Response bias deepened as cardiovascular risk rose in primary care trial invitations
A cross-sectional analysis of recruitment to a randomized lifestyle intervention trial found a familiar but clinically important pattern: the people most likely to benefit were least likely to respond. Among 8,902 patients invited from 60 general practices, only 1,489 replied to the trial invitation, leaving a large gap between the at-risk population identified in routine care and the group that actually entered the study.
The trial targeted adults with a QRisk2 score of at least 20%, a threshold that identifies substantial baseline cardiovascular risk in primary care. Response fell as risk increased, with an adjusted odds ratio of 0.82 for every 5-point rise in QRisk2 score. In practical terms, the trial invitation became less persuasive precisely as the burden of cardiovascular risk climbed. For prevention research, that is not a trivial sampling artifact; it changes who is seen, who is counted, and whose behavior is used to judge the effectiveness of the intervention.
Higher risk did not translate into higher engagement
The recruitment pattern suggests a selection process built into the first step of a prevention study. A lifestyle intervention can only show its real-world value if people accept the invitation and enter the trial, yet those with the highest calculated risk were progressively less likely to respond. This creates a trial cohort that may underrepresent patients with the greatest need for prevention, especially when the intervention depends on motivation, time, trust, and practical capacity to attend or participate.
The mean age of invited patients was 67.3 years, and 21.0% were female. Of those invited, 69.9% were recorded as white, 13.9% as belonging to ethnic minority backgrounds, and 16.2% had no ethnicity recorded in their medical records. Missing data here are not a clerical footnote. Poor ethnicity recording weakens the ability to measure whether research recruitment is equitable and can hide barriers that operate before a patient ever reaches the consent stage.
Ethnicity and deprivation shaped response in different ways
Ethnicity patterns were uneven. Patients identified as Black African or Caribbean were less likely to respond than white patients, with an adjusted odds ratio of 0.67. Patients with missing ethnicity data were also less likely to respond, with an adjusted odds ratio of 0.55. By contrast, response rates did not differ between South Asian and white patients, with an adjusted odds ratio of 1.08. Those contrasts matter because they show that under-representation is not uniform across minority groups; recruitment barriers may differ by community, previous experience with health services, trust, language, socioeconomic position, or the way risk communication is framed in primary care.
Deprivation showed a clear gradient. Patients in the fourth and fifth most deprived quintiles were less likely to respond than those in the least deprived quintile, with adjusted odds ratios of 0.70 and 0.52, respectively. In a prevention trial, deprivation can influence response through several mechanisms at once: competing priorities, lower health literacy, reduced flexibility to attend appointments, digital access barriers, and skepticism about whether a research study will lead to immediate personal benefit. Those forces can suppress participation even when clinical need is highest.
One result stood out sharply: the most deprived patients were the least likely to answer the invitation at all.
Recruitment bias can reshape the apparent effectiveness of prevention
This analysis makes clear how easily a lifestyle trial can drift away from the population it was designed to serve. If the responders are healthier, less deprived, or more engaged than non-responders, the resulting sample will not mirror routine primary care. Intervention effects can then look cleaner than they would in ordinary practice, while uptake, adherence, and implementation challenges remain hidden behind a selectively responsive cohort.
For cardiovascular prevention, this matters because risk reduction is often delivered through behavior change, repeated contact, and sustained self-management. These approaches are inherently vulnerable to response bias at recruitment, long before adherence or clinical endpoints are measured. A trial that enrolls those already more inclined to engage may overestimate feasibility and underestimate the support needed to reach patients with the highest baseline risk.
The findings also point toward a practical research problem for primary care. Invitations based on electronic risk stratification are efficient, but they are only as representative as the response they generate. When response falls with increasing QRisk2 score, the recruitment strategy is not simply inefficient; it is systematically filtering out the very group the intervention was designed to help. That has implications for trial design, service evaluation, and implementation research in cardiovascular prevention.
Ethnicity recording and outreach methods remain central to trial validity
The analysis also exposes a data-quality issue with direct consequences for research interpretation. Better recording of ethnicity in primary care records is necessary if trialists are to understand who is being reached and who is not. Without reliable ethnicity data, it becomes difficult to distinguish between true differences in engagement and missing information that conceals them.
Future trials in high-risk cardiovascular prevention may need more than standard mailed invitations or routine contact methods. Recruitment strategies may need to be adapted for groups that are harder to reach, including people living in more deprived areas and some ethnic minority communities. That could mean more personalized outreach, more accessible study information, or recruitment methods aligned with the communication preferences of the target population. The study does not test which approach works best, but it shows why the question is unavoidable.
Trial participation and disease burden are already out of sync
The registration details identify the study as ISRCTN84864870, registered on 15 May 2012. Its broader lesson is straightforward and uncomfortable: in prevention research, the burden of disease and the likelihood of response do not always move in the same direction. When that happens, the sample that answers the door may not be the sample that most needs help.
The result is not just a statistical concern. It is a direct threat to the external validity of lifestyle intervention research in cardiovascular prevention, especially in populations identified through primary care risk scores. If response continues to fall as risk rises, investigators will keep measuring success in a group that is easier to recruit rather than more representative of the patients whose outcomes matter most.