The Florey Adelaide Male Ageing Study (FAMAS): Design, procedures & participants

The well documented needs of an ageing population have lead to an increased demand for research examining the determinants and conditions that promote healthy ageing [1, 2]. Longitudinal studies have been identified as an important source of information on the ageing process through their unique ability to track a variety of events and conditions and the interactions between them over an extended period of time [3]. A large number of cohort studies now exist, both in Australia and abroad (see [4] for review), that to a varying extent have sought to address issues associated with the ageing process.

Despite this, there still remain few comprehensive studies investigating the biomedical physical, psychological, social and behavioural elements of the ageing process in Australian men, in spite of poorer health outcomes when compared to ageing women [5]. Compared to women, Australian men have higher rates of physical and psychological disease, and death. Generally, men are also less likely than women to adopt a healthier lifestyle [6]. Despite these disparities, the health of men and changes in their health status with ageing is one of the most understudied areas of health research. This has lead to a number of peak and government bodies to call for a comprehensive men's health longitudinal study [7–9]. Accordingly, a well-characterised cohort of men representing a range of age groups, and followed for an extended period, is considered of value.

The Florey Adelaide Male Ageing Study (FAMAS) is a multi-disciplinary population cohort study examining the health and health-related behaviours of men, aged 35–80 years and living in the north-west regions of Adelaide. It employs a broad range of investigative procedures in assessing the variable interactions that contribute to the health and health related behaviours of men. The principal aims of FAMAS are to investigate: 1) Incidence of and risk factors associated with chronic physical and psychological disorders in a representative group of Australian men; 2) Endocrinology of the ageing male and its relationship with age, health status and male-specific conditions (e.g. prostate health, erectile function, lower urinary tract symptoms); 3) Determinants of the utilization of health services amongst males. The main outcome variables are incidence of chronic health conditions (e.g. diabetes, cardiovascular disease, asthma, musculoskeletal conditions, cancer, obesity); sexual health and function; prostate health and function (including lower urinary tract symptoms); muscle strength and function; psychological disorders and function, male health service utilization, general health and well-being; physical activity levels; medication usage. In addition, the project involves a number of cross-sectional sub-studies investigating specific age related conditions.

This paper details the study methodology of FAMAS, including the design, sample techniques and clinic and research protocols used. To date there is little data on the factors associated with participation in the few longitudinal studies that specifically investigate the health of men in Australia. Research from other ageing cohort studies suggest that time constraints, low socio-economic status, smoking status and a lack of perceived benefit are all significant barriers to participation in men [10–12]. This is an area requiring further investigation in the Australian male population. Potential sources of selection and observational bias will be also examined through analysis of sampling and attrition. Finally, the cohort will be compared with available national data on key demographics for comparability with the target populations.

The population sampled for the Florey Adelaide Male Ageing Study (FAMAS) has yielded participants broadly representative of the sampling region and the broader Australian population of Australian men aged 35–80 in key demographics. Efforts will now focus on maintaining participation in a unique and expanding investigative base.

All longitudinal studies must include an assessment of their baseline response rates [33]. Studies with low levels of initial recruitment can potentially introduce significant levels of selection bias into the sample. Response rates can be difficult to interpret as the methods used to calculate them are often not specified. Some studies report response rates as a percentage of eligible people in the sample, whilst others report it as a percentage of the entire sample. The former often leads to much higher rates. Our chosen method concurs with that most often used in research of this type [33]. By way of comparison, our observed response rates (sample: 67.8%; clinic: 45.1%) appear much higher than those of another study (clinic response rate: 29%) using very similar sampling procedures [34]. Certainly given the study type and investigations (including detailed sexual health measures), the observed response rates seem sufficient.

Several other studies have employed additional methods to bolster their recruitment rates (see [32] for review). These have included the use of "volunteer enrichment" (e.g. Framingham Heart Study, [35]), whereby investigators allowed unsolicited participants to enter the study. Whilst many approaches from suitable candidates were received during promotional activities for the study, such requests were declined in favour of maintaining a randomly-selected sample. We have, however, managed to achieve reasonable levels of initial recruitment from our target population by minimizing exclusion criteria and incorporating measures in our study design that would appeal to our target population. Indeed, most studies [36, 37] report that participation seems largely related to the perceived benefits of participation. Thus, measures such as a comprehensive health review and relatively low contact time appear to have helped increase involvement in a population traditionally hesitant to participate [38, 39]. Despite this, the most commonly cited reasons for study refusal were householders being either 'too busy' or 'not interested', universal responses in any study of this type [40]. Interestingly, people citing being 'too old', 'too sick' or for having 'no health problems' were only minor contributors to study refusal, in contrast to previous findings [41, 42]. Whilst the reasons given for refusing participation in health research have been well documented elsewhere, there is a scarcity of such data from Australian male cohort studies. Such findings will have implications for future studies and preventative programmes involving ageing men.

Comparisons of non-responders and study participants in any study is useful in highlighting key points of difference between the two groups and allowing for an estimation of any sources of selection bias. The under representation of single men in the cohort provides further agreement with a volume of research [43] that suggests that people (especially the elderly) with live-in support systems generally demonstrate higher levels of attendance in health studies (and, more generally, better health outcomes) than people who live alone. There is also strong evidence to suggest that non-responders tend to display poorer health outcomes (e.g. an increased incidence of strokes [44] and diabetes [45]). However, in the present study non-participators were less likely to have had previous and current episodes of hypercholesterolemia. During pre-recruitment promotion, strong emphasis was placed on the free cholesterol testing and, combined with a recent national emphasis on elevated cholesterol levels, this may have dissuaded men who had previously or were currently receiving treatment. In all, our analyses do suggest that our study participants are broadly representative of the sampled population and that those that were unwilling to participate did not create a substantial bias. However, the observed discrepancies will have to be taken into account when applying study results to the overall population.

In any longitudinal design, particular effort is needed in characterizing loss to follow up (LTFU) in order to help identify potential sources of future cohort loss [46]. Previous research has identified a host of demographic factors that are associated with LTFU (see [31] for review). In most population studies, those greater than 60 years are less likely to withdraw [47], although this is likely offset by an elevated death rate. Indeed, in the present study, almost half of all study losses were as a result of participant death, the large majority of which came from the initial 65–80 age bracket. Aside from age, no other demographic factor appears to have influenced LTFU. The current LTFU rates are extremely low in comparison to other studies of this type [4]. Whilst it is acknowledged that our second wave of clinic visits is yet to begin, we have none the less been able to maintain a good rate of contact with most of the cohort to date through a host of activities. Although efforts to maximize participation appear to have been effective, other successful measures adopted by similar studies (especially in the critical elder and younger brackets [47–49]) will also be utilized to maintain the stability of the cohort.

Whilst the ability to generalize results from a given study to the broader population is one of the key elements of longitudinal research it is quite often not the primary focus of the study. Certainly many of the more noted cohort studies (e.g. British Doctors Survey [50], CARDIA [51], Framingham [52]) were non-representative by design. In the present study however a broadly representative cohort was sought. As argued by Szklo [33], an important determinant in realizing a representative cohort is the selection of a complete sampling frame. In this context, our use of the Electronic White Pages may be seen as a limitation given that this restricts potential respondents to those with a connected line. However, research [14, 53] using a comparable sampling technique, has suggested that this is not the case.

In terms of the age profile of our cohort, the lower proportion of young males in the cohort is notable. This is a common phenomenon in studies with similar sampling designs [31, 53] and a number of initiatives to maximize the ongoing involvement of this group of men have been adopted in consequence (e.g. family days, targeting specific media outlets). The overrepresentation of the elderly has also been observed previously, and whilst the reasons for this are situation-dependant, it is generally considered that such participants can offer more time to community-based research [54]. Whilst the profile of FAMAS participants did significantly differ from the broader community in selected age groups, the careful application of appropriate weights to future results should offset any age-influenced bias. Additionally, the present age profile will help to ensure adequate inter-age group comparisons can be performed, an important consideration in any ageing longitudinal study. Indeed, we would argue that our cohort is well placed long-term to cope with the attrition normally associated with studies of this type.

The low numbers in some regions of birth means that a statistical comparison was not warranted. Most notably, we do appear slightly overrepresented with participants from the UK/Ireland. Whilst no clear reasons exist for this discrepancy, it is interesting to note that some of the highest response rates for population studies are recorded in British surveys [55]. The apparent lower proportions in other regions (whilst not statistical) are slightly more difficult to explain, and may reflect a general trend towards non-participation in this country [56]. Similar patterns have been observed in a nationwide telephone-based survey of the reproductive health of approximately 6000 men age 40–80 (Men in Australia: Telephone Survey; MATeS) [57, 58].

The income distribution for the cohort also appears broadly reflective of the wider populations. The disparity between the proportion of high household incomes in the north west Adelaide and Australian populations is indicative of some of the socio-economic disadvantages of the region [59].

Many population studies report that a respondents level of education strongly predicts their likelihood of involvement [60, 61]. This notion received support in the present study with a higher proportion of participants with some post-school qualification as compared to the general population. These figures bode well for the stability of the cohort, with previous research demonstrating that those with a higher educational standard are less likely to withdraw [62, 63]. Taken together, these figures suggest that the men in the present study are generally reflective of the community from which they were randomly selected and the Australian population as a whole.

In terms of the overall study frame, the occurrence of clinics at every five years should be adequate for the conditions under investigations. Whilst other cohort studies have benefited from more frequent clinical investigations [62, 63] these studies were investigating inherently dynamic disease conditions or outcomes. In addition to the logistical and financial expense associated with clinics, participant contact beyond that which is necessary is likely to lower recruitment and increase LTFU. Also, in the present study there is a sustained effort to keep regular contact with participants in between clinic waves through annual follow-up questionnaires, sub-study investigations and cohort maintenance initiatives. In comparison to other longitudinal studies of ageing the incorporation of a "younger" subset is quite unique [4] and, provided these numbers can be maintained, will provide an invaluable opportunity to study the progression of various conditions and changing behaviours in men.

An a priori target of around 1200 participants was selected before recruitment ensuring sufficient power for the main outcome variables. Even relatively minor correlations between outcomes could be reliably detected. Whilst this cohort size may be smaller in comparison to other longitudinal studies of ageing we would argue that the focus on a single gender and the high stability of the cohort observed to date (in part due to the conscious efforts to minimize losses to follow-up) will ensure a relatively large number of active, participating subjects for the duration of the study.

The north west region of Adelaide appears an ideal community in which to investigate the ageing process, with a population that is largely reflective of the state with the highest proportion of elderly in Australia. The urban-specific focus of the study may limit the broader applicability of study results, however in the present circumstance incorporating regional participants was not feasible. Also, at present there are no home visits planned for participants with limited mobility given the logistics and expenses involved. There is still considerable debate on whether such measures can reduce the high attrition associated with elderly participants in cohort studies [64].

The Florey Adelaide Male Ageing Study will provide an opportunity to add to existing ageing longitudinal research, both internationally and locally, with a well-characterized, broadly representative cohort that remains largely active. The study design reflects a growing trend of integrative research on ageing with inputs from a range of disciplines (endocrinology, epidemiology, gerontology, public health, psychology, sociology, ophthalmology, bone and joint disease, nutrition, genomic health, politics). The inclusion of middle aged participants is unique relative to most ageing longitudinal studies, and will allow the detection of earlier life trajectories and morbidity factors associated with healthy ageing in later life. Of the few longitudinal studies specific to conditions affecting men in Australia, most are characterised as specialized investigations of the late elderly, limiting the scope of problems addressed and the potential as a longitudinal study. By contrast, FAMAS offers a broad-based approach to many of the conditions of the ageing male, increasing acknowledged as areas that require significant research input.