Your site is ready to enroll. The screening call came back in Spanish, the participant is eligible, and the enrollment window on this cohort closes in eleven days. The translated informed consent form came back from a vendor last week with a certificate of accuracy stapled to it, and you forwarded it to the IRB assuming that was the box being checked. The IRB deferred it.
Here is what most sponsors and site teams have backwards. There is no FDA regulation requiring that an informed consent form be translated by a certified translator. Read all of 21 CFR part 50 and you will not find the phrase. What the regulation requires instead is something easier to state and considerably harder to satisfy — and a certificate of accuracy, on its own, does not satisfy it. That gap is why translated consent forms get deferred, and why the fix is a process rather than a purchase.
What the Regulations Actually Say
The operative sentence is short enough to quote in full. 21 CFR 50.20, the general requirement for informed consent, ends with this:
The information that is given to the subject or the representative shall be in language understandable to the subject or the representative.
That is the whole standard on language. It names no credential, no vendor, no method, and no standards body. It is written as an outcome — the person understood — rather than as a procurement specification. Everything an IRB asks for downstream exists because somebody has to show that outcome was pursued in good faith, and a document written in a language the reviewer may not read is difficult to show anything about.
Three other provisions do the structural work:
| Provision | What it requires | What teams misread |
|---|---|---|
| 21 CFR 50.20 | Legally effective informed consent, sought under conditions that allow real deliberation, with the information given “in language understandable” to the subject or representative, and no exculpatory language. | It sets an outcome, not a credential. Hiring a certified translator is a means, not compliance itself. |
| 21 CFR 50.25 | The basic and additional elements of informed consent that must be provided to the subject. | The translated version needs all of them. A condensed or summarized translation is a different document, not a translation. |
| 21 CFR 50.27(a) | Consent documented on a written consent form approved by the IRB, signed and dated by the subject or legally authorized representative at the time of consent. | Approval attaches to the document the participant actually signs. If they sign the Spanish version, the Spanish version is the regulated document. |
| 21 CFR 50.27(b)(2) | The short form pathway: a short written form stating that the required elements were presented orally, with a witness to the oral presentation. | It is an exception for the unanticipated case, not a way around translating a form you knew you would need. |
| 45 CFR 46.116 / 46.117 | The Common Rule equivalents governing federally supported research, including the parallel short form provision at 46.117(b)(2). | Many trials sit under both frameworks. Satisfy the stricter reading, not the more convenient one. |
The consequence of 50.27(a) is the one worth writing on a whiteboard: IRB approval attaches to the signed document. An English ICF approved in March does not confer approval on a Spanish translation produced in August. The translation is a submission, not a formatting change.
ICH E6(R3), the current good clinical practice guideline, comes at the same point from the ethics side rather than the regulatory side: it frames consent as a comprehension obligation and asks sponsors to weigh the characteristics of the trial population and the suitability of the consent method when developing consent materials. It does not hand you a prescribed translation recipe either. Neither framework does. That is precisely why sponsor SOPs and IRB checklists have filled the vacuum with back-translation, reconciliation, and certificates of accuracy — those are the profession’s answer to an evidentiary problem the regulations left open.
The Short Form Is a Safety Valve, Not a Translation Plan
The short form process at 21 CFR 50.27(b)(2) and 45 CFR 46.117(b)(2) is widely misread as “the option for languages we did not translate.” It is narrower than that. The participant signs a generic form stating that the required elements of consent were presented orally; the full IRB-approved English consent is presented orally, through an interpreter; and there must be a witness to that oral presentation. The short form itself carries no study-specific information at all — typically just headers matching the required elements.
How institutions operationalize it is instructive. NIH’s human research protections guidance reserves the short form for the unanticipated or unexpected enrollment of a non-English-speaking participant when no IRB-approved translated consent exists, and does not permit it where there is sufficient time to obtain a translated document. After enrollment by short form, the consent must be promptly translated and provided to the participant. And for minimal-risk research, NIH sets a hard stop: after three short-form uses in the same language, translation becomes mandatory for subsequent participants in that language.
That last rule exists because repeated short-form use is self-incriminating. Each instance you log is evidence that enrollment in that language was, in fact, anticipated — which is exactly the condition under which the exception stops applying.
Need an ICF in a language your protocol did not plan for? We translate informed consent forms, protocols, and patient-facing materials with ATA-certified linguists and independent revision — plus back-translation and reconciliation where your SOP or IRB requires it, and a certificate of accuracy for the regulatory file.
What Actually Gets a Translated ICF Deferred
Across life sciences work the same five patterns account for most translation-related consent problems. None of them are exotic, and all of them look cheaper than the alternative at the moment the decision gets made.
1. The bilingual site coordinator translated it
The coordinator who speaks the language is also the person recruiting the participant. That is a role conflict before it is a competence question — and it is a competence question too, because consent documents carry regulatory, statistical, and clinical terminology that fluency alone does not cover. The same reasoning that makes bilingual staff a poor substitute for qualified interpreters applies with more force to a document somebody is going to sign.
2. Raw machine output with a certificate attached
A certificate of accuracy is a representation about a process. Signing one over unedited machine translation misrepresents that process. Machine translation has a legitimate place in a controlled workflow — ISO 18587 exists precisely to define full post-editing of machine output by a qualified human linguist — but raw output is a draft, not a deliverable, and a consent form is the last place to discover the difference.
3. Reading-level drift
Sponsors work hard to bring English ICFs down to a lay reading level, then hand the file to a translator with no instruction about register. It comes back grammatically perfect and pitched several grades higher. Understandability under 50.20 is a per-language property; it does not transfer from the English source.
4. Only part of the document was translated
Signature pages and key-information summaries get translated; appendices, optional sub-study language, and injury-compensation sections do not. Those sections carry required elements under 50.25.
5. Treating certification as the finish line
A certificate documents who translated the file and attests to accuracy. It says nothing about whether the target population can read the result. Both matter, and the second one is the one the regulation actually names.
A Three-Step Process That Holds Up Under Review
None of this requires a large program. It requires deciding three things earlier than most teams decide them.
1. Fix the language set at protocol design, not at screening. Your site list and catchment demographics already tell you which languages are coming. Deciding at design time converts an emergency short-form event into a planned deliverable on a normal timeline — and translation is one of the few startup activities that runs fully in parallel with everything else.
2. Translate through a controlled, documented process. ISO 17100 is the relevant standard for translation services: a qualified translator followed by independent revision from a second qualified linguist — not the same person re-reading their own work. Where the sponsor SOP or the IRB wants evidence of conceptual equivalence, add back-translation by a linguist who has not seen the source, then a reconciliation step that documents and resolves each discrepancy. Where machine translation is in the loop at all, run it under ISO 18587 full post-editing and say so in the file.
3. Build the evidence package as you go, not at submission. The regulatory file should hold the certificate of accuracy, the linguists’ qualifications, the reconciliation record where one exists, and — the part most often missing — an explicit version linkage tying each translation to a specific approved English version and protocol amendment number.
To be plain about what this does and does not do: no translation process guarantees an IRB approval, and any vendor promising one is selling something they cannot deliver. What a documented process does is remove translation as the reason for a deferral — which is the part of that decision you control.
The Amendment Trap
The failure mode that costs the most is not the first translation. It is the fourth one.
The English ICF is version 4.0. The Spanish ICF on the site’s shelf is a translation of version 3.0. Nobody noticed, because the file name did not change.
Every substantive ICF amendment orphans every translation of it. If participants have already signed, the question stops being administrative: it becomes whether re-consent is required, and whether participants in one language received materially different information than participants in another for some stretch of the study. That is a protocol deviation with a data-integrity tail, and it tends to surface during inspection rather than during enrollment.
The prevention is unglamorous. Treat the translated ICFs as dependents of the English master in version control, so an amendment automatically opens translation tasks instead of relying on someone to remember. Sponsors running several languages across several sites need that as infrastructure, not as vigilance.
What Failure and Success Look Like
Failure is rarely a regulatory finding first. It is an enrollment window closing while a submission sits in deferral, a site quietly stopping screening in a language because the paperwork is too hard, and a cohort whose demographics no longer resemble the population the therapy is for. The worst version is a participant who signed a document that did not say what the English said — which means re-consent, a deviation to report, and a question at inspection about every other translated document in the trial.
Success looks boring. The language set was decided at design. Each ICF was translated by a qualified linguist and independently revised, with reconciliation where the SOP called for it. The certificate, the qualifications, and the version linkage are in the file. The short form is still available — for the participant nobody could have predicted, which is what it was written for.
Language Access Hub, powered by Taika Translations, supports sponsors, CROs, and research sites with clinical trial translation across the documents a study actually generates — informed consent forms, protocols and synopses, patient-facing materials, and regulatory submissions — alongside broader life sciences translation and certified translation by ATA-certified linguists. We work to an ISO 17100 process with independent revision, offer back-translation and reconciliation where your SOP or IRB requires it, and provide a certificate of accuracy with version linkage for the regulatory file. We hold a GSA Schedule and NASPO ValuePoint contracts, are SAM registered, and are an SBA-verified Veteran-Owned Small Business, so publicly funded research can transact through vehicles procurement already recognizes.