The Hertfordshire Cohort Study: an overview

The Hertfordshire birth records

In 1911, a team of midwives and nurses was assembled in Hertfordshire with the aim of improving the health of children in the county. Midwives attended women during childbirth and health visitors routinely visited the child throughout infancy and childhood; weight at birth and one year of age, method of infant feeding, illnesses, and developmental milestones were all recorded. The information (Table 1) was transcribed into ledgers (Figure 1) at the County Health Department in Hertford. The ledgers cover almost all births in Hertfordshire from 1911 until the NHS was formed in 1948.

Figure 1. An extract from the Hertfordshire ledgers.

Source: original Hertfordshire ledger, on loan from the Hertfordshire Local Studies Archive; currently stored securely at the MRC LEU.

Studies based on men and women born in Hertfordshire 1911–30

The MRC EEU computerised the Hertfordshire ledgers and traced the cohort by linkage with the National Health Service Central Register (NHSCR). A mortality study of 15,000 men and women born between 1911–30 showed that lower weight at birth and one year of age was associated with increased risk of death from cardiovascular disease by 75 years of age². Subsequently, numerous detailed physiological studies were conducted among community dwelling men and women born in Hertfordshire between 1920–30 and still living there in the early 1990s. Small size in early life was associated with greater risk of cardiometabolic disease^3–5, osteoporosis⁶ and sarcopenia (the loss of muscle mass, strength and function with age)⁷ in later life. Collectively these studies contributed to the “developmental origins” hypothesis which suggests that the nourishment a baby receives in utero, and its development in infancy and early childhood, determine its risk of disease in later life⁸.

These early studies were important for establishing relationships between the early environment and physiological markers of disease, but they lacked detailed information on adult anthropometry and diet, and the men and women born 1920–30 were becoming frail and unable to take part in further studies. A second (younger) cohort, in whom the work could be continued, was therefore recruited between 1998–2004. During this time Professor Cyrus Cooper became Hertfordshire Cohort Study (HCS) principal investigator and the MRC EEU was reconfigured as the MRC Lifecourse Epidemiology Unit (LEU), a University Unit Partnership between the MRC and the University of Southampton.

Objective of the Hertfordshire Cohort Study

The principal objective of the HCS is to evaluate interactions between the intra-uterine and early postnatal environment, adult diet and lifestyle, and genetics, in the aetiology of chronic disorders of ageing (osteoporosis, osteoarthritis, sarcopenia, obesity, cardiometabolic disease and type II diabetes mellitus). The study aims to place these interactions within a life-course model for disease pathogenesis, and to characterise the physiological mechanisms underlying the pathways to these chronic disorders.

Baseline recruitment 1998–2004

The baseline recruitment of HCS participants has been described in detail previously⁹; key stages in the process are shown in Figure 2. In brief, the ledgers contained records for 42974 births in Hertfordshire between 1931 and 1939; the National Health Service Central Register (NHSCR) in Southport (now part of NHS Digital) traced 8650 men and women who were still alive in Hertfordshire in 1998. Permission to contact 6099 men and women by letter was obtained from their GPs and 3225 (53%) agreed to a home interview with a trained research nurse. 2997 (93%) men and women subsequently attended a clinic for detailed physiological investigations. The baseline HCS fieldwork was county-wide but conducted in phases by geographical area and gender. 966 (68%) of the 1412 men and women who attended clinic in East Hertfordshire also underwent a Dual-energy X-ray Absorptiometry (DXA) bone scan and knee radiography. An overview of the data collected at the HCS baseline home interviews and clinics is provided in Table 2.

Figure 2. Participant recruitment and summary of studies carried out from 1998–2018.

Ethical permission and informed consent

The HCS baseline investigations had ethical approval from the Hertfordshire and Bedfordshire Local Research Ethics Committee and all subjects gave written informed consent. All interviews and physiological investigations were carried out according to strict protocols and studies of within- and between-observer variation were conducted at regular intervals during fieldwork to ensure comparability of measurements obtained over time. Data were collected, anonymised, processed and stored in accordance with contemporary data protection regulations. Participants are free to withdraw from the study at any time.

Follow-up studies 2004 to 2018

Since their first contact in 1998–2004, HCS cohort members have generously agreed to take part in various follow-up postal questionnaires, face-to-face interviews, clinics, intervention studies, and focus groups (see Figure 2). With their consent, data on hospital admissions experienced by the cohort members across a decade have been obtained from NHS Digital¹⁰, and cohort members remain flagged on the National Health Service Central Register for ongoing notification of mortality. Ethical approval was obtained for all elements of the study at the time they were conducted.

Table 3 to Table 11 describe the scientific objective and methodology of each HCS follow-up study and detail key data availability. Taken together, the HCS datasets provide a rich, detailed and longitudinal characterisation of the sociodemographic, lifestyle, medical and biological attributes of a contemporary group of community-dwelling older men and women in the UK.

Osteoporosis

Osteoporosis is the most common metabolic bone disease affecting older people; almost one in two women and one in five men aged 50 years will have an osteoporotic fracture in their remaining lifetime¹¹. The LEU is a renowned contributor to research on osteoporosis, in part because of the wealth of data collected in HCS (Table 2–Table 12). Estimates of bone mineral density (BMD) and bone mineral content (BMC) have been obtained by DXA; estimates of bone strength (including strength strain index (SSI)), cortical bone and trabecular bone by peripheral Quantitative Computed Tomography (pQCT); and ‘virtual bone biopsies’ by HR pQCT. Few other cohorts have characterised bone in such detail.

Predictors of osteoporosis. The Health Visitors’ records that are uniquely available in HCS have enabled investigation of the early-life origins of osteoporosis. Studies using DXA showed an increase in adult BMC with rising weight at birth, and more strongly, with weight at one year. Models that also included weight in adulthood showed impacts accruing from each time-point, with greater contributions of early growth to BMC than to BMD¹². In addition, growth during the first year was shown to alter the geometry of the adult hip¹³. Studies using pQCT showed that bone strength was similarly related to early weights¹⁴.

Risk of osteoporosis attributable to lifestyle has also been investigated in this cohort. While no social patterning has been observed¹⁵, nor any relationship found between cigarette smoking, alcohol use, or physical activity and either BMC or BMD¹⁶, alcohol has been associated with less radial cortical and trabecular bone in men and less tibial trabecular bone in women¹⁷. Clustering of risk factors (physical activity, diet, alcohol consumption, smoking behaviour, grip strength, personal and family interactions, comorbidities) has been found to increase the risk of low BMD in women¹⁸. Also in women, pQCT scans have shown that having a ‘healthier’ diet (characterised by greater consumption of fruit, vegetables and wholegrain cereals) is associated with increased bone area on pQCT scanning¹⁹. Some lifestyle risk factors appear to interact with early life factors to influence osteoporosis risk, for example, in men, having a low birth weight and being a current smoker increased the risk of low femoral BMD¹⁶.

Osteoporosis and comorbidity. HCS has provided valuable insight into the interplay between osteoporosis and other chronic conditions. For example, in the case of cardiometabolic disease, analyses have shown that: BMD is associated with serum triglycerides and HDL²²; higher BMI is associated with lower 25(OH) vitamin D²³; arterial calcification affects the structure of cortical and trabecular bone in women, but not in men²⁴; hyperinsulinemia influences BMD through BMI²⁵; and the associations between lean and fat mass and bone microarchitecture differ²⁶. In the case of chronic inflammation in later life (inflammaging), HCS has shown that an elevated inflammatory state is associated with reduced BMD at baseline and with an accelerated rate of decline over time²⁷. Relationships between muscle and bone health are discussed in the section on sarcopenia that follows.

Outcomes of osteoporosis. The primary outcome of osteoporosis is fracture, which carries high personal and societal costs. Risk of fracture has been elucidated in HCS, showing, for example, that history of falls adds to clinical risk factors and low BMD as a risk factor for fracture²⁸. Research using pQCT scans has shown that in women, cortical radial thickness and area, and tibial cortical area and density, are associated with adult fracture risk, and in men tibial SSI is associated with adult fracture risk²⁹. Finally, cluster analysis of bone microarchitecture from HR pQCT has demonstrated two separate phenotypes associated with high fracture risk³¹.

Osteoarthritis (OA)

OA is the most common joint disease affecting older people³². It is an important element of the research programme of the LEU and its incorporated Arthritis Research UK MRC Centre for Musculoskeletal Health and Work. In HCS, radiographic, clinical and self-reported markers of OA have been collected at multiple time points (Table 2–Table 12).

Descriptive epidemiology. HCS data have been used to show that the method of ascertainment of OA status affects both its calculated prevalence³³ and the demographic characteristics with which it is associated³⁴. For example, prevalence of knee OA ranged from 18% for clinically diagnosed disease through 21% for self-reported disease to 42% when the diagnostic criteria were radiographic; with higher specificity and lower sensitivity for clinical and self-reported OA than for radiographically determined disease³³. Further work utilising data for multiple methods of ascertainment showed that in the presence of radiographically determined disease in the knee, signs and symptoms differed dependent on the bones involved, tibiofemoral OA being associated with both signs and symptoms, and patellofemoral OA only with clinical signs³⁵.

Predictors of OA. Results from this study have demonstrated that growth before birth and in the first year both affect the likelihood of developing OA in later life³⁶. Specifically, an increase in number of osteophytes on hip radiographs has been associated with low birthweight and the number on knee radiographs with low weight at 1 year. An increased prevalence of clinical OA at the hand was also seen in those whose weight at 1 was low. Conversely, rheumatoid arthritis was not associated with early weight³⁷. The role of biochemical markers of bone in the aetiology of OA has also been investigated, finding that CTX-II and glucosyl-galactosyl-pyridinoline were associated with both osteophyte score and joint-space narrowing, whist CTX1 and osteocalcin were not³⁸.

Outcomes of OA. The consequences of OA include an increase in the impact of objectively measured neighbourhood problems on participants’ physical activity³⁹ and quality of life⁴⁰. Pain is an important symptom of OA which, in HCS and other EPOSA cohorts, has been found to fully explain the association of OA with objectively measured (but not with self-reported) physical function⁴¹. This suggests that individuals consider factors other than pain (e.g. comorbidity) in making an assessment of their physical function.

Sarcopenia

Table 2–Table 12 show that HCS has collected a wealth of markers of the mass, strength, function and morphology of muscle. Together, these have enabled a rich body of research on the predictors and consequences of sarcopenia.

Descriptive epidemiology. Sarcopenia may be defined by one or more of the markers collected in HCS, inter-relationships between which have been explored in detail^42,43. The prevalence of sarcopenia across HCS, estimated according to European Working Group for Sarcopenia in Older People (EWGSOP) guidelines, but with lean mass estimated by skinfold measurements, was 4.6% in men and 7.9% in women⁴⁴. For subgroups of HCS in which imaging and anthropometric data coexist, estimates of prevalence have been calculated according to more than one operational definition of sarcopenia (EWGSOP; International Working Group on Sarcopenia; Foundation for the National Institutes of Health sarcopenia project; dysmobility syndrome), enabling the methods to be compared directly^44,45. Late-life declines in muscle mass as well as strength have been observed using data from two of the time points shown on Figure 2 (MSFU and EPOSA); these data suggest that muscle mass declines less rapidly than strength or function⁴⁶.

Predictors of sarcopenia. The early-life origins of sarcopenia have been investigated using data from the Health Visitors’ records. Weights at birth and one year of age have been related to adult anthropometry, showing that whilst BMI and lean mass rise with birthweight and with weight at one, fat mass has associations only with the latter⁴⁷. This suggests that whilst the prenatal environment affects lean mass, the postnatal environment is more influential in the development of adult adiposity. Direct measures of muscle size from pQCT scanning have confirmed that birthweight is associated with muscle area in the calf and forearm⁴⁸, whilst biopsies showed that smaller babies have lower total myofibre scores (kg/mm²) than those who were bigger at birth⁴⁹. Grip strength was more strongly associated with birth weight than with growth during the first year of life in HCS participants⁵⁰. This again suggests that sarcopenia has primarily pre- rather than post-natal origins, though there is some evidence to link breast feeding during infancy with increased grip strength⁵¹. A systematic review and meta-analysis of work in HCS and other cohorts showed that the positive relationship between birth weight and muscle strength exists across the lifecourse⁵².

The effects of adult lifestyle on risk of sarcopenia have also been investigated. First, to examine the role of diet, associations of grip strength with specific nutrients, individual foods and dietary patterns were studied. Healthier diets were associated with higher grip strength, although the strongest association was with more frequent consumption of oily fish⁵³. Physical activity in old age, objectively measured by accelerometry, was found to be associated with decreased risk of EWGSOP sarcopenia and improved physical function, though not with grip strength⁵⁴. Grip strength was not associated with lifetime occupational exposure to physically demanding activities⁵⁵, a somewhat surprising finding in the light of evidence of social patterning in sarcopenia⁵⁶. The role of adult heath in causal paths to sarcopenia has also been investigated, identifying multimorbidity⁵⁷, inflammaging⁵⁸, and some⁵⁹, but not all⁶⁰, cardiovascular drugs as important influences.

Outcomes of sarcopenia. Sarcopenia has profound consequences for health and wellbeing in later life and is associated in HCS with increased risk of hospital admission⁶¹. Changes in ageing muscle have been shown to contribute to the metabolic syndrome and all its components⁶²; to self-rated health and health-related quality of life⁶³; and to bone mineral content⁶⁴. Together with strong independent relationships between muscle and bone size/strength⁶⁵, this latter finding supports the existence of a muscle/bone unit and provides some insight into the mechanisms that underlie it.

Diet

Determinants of diet quality. The HCS has described marked differences in dietary choices in later life. For example, healthier diets have been found to be associated with being female, reporting a non-manual social class, and with being a non-smoker⁶⁶. Psychological and social factors have important influences⁶⁷, greater involvement in social activities being associated with slower declines in diet quality⁶⁸. In addition, a relationship has been described between infant feeding and health behaviours in later life, such that people who were breastfed as infants were more likely to have healthier diets⁶⁹.

Outcomes of diet quality. Our studies have described links between differences in nutrient intake and physical functioning in older women⁷⁰, and shown that there is a graded increase in the prevalence of poor physical functioning in older adults as the number of risk factors (from obesity, smoking and poor diet) increases⁷¹. Less healthy diets also contribute to risk of hospital admission⁷². Additionally, higher processed meat consumption and antioxidant intake have been related to poorer lung function in HCS⁷³, while low dietary antioxidants were also associated with poorer glucose tolerance⁷⁴.

Methodology. Assessment of diet in older populations can be burdensome. In HCS, a short food frequency questionnaire was therefore developed and validated to assess compliance with a healthy dietary pattern in older community-dwelling adults⁷⁵.

Mortality

In total, the 1911 to 1939 Hertfordshire birth cohort comprises 37 000 men and women, 7916 of whom had died by the end of 1999. Risk of death from circulatory disease was lower among men and women who were heavier at birth. Women who were heavier at birth also had lower risk of death from pneumonia, injury, diabetes, and musculoskeletal disease⁷⁸.

Hospital care

HCS was one of the first English cohorts to link with HES data and produced early evidence on the descriptive epidemiology of hospital admissions among older individuals¹⁰. Such individual level data contrast with published statistics, which are at the population level. A novel methodology was developed to explore risk factors for admission in these multiple-failure survival data⁷⁹, demonstrating links between admission and grip strength⁶¹, and showing that poor lifestyle risk factors have a cumulative effect on likelihood of admission⁷².