A note from our COO

Find the gaps before you sign

I run operations at Biobase. Programs rarely stall on the science. They stall on contracts, compliance and delivery, or on data that looks complete and isn’t. This issue shows three ways that happens. We track every case from source to your team, so gaps surface before you sign. We deliver in weeks, not months. Let’s talk about where you’re stuck.

Terence O’Keeffe

Terence O’Keeffe

Chief Operating Officer, Biobase

Book time with Terence

The Biobase Brief, issue 03 · November 2026. Adapted for the web from the Life Sciences edition.

Before sourcing, check three gaps: whether a biobank cohort resembles the patients a study targets, whether follow-up comes from more than one record system, and how biomarker fields were filled. Three recent studies show each gap can be large.

Do biobank donors look like typical patients?

Not always. Researchers in Taiwan linked liver cancer patients who had donated to a biobank to national cancer registry and claims data, and compared them with patients who had not. Of 4,790 donors, 94.8% had surgery, against 50.7% of 126,547 non-donors. Donors were also more often diagnosed at stage I or II (78.6% against 57.2%) and had longer median survival (7.15 against 1.92 years).

The analysis is exploratory and covers one country. Survival was not adjusted, and the gap reflects how donors were recruited.

Before assuming a biobank cohort matches a trial population, ask for its stage and treatment mix against a registry baseline.

Chen et al., Clinical Epidemiology 18 (2026).

How much follow-up does one EHR capture?

Less than a cohort may assume. An EHR (electronic health record) holds the care a patient received within one health system. Researchers at an academic health system linked EHR records for adults aged 65 and over with type 2 diabetes to Medicare claims. The EHR alone captured 38.3% of the encounters and 68.2% of the deaths found in claims. The linked EHR and claims data captured all of them.

The study covers one health system and one condition, and claims carry their own coding errors.

For outcomes-linked data, ask whether follow-up comes from one EHR, or from an EHR plus claims or a death index.

Lee et al., JAMIA Open 9(3), ooag102 (2026).

Do FHIR extracts include biomarker results?

Often not. FHIR (Fast Healthcare Interoperability Resources) is the standard many EHR systems use to export data through an API. At one NCI-designated comprehensive cancer center (NCI is the US National Cancer Institute), researchers reviewed the records of 30 lung cancer patients and traced their data into FHIR extracts. Between 80 and 100% of the biomarker results documented in the EHR were missing from the extracts, while demographic information transferred accurately. EGFR, KRAS, and BRAF results were missing for every patient in the first cohort. Manual review of the free-text “Lab Text” field in the hospital’s cancer registry recovered 10 to 50% of the missing biomarkers.

The study covers one cancer center and 30 patients.

Ask how a dataset’s biomarker fields were filled: structured extract, natural language processing (NLP) of the notes, or chart abstraction by a person.

App et al., JAMIA Open 9(3), ooag052 (2026).

What should a request ask about a dataset?

Three questions cover the gaps above:

  • Who is in the cohort? Its stage and treatment mix, compared with a registry baseline for the indication.
  • Where does follow-up come from? One EHR, or an EHR linked to claims or a death index.
  • How were biomarker fields filled? Structured extract, NLP, or chart abstraction.

Put them in the request, so a supplier answers them before a quote. The biospecimen request guide brings cohort, material, and linked-data requirements into one brief, and a feasibility count shows what matches before you commit.