Potential for enhancements

2.5.1 Repeat assessments in representative subsets

Typically in prospective studies of the relevance to disease of risk factors (such as blood pressure or blood lipids), various characteristics of the cohort are recorded at the initial "baseline" assessment visit and these baseline characteristics of individuals who subsequently develop a particular disease are then compared with those of individuals who do not. But, because of fluctuations in the measured values of a risk factor at baseline, such comparisons often substantially underestimate the strength of the real association between the

"usual" (i.e. long-term average) level of that risk factor during a particular exposure period and the disease rate during that same, or a later, period [114].

This "regression dilution" effect may be caused by measurement error, by short-term biological variability (including both transient fluctuations and any diurnal or seasonal variation), or by longer-term within-person fluctuations in risk factor values (which may occur for several reasons, including physical activity, diet, treatment, disease or age).

Information from repeat measurements of the risk factor after just a year or two in a reasonably representative sample of individuals can be used to correct for the effects not only of random measurement error but also of short-term variability in risk factor levels. If, however, the aim is to estimate the usual risk factor levels 10 or 20 years later then corrections based on re-measurements made relatively soon after baseline may not allow properly for the effects of longer-term within-person variability. Moreover, since the interval between the baseline survey and the occurrence of an event in prospective studies is typically longer among those who suffer events at older ages, such underestimation may well be greater in the elderly. In order to make appropriate "time-dependent" corrections for these effects of regression dilution, re-measurements during prolonged follow-up can be used to estimate the usual risk factor levels at some particular fixed interval prior to death in each decade of age [49, 52]. In order to be able to adjust sufficiently reliably within various subsets of the cohort (e.g. for different ages at risk), such re-assessments need to involve at least a few tens of thousands of individuals on each occasion. Consequently, in UK Biobank, it is planned to repeat the baseline assessment (i.e. questionnaire, measurements and sample collection) in about 25,000 participants during the recruitment phase and then every 2-3 years during follow-up in a similar sized cohort.

2.5.2 Additional measures at re-assessment

Typically, in order to allow correction for regression dilution, the measures of interest made at baseline are repeated during the periodic re-assessments in representative samples of the cohort (Section 2.5.1). But, such repeat assessments can also provide an opportunity to conduct more intensive phenotyping of the participants being re-assessed. Whereas it might not be feasible (e.g. for reasons of cost) to undertake such intensive phenotyping in the whole cohort, more detailed assessment in several thousand individuals could still help to inform the whole cohort [115, 116]. For example, if for some reason it

was only feasible to estimate blood pressure as “below average”, “average” or

“above average” (rather than to measure it directly) in all participants at baseline, then the informativeness of this estimate of blood pressure as a predictor of disease would be limited. But, if it was then possible to measure blood pressure in a representative subset of the cohort (e.g. during a subsequent re-assessment), these measured values could be used to determine the measured long-term usual blood pressure for each of the baseline-defined groups (i.e.

below average/average/above average). That is, more precise measurement of some particular factor in a reasonably representative subset of the cohort would allow adjustment not only for regression dilution but also “calibration” for other sources of imprecision in baseline measures conducted in the cohort as a whole.

This calibration approach is likely to be particularly useful for various measures that it has not been possible to include in the baseline assessment of all participants in UK Biobank. For example, as described in Section 1.3.3.6, it is intended to develop an internet-based dietary recall questionnaire that could be completed by a substantial proportion of the cohort and so supplement the more limited food frequency information being sought in the whole cohort. Similarly, the repeat assessment visits planned for about 25,000 participants every few years (Section 2.5.1) provide an opportunity to conduct some more intensive measurements (e.g. the questionnaire-based estimates of physical activity being obtained at baseline could be supplemented by some more objective validated measure of energy expenditure, such as heart rate monitoring [117]).

Development and conduct of the internet-based dietary recall questionnaire has been included in the budget for UK Biobank, and so too have the costs of repeating the standard baseline assessment visit every few years in about 25,000 participants. Separate funding will need to be sought, however, for the additional costs of conducting some more intensive measure in a subset of the participants attending for re-assessment. Given the potential value of such add on studies (and their relatively modest marginal costs), it seems likely that researchers interested in enhancing the UK Biobank resource in this way would be able to raise this funding through the regular peer-review mechanisms.

2.5.3 Intensive phenotyping at baseline

As discussed in Section 1.4.3, a large number of physical measures potentially associated with various health outcomes were excluded from the baseline assessment of the whole cohort for reasons of feasibility (i.e. available funding would not allow a more prolonged visit). These included electrocardiogram;

continuous or ambulatory blood pressure and pulse rate; ankle-brachial index;

pulse wave velocity; carotid intimal-medial thickness; cardiac echocardiogram;

skinfold thickness; spirometry reversibility; bone densitometry; quadriceps strength; timed shuttle walk test; aggregated locomotor test; and visual and auditory acuity. Section 1.2 provides the rationale for recruiting at least 500,000 individuals aged 40-69 and following them for several years in order that there will be sufficient numbers of cases of any particular disease to allow the reliable assessment of plausible risk associations. Indeed, even with the more common conditions (such as coronary heart disease or diabetes), it is likely to require at least 5 years of follow-up before 5,000 cases have developed. But, as follow-up

baseline measurements made in only a substantial subset of the whole population might well become informative. This would be the case especially if such measures were more precise and strongly related to health outcomes than those made in the whole cohort (e.g. heart rate monitoring rather than a questionnaire for physical activity) [118, 119].

As discussed above, variability and other sources of imprecision in the baseline assessment can be allowed for in UK Biobank by conducting repeat assessments that include some more precise measures in several thousand reasonably representative participants. As a complementary strategy, it has been proposed that some additional measures be conducted at baseline in about 100-200,000 of the participants. This option for an intensively phenotyped sub-cohort within UK Biobank has not been included in the budget and additional funding will need to be obtained to cover the full costs of its inclusion (including the impact on the assessment centre throughput and any changes to IT or other systems). Nor have there been detailed discussions as to what (if any) additional measures might be conducted in such an intensively phenotyped cohort. Instead, what is planned is that there be wide consultation during the early phase of recruitment among interested researchers in the UK (and elsewhere) as to what additional measures might be included. Funding will then be sought from relevant sources (e.g. heart disease charities for vascular outcomes; cancer charities for neoplastic outcomes) by those researchers, in collaboration with UK Biobank, with a view to incorporating these additional measures into the assessment visits during the latter phase of enrolment (e.g. the last 100-200,000 recruited).

2.6 Long-term follow-up