Post-merger R&D integration, explained.

R&D integration explained: what a buyer inherits when the science is the reason for the deal

One acquired program, from closing day to a signed section

1. Find out what you boughtA request list on Day 1: every source, every archive, every retention date. Merrow hands over 38,412 files from seven sources.
2. Decide what each program is nowContinue, change or stop, judged against the buyer's own portfolio and what the deal contract promised.
3. Read the record and compare itFinal reports against summaries, notebooks against reports, minutes against plans. This is where 50 becomes 15, and where a missing assay turns up.
4. Move the regulatory ownershipLetters to the agency, the trial registry and the investigators, with every inherited commitment listed and owned.
5. Run it as one R&D organizationOne committee, one owner per function, and old records kept as they are until someone has a reason to move them.

Four things that catch R&D integration teams out

Four terms an R&D integration file uses on every page

Fourteen terms are defined in the one-page summary.

Why acquired science now decides pharma pipelines: 41% of new drugs and a shift to earlier deals

In shortLarge drug companies get more new medicines by buying them than by inventing them. Deals in 2026 are bigger and aim earlier in development, so more of the price rests on a scientific record the buyer has not yet read.

A wider count points the same way. Of 371 novel drugs FDA approved from 2010 to 2020, large companies originated, developed and marketed 45 (12%) entirely themselves, and secured 228 (61%) through collaborations and external sourcing.9 That paper is a working paper, and we read its abstract.

Published deal totals differ by more than half because each counts something different. EY's first-half 2026 total includes medical devices, and IQVIA counts only majority biopharma purchases of at least $250 million.10,8,11 Compare a number only with one from the same source.

IQVIA's first-half count is 42 deals, or 84% of its 2025 total, at an average of $3.1 billion.11 Each brings science someone has to read. No public source counts how long those integrations take or how many files they hand over. Measure both on your last deal.

Deciding which acquired programs continue: overlap, odds, earn-out promises and a written reason

In shortEvery inherited program needs a decision: continue, change or stop. The odds are published, the antitrust and contract limits are real, and a decision made to meet a savings target can cost more than the savings. Write down why, with the evidence you read.

What a portfolio committee has to weigh for each acquired drug

The base rates a portfolio review starts from

BIO's 2021 analysis, which its authors call the most comprehensive yet, covers 9,704 development programs from 2011 to 2020. Of those entering Phase 1, 7.9% reached approval. Phase 2 was the steepest step, with 28.9% moving on.12 Rates differ widely by disease and drug type, so treat this as the starting line for a discussion.

Overlap: the study behind the phrase "killer acquisition"

Cunningham, Ederer and Ma followed drug projects through acquisitions. Those that overlapped with the buyer's own products were 23.4% less likely to have continued development than drugs bought by a buyer with no overlap. They estimate conservatively that 5.3% to 7.4% of acquisitions in their sample were made to end a competing project, and that such deals cluster just below antitrust reporting thresholds.13 Most stops have ordinary reasons. The finding is why a stop decision on an overlapping program should carry its scientific reasons in writing, dated when they were made.

Earn-outs: when a portfolio review becomes a court record

An earn-out is a later payment owed to the seller if a drug reaches agreed milestones. The best-documented case is Shareholder Representative Services v. Alexion. The buyer agreed to use commercially reasonable efforts to reach later milestones on an acquired drug. Delaware's Court of Chancery measured that promise against a hypothetical typical company of the buyer's size, working on a similar molecule at a similar stage.14

Contract wording decides these cases, and yours will differ. The lesson for R&D is practical. Before any acquired program is cut, ask counsel whether the deal holds an efforts clause, and record what a reasonable company would have considered at that date.

What the science team may see before closing

Until a deal closes, the two companies remain independent. FTC staff wrote in 2018 that clean teams, the small groups allowed to see sensitive data before closing, "should not include any personnel responsible for competitive planning, pricing, or strategy".17 That note concerns diligence between competitors, and the limits on your deal are counsel's to set. In practice, prepare the request list, the dossier format and the names of section owners before closing, and start reading on Day 1. The wider integration guide covers gun jumping in full.

Where the example's reading effort has gone by Day 50

The map carries a warning. Stopped programs are easy to finish because nobody is waiting on them, while the program the deal was done for still has open questions. Reading effort should follow the decision that is due, and that takes a plan.

What an acquired company's scientific record holds, and where GLP raw data, specimens and trial files really sit

In shortThe inheritance arrives from several places, with different authority. Final and amended reports outrank drafts and summaries. Raw data and specimens sit in archives, often at contract labs, and each has a retention clock. Find the archives early: they are harder to find after people leave.

Seven sources behind one fictional inheritance of 38,412 files

LIMS is the laboratory information management system, the database of samples and results. An ELN is an electronic lab notebook. SOPs are standard operating procedures. Our example's sizes are fictional. No public source says how many files a biotech deal hands over, so count yours in the first week.

Which document wins when two say different things

The GLP archive rules a buyer inherits

FDA's GLP rules require that raw data, documentation, protocols, final reports and most specimens be retained in archives "for orderly storage and expedient retrieval". A testing facility may contract with a commercial archive. An individual must be identified as responsible, and material must be indexed.2 The study director must see that everything reaches the archive at the close of the study.21 Each final report must name the study director and the other scientists and supervisors involved.18 A shelf of final reports is therefore also an index of who worked on what.

The GLP clock is "whichever of the following periods is shortest", so a 2019 study on a stopped program may already be past its minimum.22 Nothing obliges a contract lab to keep it longer without an agreement, so ask what it plans to destroy before you assume the material is there.

When a contract lab closes, moves or is itself sold

For clinical trials in the EU, "Any transfer of ownership of the content of the clinical trial master file shall be documented. The new owner shall assume the responsibilities set out in this Article."24 ICH E6(R3), the global good clinical practice guideline, says the sponsor "(or subsequent owners of the data)" retains the essential records and should report any transfer of ownership as the rules require.26

Transferring IND sponsorship after an acquisition: the rule that does not exist and the papers that do

In shortPeople say "transfer of obligations" for at least seven separate acts, each with its own paper. We read all of 21 CFR Part 312 in its 2025 edition (revised as of April 1, 2025, the newest on govinfo in November 2026) and found no section on changing the sponsor of an IND, and no words such as "new sponsor" or "change in ownership".27 The procedures we could read borrow from the rule for approved applications. Decide first whether your deal structure moves anything at all.

This table is partial. Subpart D of Part 312 sets out sponsors' responsibilities in full, and sections 312.50 and following are the list to check duty by duty with your regulatory lead.33

The first question: does your deal structure move the application at all?

If the buyer purchased the target company's stock, or merged it into a surviving subsidiary, the legal entity holding the IND may be unchanged. US registry rules consider "the IND or IDE holder" the sponsor.1 Delaware's Meso Scale opinion held that a reverse triangular merger "was not an assignment by operation of law or otherwise", deciding a license's anti-assignment clause.34 Neither text speaks to INDs.

Whether a move is needed, and when, depends on your structure and on the group's plans for that entity. That is a question for counsel and your regulatory lead, and it comes before any letter is drafted.

Many groups choose to move applications later, so one company holds them all. Moving creates a date. Everything else then has to agree with it: contract notices, registry records, investigator letters and the name on the protocol.

ICH E6(R3) adds one plain duty: the sponsor "should also inform the investigator if sponsorship of the trial changes".26 ICH Q10, the quality system guideline, sets the general principle: when ownership changes through an acquisition, management should ensure "the ongoing responsibilities are defined for each company involved" and "the necessary information is transferred".35

The promises you inherit: turning agency minutes and letters into an IND commitments register

In shortAn application is a stack of promises as well as data. They sit in meeting minutes, responses to questions and cover letters. Pull each one out with its source, give it an owner and a due point, and keep the fixed clocks beside them. None of the clocks restarts for your deal.

The rule for approved applications makes the new owner commit to "agreements, promises, and conditions made by the former owner and contained in the application".3 INDs have no such section, yet the same wording appears in two universities' transfer procedures.4,5 Whatever the paper says, an inherited trial plan rests on what the agency was told and what it asked for.

What belongs in a commitments register, in the fictional MRW-217 file

The fixed clocks that keep running through a merger

Working through a stale NOAEL: exposure margin, starting dose, a metabolite above 10% and a new crystal form

In shortThree short calculations show why comparing a summary with its final report is worth doing by name. The exposure margin to a planned dose falls by more than two-thirds. The starting-dose multiple halves. And the metabolite test has two parts, of which most teams quote one.

Why a final NOAEL can differ from a draft

In the fictional dog study, a second pathologist read the liver slides and agreed with the study pathologist that minimal cell death at 50 mg/kg/day was caused by the drug and harmful. So the no-effect dose fell to the next dose down, 15. FDA says pathology peer review is not required by the GLP rules but can be valuable, and that a draft pathology report can change until it is signed and dated.20 The correction then has to appear as a signed amendment.18 Nothing in that sequence reaches a summary already sitting in a filing.

Calculation one: the exposure margin to the planned Phase 2 dose

A margin compares the drug exposure in the animal at its no-effect dose with the exposure expected in people at the planned dose. Exposure here is AUC, the area under the drug-concentration curve over a day. The example's screens give 68 µg·h/mL at 50 mg/kg/day, 21 µg·h/mL at 15 mg/kg/day, and margins of 6.2× and 1.9×. Those imply a human AUC of about 11 µg·h/mL at 200 mg (our arithmetic).

Whether a margin of 1.9 is acceptable depends on the finding, whether it reverses, and how closely patients are monitored. Either way, it changes the conversation with toxicology, with the agency and with whoever is writing the protocol.

Calculation two: what the correction does to a starting-dose argument

FDA's 2005 guidance converts an animal NOAEL to a human equivalent dose using body surface area, then applies a safety factor of "at least 10". For a 60 kg adult, divide a dog dose by 1.8 and a rat dose by 6.2.38 Take the example's 10 mg first dose and compute how far each study's human equivalent sits above it (our arithmetic).

Fifty-fold is well above the default of 10, so the example's first dose still clears FDA's default. What changes is the most sensitive species, and every sentence written on the old one.

Calculation three: when a human metabolite needs its own animal cover

People quote "the 10% rule" as one test. ICH M3(R2) has two: a metabolite warrants nonclinical characterization when it is "greater than 10% of total drug-related exposure and at significantly greater levels in humans than the maximum exposure seen in the toxicity studies", and such studies "should be conducted to support Phase III clinical trials".39 FDA's 2020 guidance says greater than 10 percent "at steady state".40

Repeat-dose studies a longer trial needs

ICH M3(R2) ties the length of animal studies to the length of the trial: up to 2 weeks needs 2 weeks; between 2 weeks and 6 months, the same as the trial; beyond that, 6 months in rodents and 9 in non-rodents.39 A Phase 2 plan dosing longer than 28 days needs studies the example's 28-day package does not include, which matches what the agency asked for in the sample minutes.

A notebook against a report: Form A, Form B and why a crystal form matters

In the example, a March 2024 report says Form A is the only crystal form seen in a 96-condition screen. A June 2025 notebook entry records a new Form B, with a higher melting onset (186 °C against 171 °C) that suggests it is more stable at room temperature. Both can be true. The report was right when written, and the notebook is later.

This matters because a more stable form can appear on its own and behave differently. The published case is ritonavir: in summer 1998 capsule supplies "were threatened as a result of a new much less soluble crystal form", which the authors explain as a more stable packing that seldom nucleates.41 The right response is to keep both records, ask the chemist before he leaves, and let the CMC owner decide what testing follows.

Key person risk after an acquisition: what the research measures and how to plan the questions

In shortThree studies find inventors leave more often after a deal, and those who stay produce less. Headcount is the wrong measure of the risk. Count the open questions only one person can answer, each with that person's last day, and ask those first.

Each figure measures something different, and none is a forecast for your deal. All three use patent records, which do not show why a person left or whether they retired.

Ornaghi and Cassi add two findings in their summary of the study (we read that summary; we did not read the journal paper). Normally, top-decile inventors are 18.2% less likely than their peers to have stopped patenting five years after their first patent, but during an acquisition their chance of exit rises about 22%, so the advantage largely disappears. Inventors who stay produce about 0.92 fewer patents after three years than comparable colleagues at other firms.43

Evidence outside pharma points the same way: in one study of US technology startups, 33% of acquired workers left in the first year, against 12% of comparable hires (we read MIT Sloan's report of the working paper; we did not read the paper itself).44

Measuring knowledge concentration: who wrote what, in the fictional Merrow record

Three records answer this without an interview. Final GLP reports name the study director and other scientists.18 Notebook entries carry an author and a witness. Archive rules require a named archivist.2 The useful measure is how much of a workstream one departing person wrote.

Writing questions a scientist can answer in an hour

• Start from a gap. "Was M3 measured in any rat or dog plasma sample, even outside the GLP studies?" beats "tell me about the DMPK package".
• Ask for a yes, a number or a location. "Where are the raw toxicokinetic files for rat study TX-004?" has an answer that can be checked.
• Ask what never reached a report. Pilots, stopped arms, reasons for a change of method.
• Book it early, with the owner of the gap in the room. Late in a retention period people have less reason and less time. Start from the last day and work back.

Why an interview answer cannot correct a record

GLP raw data are "the result of original observations and activities" of a study.19 A scientist's memory two years later is useful evidence of where to look, and it stays an interview. Keep it as a dated interview note, linked to the question. If she names a file, go and find the file. The MHRA's data integrity guidance rests on the ALCOA principles: data should be attributable, legible, contemporaneous, original and accurate.45 An answer given today is contemporaneous with nothing in 2024.

Inventors own their inventions until they assign them

As the syllabus of the Supreme Court's 2011 Stanford v. Roche decision puts it, US patent law "has operated on the premise that rights in an invention belong to the inventor".46 An employment clause and an assignment are what move them. Check that every named inventor at the target signed one, for the right company, before those people become hard to reach. An exit interview is the cheapest time to fix a missing signature, with counsel's wording.

R&D governance after close: who signs, which procedures apply, and when to migrate notebooks and LIMS

In shortAfter closing, every decision needs one owner, and old records keep the status they were created under. Reading a record in place carries little risk, while moving it needs validation. Decide what moves after you know what it holds.

Five decisions to settle in the first month

Reading in place against migrating: why order matters

Two notebook systems and two LIMS look like an obvious first job. Migration moves the files without telling you what is in them, and it brings its own duties.

EU GMP Annex 11, the EU's manufacturing rule on computer systems, still at its 2011 revision, asks for checks "that data are not altered in value and/or meaning during this migration process".47 ICH E6(R3) asks that data transferred between systems retain integrity, with reconciliation as appropriate.26 The MHRA warns that the challenges of migrating data "are often underestimated, particularly regarding maintaining the full meaning of the migrated records".45 FDA's electronic records rule, 21 CFR Part 11, requires accurate copies and ready retrieval throughout the retention period.48

A migration is often worth doing, in its own time. Retiring a system saves money, and a lab must keep working. But keep access to the old system, or a validated readable export, until the retention period ends. And plan the move after the readers have finished, since they will find which records matter.

The status a record was born with stays with it

Summarizing a GLP study in your own dossier does not make the summary a GLP record, and a research notebook does not become GLP because a GLP report cites it. Label each inherited record with the status it had when created: GLP, GMP (good manufacturing practice), GCP (good clinical practice), research or agency record. FDA's 2016 proposal for a GLP quality system had no final rule that we could find as of November 2026, so current Part 58 is the text to read.49

Has the acquired company already run this study? Finding reuse before booking new animals

In shortThe cheapest study is the one the seller already ran. Rules and regulators push the same way. But a prior study only helps if you know its status, its exposure and whether it finished. Ask about pilots and stopped work by name.

Three texts support looking first. ICH M3(R2) aims to reduce animal use under the 3R principles (reduce, refine, replace).39 EU Directive 2010/63, Article 46, is headed "Avoidance of duplication of procedures" and has member states accept data from one another, unless further procedures are needed to protect public health, safety or the environment.50 FDA announced a plan on April 10, 2025, to reduce, refine or potentially replace animal testing for monoclonal antibodies and other drugs.51

Five checks before you rely on a prior study

• Did it finish? In the example, a genetic knock-in study stopped at week 3 because the mouse colony was unstable, with no interpretable result.
• What was its status? A research record can guide dose selection. It cannot stand in for a GLP study a regulator expects.
• Does the exposure match? A result at an exposure far from the planned human one answers a different question.
• Was it the same material? Lot, crystal form and formulation. A notebook's Form B may matter here.
• Is the raw data reachable? A slide deck is a lead. Find what sits behind it before you cite it.

We found no reliable public figure for what a repeated study costs. The figure usually quoted, about $3 million for a 3-month GLP package in two species, traces to a teaching deck we could not find, so we do not use it. Ask your contract lab for a quote on the study you are about to book, and use that number in the calculator.

Eight ways an R&D integration fails, and the earliest place each one shows

In shortMost failures are quiet. A number goes unchecked, a person leaves, an archive moves. Each leaves a mark early if someone is looking. No public source counts how often they happen, so measure them on a closed deal of your own.

The published warning about records you bought: a Delaware opinion

In Akorn v. Fresenius, a buyer walked away from a deal over regulatory and data integrity problems the court found at the target. The court called $900 million a rough estimate of the valuation hit: 21% of about $4.3 billion of implied equity value, wider than the cost of fixing the problems. Access to audit trails came after signing, in a whistleblower investigation, and not in diligence.52

Your deal has already closed, so that exit is gone. The point for integration is that a problem hiding in a record can be large, and reading is how you find it while you can still act.

What software changes in R&D integration, and where regulators stand on AI reading a scientific record in 2026

In shortSoftware can read a whole data room, pull a fact with its page, compare versions and count authorship. It cannot read what nobody wrote or decide what a finding means. Regulators' AI texts mostly speak to evidence for a filing, so a buyer's internal dossier sits at the edge of them. Once a value enters a filing, ordinary responsibility applies.

Computer system controls such as audit trails, access limits and electronic signatures support your compliance. None of them makes a product "certified" for GxP or Part 11, and no such certificate exists. Ask each vendor what it has been audited for and read the report.

A hundred days on the R&D integration team: who reads what, and where five weeks per asset go

In shortA small team with day jobs reads what it can, usually in the order people ask. In the example a function lead takes about five weeks per asset to assemble a dossier by hand, a team estimate; no public benchmark exists. In our illustrative split, most of that time goes on finding and checking, with little left for judging.

Who touches an acquired asset's record

Three ways companies staff R&D integration today

Day 50 of the Merrow integration: from a map of the inherited science to a signed toxicology section

In shortThis is the example from the top of the page, inside the product. Priya Ranvelle, the R&D integration lead, sees what has been read and who is leaving. She books Dr. Solheim's walkthrough. Dr. Maya Chen, head of toxicology, sees the 50 and the 15 side by side, takes the final report, and signs. Six agents did the reading; the people decided.

1

All of Merrow's science on one map, shaded by how much has been cited

Each block is a program, split by function. A tile's size is its file count, and its green shows the share now cited in a dossier. Overall, 57% is cited and 14,926 facts carry a page.

Hatched tiles mark work written mostly by someone who is leaving. MRW-217's DMPK tile says "Only author · 19 days left". The side panel explains: Dr. Ingrid Solheim wrote all 212 M3 assay notebook entries, and two open gaps need her.

2

Replaying the integration since closing shows what moved and what did not

Replay since close runs the map from Day 0. By Day 28 the map is at 32% cited, and the timeline marks milestones above and departures below. Tiles go from gray to green as the record is read.

The DMPK tile stays pale and hatched. That is the point of the view: progress overall can hide the one place it matters.

3

Booking the only author, with four questions she can answer

Priya opens the booking and picks Wednesday, October 14, twelve days before Dr. Solheim leaves. Sam Patel, who owns the gap, joins. Four short questions go with the invite, each tied to a gap. The first asks whether M3 was measured in any rat or dog plasma sample, even outside the GLP studies.

The note at the bottom matters: answers are kept as an interview and linked to the gaps, and they never become a source record. After booking, the tile reads "Walkthrough Oct 14".

4

The IND summary and the final report, side by side

In MRW-217's toxicology section, the agents raise a disagreement. The IND summary from November 2023 says 50 mg/kg/day, based on a draft. The final report's Amendment 1 from April 2025 says 15, backed by the pathology peer review. The recommendation and its reason are shown, and both numbers keep their citations.

Only the section owner can decide. Dr. Chen takes the final report. "Keep the IND value" stays available, and a person must write a reason to use it.

5

The section rewrites itself around the decision, and the owner signs

Every sentence that used 50 is rewritten, and the margin now reads 1.9×. A new gap goes to Dana Okafor, the regulatory lead, to correct the IND summary, scheduled on the screen for the next annual report in January 2027. That timing is the sample's choice, and this guide does not endorse it. FDA names "new toxicology, chemistry, or other technical information" as an example for an information amendment, so the regulatory lead decides how and when the agency hears.37

Dr. Chen signs with a stated meaning, "I reviewed the cited evidence and approve this section for the integration dossier", and her password. A section with an open disagreement cannot be signed.

The data room, the page behind each fact, the commitments, the study reuse check and the people leaving

In shortThe rest of the work lives on these screens. Each fact opens its page, each disagreement shows both sides, each gap has an owner and each departure has a date. Select any picture to enlarge it.

The inherited record, read in place

Disagreements a person has to settle

Regulatory ownership: the letter, the minutes and the register

Reuse, gaps and people

The six agents that read an acquired record, and the function owners who sign for it

In shortEach agent has one job and one limit. None signs, none changes a source record, and none contacts a scientist without the integration lead. The people below hold every decision.

Who holds which decision on an acquired asset

Controls that support your compliance on inherited records

Inherited GLP, GMP and GCP records are read in place, never copied, relabeled or changed. Every read, decision, booking and signature goes on the audit trail. Acquired-company experts see only their own question pack. Roles keep the maker and the checker apart, and a section with an open disagreement cannot be signed.

These controls support your compliance with your own rules for records and signatures. They do not make any record GLP or GMP, and your quality unit decides how to qualify the system for your use. The dossier is a research summary that cites its sources, and it goes to the agency only if your regulatory lead puts its content there. Botminds' own credentials are SOC 2 Type II audited and information security managed to ISO 27001.

From a data room export and notebooks to signed asset dossiers, gaps with owners and a searchable science registry

In shortIt reads the exports you already have, read only, and returns one dossier per asset plus the lists an integration team works from. Timing below is a design target; the sample's own team estimated the hand-built figure.

The solution also carries a second design target: every departing author interviewed before their last day. On your deal the result depends on when you start. In the example, Dr. Solheim's walkthrough is booked with 19 days left, later than the product's default 30-day reminder.

Connects to the systems an R&D organization already runs

Virtual data rooms by export; electronic lab notebook exports; laboratory information management systems; contract lab file shares; regulatory archives; retention dates from HR; and your calendar for booking. Sources connect read only, across every program in a deal, and roles control who sees what, so acquired-company experts see only their own question pack.

What a late R&D integration finding costs you: delayed trials, repeated studies and a reading bill nobody budgets

In shortNo public study prices a failed R&D integration. What can be priced is a day of trial delay, an efforts breach in one case, and your own team's reading time. The costs that matter most, such as a regulator's trust, do not fit in a calculator.

Pricing your own last acquisition in a week

Costs of a badly read inheritance that no spreadsheet counts

Your return on reading every inherited file: reviewer weeks, one study not repeated, and days of trial delay

In shortThree levers count: reading time, repeated studies and trial delay. At the default inputs, reading time alone does not repay one four-asset deal. A single avoided study or a few avoided days can change that, but those are the numbers to be most skeptical of, so run the calculator with them at zero first.

Where the value comes from when an acquired record is read in days

The calculator: your assets, one deal

How ready your R&D integration is: access, owners, retention dates, rules for disagreement and security

In shortEight questions, each scored 0 to 2, and one to reflect on. Answer for the deal you have closed or expect to close. Each question says why it matters and what a strong answer sounds like.

A low score usually means naming section owners before anything else

What to ask an R&D integration software vendor or consultant about citations, disagreements and acquired records

In shortFifteen questions in five groups. Ask us too. Each says why it matters, a strong answer and a warning sign.

Test it on one asset from a closed deal, where your team already knows what surfaced late

In shortChoose one asset from your last deal. Your people already know what they found, and when. In eight weeks you can tell what was caught, what was missed, how long signing took and what the function owners thought.

What to measure on the pilot asset

Post-merger R&D integration: answers to what R&D and regulatory teams search for

In shortEach answer stands alone, so you can send one to a colleague.