More data, more answers? A 1.9 million-person health study shows why scale has limits

One of the world’s largest health cohorts provides researchers with an unusually detailed view of disease across Britain, but comparisons with national surveys and established biobanks reveal that sample size is only part of the picture.

Study: Phenomic profiles and disease patterns of 1.9 million participants from Our Future Health. Image Credit: Shutterstock

In a recent study published in the journal Nature Medicine, researchers evaluated baseline phenotypic data from more than 1.9 million adult participants enrolled in the Our Future Health (OFH) study.

The analysis focused on OFH participants from England, Scotland, and Wales and assessed self-reported medical histories, physical measurements, and linked electronic health records (EHRs) to map disease patterns against UK national benchmarks.

The findings showed that OFH broadly reflects previously reported national demographic patterns, but younger adults, most minority ethnic groups, and people in the most deprived areas remained proportionally underrepresented. Its large scale still provided unusually high absolute numbers in several underrepresented groups. Prevalence rates across 109 common health conditions strongly correlated with data from the UK Biobank (UKB; Pearson correlation coefficient r = 0.784).

The authors concluded that the OFH cohort already provides a major population resource for biomedical and clinical research, with particular potential for stratified analyses and less common conditions, while selection and ascertainment biases require careful consideration.

Background

A growing body of research emphasizes that modern healthcare systems face a growing burden of chronic disease linked to population aging, lifestyle changes, and environmental exposures.

Researchers have increasingly used volunteer biobanks to investigate the underlying mechanisms of chronic diseases and identify interventions at both population and individual (precision medicine) scales.

While the coupling of participants’ genomic data with their health records has reshaped research in modern medicine, reviews of existing volunteer-based biobank cohorts indicate that these resources can be affected by selection bias and underrepresentation of some population groups.

Traditional cohorts often lack the sample size required to investigate low-prevalence conditions, necessitating the establishment of unified, population-scale resources that reflect diverse demographic groups and integrate multiple clinical data streams to capture disease trajectories.

About the study

To address these empirical limitations, the researchers used data from the Our Future Health (OFH) study, a prospective UK cohort supported by government, industry, and charity sectors that began recruitment in late 2022. OFH aims to recruit five million United Kingdom (UK) resident adults, which would cover about 10% of the UK adult population.

The analysis drew on phenotypic profiles derived from questionnaire responses, physical measurements, and linked health records, with several complementary comparisons.

First, participants’ sociodemographic, lifestyle, and physical traits were compared against the nationally representative Health Survey for England (HSE) dataset. Second, participants’ self-reported disease prevalence was compared with UKB and GBD data, while associations between selected conditions and clinical correlates were benchmarked against UKB.

Finally, the researchers compared selected EHR-defined diseases with those in other large biobanks and examined medication use, cancer prevalence, and agreement between self-reports and health records. For an exploratory analysis of 187 rare, ultra-rare, and other low-prevalence conditions, they mapped ICD-10 diagnoses from linked inpatient records to corresponding Orphanet rare-disease definitions.

Study findings

Sociodemographic comparisons found that participants residing in the most deprived Index of Multiple Deprivation (IMD) quintile were underrepresented (13.0% in OFH versus 20.0% nationally), whereas those in the least deprived quintile were overrepresented (26.7% versus 20.0%), showing that recruitment did not fully mirror the national population.

OFH included more than 40,000 participants who identified as Asian British or of another Asian background and were aged 18 to 39 years, more than in any existing UK-based cohort.

Physical and lifestyle assessments showed that average body mass index (BMI) was 27.68 kilograms per square meter (kg/m²) in men and 27.04 kg/m² in women, closely matching HSE figures (27.64 kg/m² and 27.44 kg/m²).

Current smoking was substantially lower in OFH (7.38% among men, 6.56% among women) than in HSE surveys (13.22% among men, 10.55% among women). Conversely, frequent alcohol intake (three to four times weekly) was higher in OFH (22.85% of men versus 17.08% in HSE).

Across 109 conditions, disease prevalence showed strong epidemiological consistency with the UKB (r = 0.784), while clinical associations also showed broad agreement across cohorts (r = 0.783). OFH participants reported a lower prevalence of several age-associated cardiometabolic conditions, including hypertension, myocardial infarction, and stroke, while depression, anxiety, and related mental health conditions were reported more often than in the UKB. Agreement between self-report and hospital records varied from a Cohen’s kappa of 0.29 for high cholesterol to 0.66 for cancer.

OFH also contained larger rare disease cohorts than the UKB, including 668 cases of myasthenia gravis (versus 266), 464 cases of primary biliary cholangitis (versus 312), 172 cases of cystic fibrosis (versus 25), and 1,739 cases of idiopathic intracranial hypertension (versus 132). The authors cautioned that some differences may reflect age structure, survivorship, healthcare use, and recruitment patterns rather than underlying population prevalence.

Conclusions

Overall, the findings demonstrate the scale and research potential of the OFH cohort as a resource for biomedical discovery and translational research. Its large participant numbers and multiple phenotyping sources can support analyses of common, uncommon, and selected rare conditions and facilitate recruitment into clinical research.

The authors estimated that about 4.5% of invited people had consented at this stage of recruitment and cautioned that OFH should not yet be used to derive generalizable prevalence or incidence estimates across all conditions. The current data also rely heavily on self-report, and primary care EHR data were not yet included.

The future integration of expanded multi-omics data, longitudinal questionnaires, and genetic kinship mapping could expand research into disease trajectories, family-level influences, causal relationships, and understudied conditions. Continued assessment of cohort composition and additional health-record linkage will be needed as OFH develops.