One record carries the study from the pasted protocol to the bioequivalence answer, and the business around it. No retyping, no binders, no midnight spreadsheet quote.
From the first sponsor approach to the final report and the work that follows it, the CroW moves the study through every stage in order, so nothing is skipped and nothing waits in an inbox.
The engine reads the protocol and drafts the design, the randomisation, the sampling schedule, the eCRF domains, and the analysis-plan rules, then opens each one as a real document you can read, edit, and print. Every field shows where it was read from and how sure the read is.
A team rebuilds the CRF, the consent, the randomisation, the SAP, and the worksheets from the protocol. One real account puts it at two months.
One paste, and the study spine plus every document the study and your CRO require are drafted in minutes, ready for review and sign-off.
Vitals, dosing, blood samples, adverse events, concomitant medication, and deviations are captured against the study and the subject, and every value runs range, protocol-window, and consistency checks as it is typed. A validation brain then reads across records for what a single field cannot show, a sample drawn before its dose was recorded, a duplicate timepoint, a cohort outlier caught on the median and MAD rather than the mean, which an outlier hides in. Nothing moves to analysis until the subject is source-verified and locked, and corrections keep the original. A batch from an instrument is checked the same way on the way in, and a study query list is drafted for the data manager with a suggested resolution on each line.
Data on paper, keyed later, queried days afterward by a server you cannot see, and priced for large pharma.
Checks at entry, in the browser, offline. A study that locks only when every subject is verified, and only then flows to the verdict.
Sample size and power, non-compartmental analysis, the 90 percent confidence interval against 80.00 to 125.00, the full crossover ANOVA table with the sequence, period and formulation terms a reviewer asks for, a distribution-free Wilcoxon and Hodges-Lehmann treatment for Tmax, reference scaling for highly variable drugs, two-stage adaptive design, the f2 biowaiver, and stability shelf-life. Every engine is checked against its published reference, and a stress harness re-runs sixty-two invariant checks across a thousand randomised studies against the exact shipped code.
PK in one tool, statistics in another, the bioanalytical numbers retyped in between, and the verdict reached three programs later.
One record runs from clean data to the verdict, with no retyping and the maths traceable.
The whole operation as a graph you can walk. Studies, sponsors, volunteers, samples, and billing milestones link the way they really relate, and work moves along the edges as it happens. A completed study freezes into an archive you can still open years later. Click a node to open its details, hover to trace its connections, or drag one to reshape the layout.
Status lives on a whiteboard, in an inbox, and in the project manager's memory.
One graph everyone reads the same way, with archives that outlive the study and monitoring that lights up as it runs.
Every action is signed with a name, an employee ID, a designation, and a timestamp. Each timestamp carries its own trust label, server-synced under the hosted app or an honest untrusted local clock offline, and never claims to be more than it is. QA issues each document set with a digital stamp and a copy number. The audit trail is hash-chained with SHA-256 over a canonical form, so it breaks the moment it is touched. A correction chains to the latest entry for that field, so a correction of a correction stays unbroken and in order. Controlled prints are logged. Excluded data goes into a register and is kept, never deleted.
Paper binders, wet signatures, a stamp pressed by hand, and an audit trail you have to take on trust.
Signed, stamped, copy-numbered, hash-chained, and exportable, with every exclusion explained on the record.
From 7 March 2026 the amendment replaced prior approval with a prior-intimation route through the National Single Window System for lower-risk BA/BE studies, and kept approval for named higher-risk categories. Filing for approval when an intimation would do costs weeks. Starting on an intimation when approval was required is a breach.
Enter the category, the purpose, and the site status. The console returns the route, lists every rule it applied and what each one decided, and separates the route from the blockers, because a missing Ethics Committee approval stops the study without changing which form you file. It then builds the filing checklist and holds the determination against the study record, so the answer to why a study started on an intimation is on file rather than in somebody's memory.
A judgement call made from a circular, repeated per study, with nothing written down explaining it.
A recorded determination naming the rule, the inputs, the person and the time, ready for the day an inspector asks.
Every BA/BE centre in India lives with the Ethics Committee, and almost no software holds it properly. It becomes a scanned letter in a folder and a date somebody remembers. Then a study doses on an approval that expired last month, or on a protocol version the committee never saw, and it is found at inspection rather than at the bedside.
The committee is a registered body with members, not a checkbox. The console holds its registration and its composition, and checks that composition against what the rules prescribe: at least seven members, an independent chairperson, a member secretary, a basic medical scientist, a clinician, a legal expert, a social scientist, a lay person, at least one woman. Each requirement carries the reason it exists, so a gap reads as something to fix rather than a code.
An approval is tied to one study and one protocol version, for a bounded period. Then the part that matters: the eCRF asks before it will accept a capture. No approval, an expired one, a suspended one, an approval for a different protocol version, an overdue continuing review, or a committee whose own registration has lapsed, and the capture is refused. Nothing is written. The refusal names which of those it was.
A system that faithfully records that you dosed a subject without a current approval.
A system that does not let you, and says why on the screen where you were about to do it.
And the clocks. Continuing review runs from the last review, warns as it approaches and blocks once it is overdue. A serious adverse event starts the two reporting clocks the rules set, twenty four hours for the initial report to the sponsor, the committee and the licensing authority, then fourteen days for the detailed analysis. Both run from when the event became known, not from when somebody got round to typing it in, which is the distinction an inspector asks about.
Where the line is. The composition rules are encoded from the New Drugs and Clinical Trials Rules 2019 and the ICMR guidelines 2017, and the module tells you on screen to confirm them against the gazetted text before a filing relies on them. Recording an approval in the console is not the same as the committee granting one, and it does not pretend otherwise. It records what the committee decided, and then holds you to it.
Sign-in is a named account with an employee ID and one fixed role that cannot be switched afterwards. Passwords are checked against a PBKDF2-SHA-256 derivation with a per-user salt, five failures lock the account, and the session ends when it goes idle. Signing a record re-verifies the password rather than trusting the open session, and the signature carries the account holder's identity rather than a name typed into a box. A hash-chained audit trail proves a record was not altered; the account is what proves who made it.
An audit trail that proves integrity but attributes every entry to a role anyone at the desk could select.
A named, employee-numbered person on every entry, signature and determination.
Two configurations, and we say which one you are in. Run the console on its own and it gives you attribution: every entry names a real person with an employee ID and a fixed role. That is honest and it is not access control, because a file you can edit cannot enforce anything. Run the sync server alongside it, one process on your own machine or your own India-region server, and the decision moves somewhere nobody using the console can reach. Your role and every permission are checked there. Two coordinators cannot silently overwrite the same subject-period, because the second one is told who is holding it. Timestamps come from a clock the user did not set.
The console states which of the two is in force on the sign-in card, checked live rather than asserted, so nobody has to take our word for it. That is the sentence to put in front of a quality unit: the deployed configuration enforces access control, controls concurrency and issues its own timestamps. The offline file remains what it always was, the thing you can evaluate in an afternoon without involving IT.
Set the study and the build-up appears across clinical, bioanalytical, data and statistics, and pass-throughs, with the CRO cost beside the sponsor price on every line. It reads live capacity from operations and, when clinical beds or bioanalytical is the binding constraint, suggests a yield adjustment you confirm before it touches the total. Put in the sponsor's target budget and it tells you whether the margin survives, and which levers recover it if not.
A spreadsheet quote, a margin half-guessed, and a single number shown to the sponsor.
A transparent build-up, the margin you actually keep, and a clean quote in one click.
Open the app and the console runs on your own machine, saving every change to a file you control. Nothing leaves the building. Refresh the page, restart tomorrow, and your work is still there.
Files scattered across drives and inboxes, the latest version anyone's guess.
One local store, saved automatically, with a backup of the last save kept beside it.
Sample size, power and non-compartmental analysis are not things you take on faith. We ran PowerTOST and NonCompart, the reference implementations regulators and biostatisticians already recognise, across seven study designs and the whole coefficient of variation range, and stored what they produced. The console reproduces 610 of 610 of those values, and it re-runs the whole set every time it loads.
Then press the button marked Prove the harness can fail. It bends each engine by a known amount, shows you the checks going red, and puts them back. Anyone can display a green tick. Being able to make it go red on demand is the part that means something.
A tool that says it is validated, with nothing you can run yourself to check.
A qualification report you generate on your own machine, naming every check, every deviation, and the package version it was compared against.
Where the line is. Those reference packages are licensed in a way that would not let us ship their code, so we did not. We ran them and kept the numbers. Numbers are facts and facts carry no licence. Nothing of theirs is inside the product.
One click drafts a risk-based validation pack from the live system. The user requirements, an FMEA risk register, and a traceability matrix mapping each engine to its published reference. The artefact your quality team finishes, instead of starts.
Months assembling the computer system validation file from a blank page.
A structured pack generated from the build, ready for QA to extend.
Run one of your own protocols through the console and watch the design, the documents, the analysis, and the verdict appear inside the same hour.