Randomize by generating the allocation list before the first animal is enrolled, with a method that leaves chance in charge and a block structure nobody at the site can predict. Conceal it by keeping the list away from the people who enroll, so the arm for an animal is revealed only after that animal is confirmed eligible and its owner has consented. Keep the blind by giving the owner, the assessor and the investigator products that look the same and no way to see the arm, by telling the owner in the consent form whom they may discuss treatment with, and by making every code break a logged event with a reason, a name and a time. The rest is writing it down in the protocol first and keeping the record afterwards.
What the rules ask
For a study meant to support FDA approval of a new animal drug, 21 CFR 514.117 defines an adequate and well-controlled study. Paragraph (b)(4) asks for a design that permits a valid comparison with a control: a placebo, no treatment, an active treatment or, in special circumstances, a documented natural history. Paragraph (b)(6) asks for "a method to assign a treatment or a control to each experimental unit of animals that is random and minimizes bias", with groups comparable on age, sex, class of animal, severity and duration of disease, diet and level of production. Paragraph (b)(7) asks for "methods to minimize bias on the part of observers and analysts" and says the protocol and the study report must document the methods, "such as 'blinding' or 'masking'", used to exclude bias from the observations. Paragraph (c)(2) applies the same to field studies: an appropriate control, procedures to minimize bias, and observations recorded in enough detail to be evaluated.
21 CFR 511.1 is the rule that lets an unapproved new animal drug be shipped for a clinical investigation. It says nothing about randomization or blinding. What it adds is the file: the notice before shipment in (b)(4), the investigator's complete records kept for two years in (b)(7), the sponsor's monitoring and hazard reporting in (b)(8), and the bar in (b)(8)(iv) on representing the drug as safe or effective while it is under study.
FDA Guidance #85, which is VICH GL9, is the standard of conduct. It defines randomization as assigning animals "using an element of chance to determine the assignments in order to reduce bias" (1.23) and blinding as a procedure in which "designated study personnel are kept uninformed of the treatment assignment(s)" (1.5). The protocol checklist asks for the randomization method and the practical arrangements for allocating animals (6.3.8.3), the experimental unit (6.3.8.4), and the extent and methods of blinding together with "the provisions, including procedures and personnel, for access to treatment codes" (6.3.8.5). The investigator manages "randomization envelopes, blinding information" with professional care and breaks a code only in accordance with the protocol and with the sponsor's knowledge and consent; people who cannot be blinded take part "to the minimum extent necessary" (3.2.18). Locating and breaking a code so an animal can be treated is a planned adverse event procedure (6.3.19.2). The treatment assignment itself is raw data (8.2.2.1).
The EMA guideline on statistical principles for veterinary clinical trials, EMA/CVMP/EWP/81976/2010-Rev.1, spells out the mechanics. Allocation concealment means "keeping investigators and animal owners unaware of upcoming treatment assignments", and without it "randomisation might become corrupted" (4.3.1). The sequence and the block length should not be in the protocol, and the schedule "should be filed securely by the sponsor or an independent party". The guideline prefers a double-blind trial in which the investigator, the staff, the owner and anyone assessing eligibility, endpoints or compliance are unaware of the arm. A single-blind trial must say who is blinded, whether the owner or carer is, and from what stage (4.3.2). Breaking the blind is for when knowledge of the arm "is deemed essential to the veterinary care and welfare of the study animal", and every break, intentional or not, is reported and explained at the end of the trial.
Draft FDA Guidance #282 on informed consent forms carries the owner's side. The form should describe the treatment groups, including the control, and give the likelihood of being assigned to the control (VI.A.2). Among the owner's responsibilities it lists: "To maintain masking, only discuss treatment information with pre-specified study personnel" (VI.A.4.a). It asks for a plain statement that the animal may be in a control or placebo group (VI.B.4). It is a draft, so current thinking rather than a requirement.
Where the allocation list and the code-break log are electronic records kept under an FDA records requirement, 21 CFR Part 11 applies to them: system access limited to authorized individuals (11.10(d)), a secure, computer-generated, time-stamped audit trail whose changes do not obscure earlier entries (11.10(e)), and authority checks on who may alter a record or perform an operation (11.10(g)).
How the design changes the answer
- Observational study, nothing assigned. There is no allocation to conceal, and 514.117 and 511.1 do not apply. Blinding the person who scores an outcome to the animal's exposure is still a design choice that strengthens the result, and the method belongs in the protocol.
- A marketed product used as labelled. No INAD. If the study is meant for FDA, 514.117 and GL9 apply as above. If it is meant for a journal or a practice's own decision, the rules are the protocol and the ethics review, and the same methods are what a reviewer looks for.
- An unapproved product, including a compounded product or a research-use peptide. An INAD under 511.1 before shipment, 514.117 for the design, GL9 for conduct, and Part 11 for the electronic records. Whether a supplement study falls here turns on the product's intended use, which is a question for the sponsor and counsel.
Randomize and conceal, in order
- Write the scheme in the protocol: the method, the ratio, the experimental unit, any stratification, and who holds the list. Leave the sequence and the block length out of the protocol.
- Generate the list before the first animal is enrolled, from a documented procedure, so it can be reproduced later.
- File the list with the sponsor or an independent party, not at the site.
- Enroll in order: confirm eligibility, record the consent, then reveal the arm for that animal and no other.
- Record the assignment as raw data, with who revealed it and when.
- Write the code-break procedure before the study starts: who may ask, who approves, what is written down, and what happens to that animal's data afterwards.
Keeping the blind with owners at home
- Supplies look alike. Matched product and placebo, labelled with kit numbers rather than arm names. Where the two formulations cannot be matched, a double-dummy design gives every animal one active and one placebo.
- The owner's diary never shows an arm, and its questions are worded so an answer cannot reveal the product.
- The consent form says whom the owner may talk to about treatment and names that person. What the form should contain is in Informed consent for client-owned animal studies.
- Where the product cannot be masked from the person giving it, the dispenser and the assessor are different people, and the dispenser scores nothing.
- The dosing record names a kit, not an arm.
- Emergency unblinding has a route the owner can reach at any hour. The break is logged with its reason, the sponsor is told, and the animal's later data are flagged.
What the protocol section must contain
| Item | Where it is asked for |
|---|---|
| The overall design and the control | 514.117(b)(4); GL9 6.3.8.1 and 6.3.8.2 |
| The randomization method and practical arrangements | 514.117(b)(6); GL9 6.3.8.3; EMA 7.1.1 |
| The experimental unit | GL9 6.3.8.4; EMA 5.4 |
| Who holds the list and how allocation is concealed | EMA 4.3.1 |
| The extent of blinding, who is blinded, from when | 514.117(b)(7); GL9 6.3.8.5; EMA 4.3.2 |
| Access to codes, the code-break procedure and its record | GL9 3.2.18 and 6.3.19.2; EMA 4.3.2 |
| What the owner is told | GFI #282 VI.A.2, VI.A.4 and VI.B.4 |
Whose job it is, and where a data system fits
The protocol, the choice of control, the ethics review, the consent text, the INAD file and any report to FDA stay the sponsor's and the investigator's. A data system holds the list, reveals one arm at a time, hides it from the people the protocol says are blinded, and keeps the log. Signalment randomizes in permuted blocks with concealed allocation, supports single or double blinding, and logs each code break; the sponsor page shows the parts in use. It does not yet keep a drug accountability ledger, so the reconciliation of kits against the allocation stays in the investigator's own records.
Sources
- 21 CFR 514.117, adequate and well-controlled studies: https://www.ecfr.gov/current/title-21/section-514.117
- 21 CFR 511.1, new animal drugs for investigational use: https://www.ecfr.gov/current/title-21/section-511.1
- 21 CFR Part 11, electronic records and electronic signatures: https://www.ecfr.gov/current/title-21/part-11
- FDA Guidance #85, VICH GL9, Good Clinical Practice: https://www.fda.gov/media/70333/download
- FDA draft Guidance #282, Informed Consent Forms for Studies that Enroll Client-Owned Companion Animals: https://www.fda.gov/media/172056/download
- EMA, Guideline on statistical principles for clinical trials for veterinary medicinal products (pharmaceuticals), EMA/CVMP/EWP/81976/2010-Rev.1: https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-statistical-principles-clinical-trials-veterinary-medicinal-products-pharmaceuticals-rev-1_en.pdf