm parallel, crossover and staggered cohorts. In practice most designs I have ever been asked to model are one of those with the numbers changed.

Deliberately deterministic

I want to be clear about what is happening under the hood, because it is easy to assume more than is there. Simulator-C does not make decisions and it does not override anything you set. Every derived date is computed directly from the parameters you entered. Same design, same dates, every time. That is the human-first foundation we build on across the Aurora Suite: rule-based engines designed by people who have done the work, with the judgment left where it belongs.

It is also why Simulator-C and Simulator-L are siblings rather than rivals. Simulator-L answers a question of uncertainty: given a single cohort and a realistic range of assumptions, how likely are we to hit the date? Simulator-C answers a question of structure: given this multi-cohort design, what is the timeline it implies, and how does it change if I rearrange it? One is about probability, the other about shape. Most programs eventually need both, which is why the two are offered together.

What it will not do

Simulator-C will not tell you whether a design is the right one. That is a clinical and strategic judgment, and it belongs to the people making it. What it does is show you the timeline cost of the design you have chosen, instantly and consistently, so that the conversation about alternatives can happen in minutes rather than after a week of rebuilt tabs.

Had I had it on that SAD and MAD study, the sponsor's question would have taken a drag and a glance rather than an afternoon. The overlap between the two tracks, the reviews that paused one and not the other, the cohorts waiting on a site that had not activated yet: all of it would have been visible on one screen instead of scattered across a sheet. More importantly, I would have been able to show my working.

If your team is still planning dose-escalation, basket or adaptive studies in a spreadsheet, the quickest way to see the difference is on a design you already know. Book a demo and bring one with you. I would be glad to rebuild it on the canvas while you watch.