Physical inactivity is an established risk factor for cardiovascular disease, cancer and diabetes, which along with chronic respiratory disease account for more than 60% of all deaths [1]. More than 80% of chronic disease deaths occur within low and middle income countries [LMC] [2]. The most recent estimates suggest that almost 2 million deaths per year worldwide are attributable to inactivit[1], leading to physical activity being described as 'the best buy in public health.' [3]. Despite global concerns about non-communicable disease in LMC [4], increasing obesity, and rapid changes in patterns of work, transport and recreation, physical activity surveillance and monitoring is only carried out in a few countries [5].

There is a significant gap in international physical activity surveillance, compared to surveillance of other chronic disease risk factors [6]. This gap makes it difficult to estimate the impact of physical inactivity on health outcomes. The World Health Organization pooled prevalence estimates to estimate the attributable risk of physical inactivity in the global burden of non-communicable diseases [1, 7]. The lack of comparable data, along with the recent Global Strategy for Diet, Physical Activity and Health [DPAS] [8], have created a compelling need for internationally-comparable measures of physical activity that can be used to quantify population levels of exposure and monitor trends over time within and among countries.

Although some countries conduct national physical activity surveillance, use of different questionnaires makes it difficult to assess inter-country differences in physical activity. In 1997, the International Physical Activity Questionnaire (IPAQ) was developed as a surveillance instrument to measure multiple domains of physical activity. This was the first effort to develop an instrument suitable for global physical activity surveillance. The goal was to identify a common questionnaire that all countries could use that would permit comparability among countries on various domains of physical activity. The IPAQ items capture moderate and vigorous intensity leisure-time, domestic, occupational and transport-related domains, which summate to total physical activity [9–11]. Assessing multiple domains of activity is particularly important in developing and transitional countries, where measures confined to leisure-time activity may miss substantial daily physical activity undertaken for the purpose of work or travel [5]. Sitting time is assessed separately by hours of sitting time per weekday and weekend day. Inactivity measures are useful for monitoring the effect of health promotion strategies that encourage people to sit less and engage more in ambulatory activity.

The aims of the International Prevalence Study on Physical Activity (IPS) were to collect and compare for the first time nationally representative prevalence estimates on physical activity from a diverse set of countries and to qualitatively assess the use of IPAQ in large-scale representative population physical activity data collection.

Development and international validation of the IPAQ has been previously reported [11]. In brief, surveillance-oriented questionnaires were developed for adults aged 18 to 65 years. Reliability and validity studies were conducted in 12 countries across 6 continents using standardised methods, and demonstrated reasonable test-retest reliability (intra-class correlations range 0.7–0.8) and inter-method validity (median rho = 0.67), with criterion validity around rho = 0.3 based on comparisons with accelerometer data. Measurement properties were similar to those of other physical activity surveys used in developed countries [12] with the IPAQ offering broader applicability to a wide range of countries and cultures. The IPAQ short form instrument was considered brief enough for physical activity surveillance, flexible enough to be used in telephone interview or self-administered applications, and adaptable enough to apply across cultures [11]. IPAQ has been used for physical activity surveillance activities in individual countries and in the European Union [13, 14].

The International Prevalence Study (IPS) was coordinated by an international group of scientists. The study involved 3 primary phases: [i] country recruitment and development of comparable study methods; [ii] within-country data collection with IPAQ nested within existing surveys or as a stand-alone survey and [iii] analysis of the pooled data using standardised protocols. A Data Management Centre (DMC) provided technical support to participating countries, collated country-specific data sets, and undertook pooled analyses. A statement of ethics approval was obtained from all centres. Informed and voluntary consent was provided verbally or in writing from all participants.

Countries were invited through existing international and regional global health and physical activity networks, key organizations (World Health Organisation; U.S. Centers for Disease Control and Prevention) and various non-communicable disease networks. Expressions of interest were received from 24 countries or large within-country regions during 2002–2004. The inclusion criteria required a representative population sample of at least 1500 adults, using comparable data collection methods in spring or autumn, and using approved cultural translations of the IPAQ instrument [see Appendix 2]. A manual of operations specified the required study protocols and indicated where modifications could be made to accommodate the local context.

IPAQ was designed to measure physical activity across all domains of leisure-time, work, transportation, and household tasks. The IPAQ short form asks respondents to report frequency and duration of walking, moderate-intensity and vigorous-intensity activity performed for at least 10 minutes duration per session. IPAQ also collects information on total sitting time, but these data are not reported here. Weekly minutes of walking, moderate-intensity and vigorous-intensity activity were calculated separately by multiplying the number of days/week by the duration on an average day. Reported minutes per week in each category were weighted by a metabolic equivalent (MET; multiples of resting energy expenditure) resulting in a physical activity estimate independent of body weight, expressed in MET-minutes/week and computed by multiplying METs by minutes/week [11]. The summary indicator was used to categorise a population into three levels of physical activity: "low" [physically inactive], "moderate" and "high" levels of physical activity. The "moderate" level nominally indicated meeting physical activity guidelines of 30 minutes of moderate intensity activity 5 days a week, 20 minutes of vigorous activity 3 days a week, or a combination [15]. Because guidelines were based mainly on leisure time activity but IPAQ assessed four domains, the "high" activity level was developed and reflects approximately twice the MET-minutes of the "moderate" level. These categories were based on standard scoring criteria http://www.ipaq.ki.se.

• Low: Meets neither 'moderate' nor 'high' criteria.

• Moderate: Meets any of the following three criteria: (a) 3 days of vigorous activity of at least 20 minutes/day; (b) 5 days of moderate-intensity activity or walking of > 30 minutes/day for > 10 minutes at a time; or (c) 5 days of any combination of walking, moderate-intensity or vigorous-intensity activities achieving at least 600 MET-minutes/week.

• High: Meets either of two criteria: (a) vigorous-intensity activity on > 3 days/week and accumulating at least 1500 MET-minutes/week; or (b) >5 days of any combination of walking, moderate-intensity, or vigorous-intensity activities achieving at least 3000 MET-minutes/week.

For country specific prevalence estimates, data were weighted to the national population based on study design. For international comparisons, all prevalence rates were age standardised to the world population distribution to enable comparability for totals across countries with varying population distributions by age [16, 17]. Data analyses were conducted using SAS software V8.02 [18].

Data were from 20 countries with a total of 52,746 individuals aged 18–65 years. For most countries there was a fairly even balance of males and females with the exception of Colombia, Japan and Lithuania (where around 60% of the samples were female), and Saudi Arabia (males 65%). Data on education level were available for 19 countries and showed a wide variation among countries (Table 1).

Figure 1 shows the relative contributions of walking, moderate-intensity, and vigorous-intensity physical activity to total MET-minutes/week for each country. In all countries at least 20% of total MET-minutes/week was accrued through walking, and walking comprised at least 50% of total MET-minutes/week in Hong Kong SAR, and China. By contrast, the contribution of vigorous-intensity activity to total MET-minutes/week varied widely among countries. For example, vigorous-intensity activity contributed less than 5% of total MET-minutes/week in India, but more than 45% of total MET-minutes/week in Australia, Brazil, Canada, New Zealand and the USA.

International comparisons of key non-communicable disease (NCD) risk factors, such as obesity and tobacco use, are commonplace [19, 20], but comparisons of physical activity prevalence among countries have not been possible due to the lack of standardised and validated instruments. Previous reviews have demonstrated very different prevalence estimates across countries, and these differences were as likely to be due to variations in questions and survey methods as they were to true differences in prevalence [21, 22]. The present study reports population-level prevalence estimates and patterns of physical activity in a diverse set of countries, using a comparable, reliable and validated survey instrument. In this study, the IPAQ short form was administered to over 52,000 adults aged 18–65 years, using a standard protocol in 20 countries.

The results show substantial variation in the population estimates of meeting the IPAQ "high active" category, a threshold developed to reflect an amount greater than than standard recommended levels, but more suited for use with a multiple domain measure such as IPAQ. Eight of the 20 countries had "high activity" rates over 50%, and these countries came from several continents. We also observed different patterns by gender and age, with most countries showing younger men more active than younger women, but this gender difference was less marked among older adults. Further, in these countries, age-related declines in physical activity were much more frequently observed among men than among women.

As shown in Figure 1, countries achieved high physical activity in different ways, with four of the most active countries (Australia, Canada, New Zealand, USA) showing a greater volume of vigorous-intensity physical activity relative to moderate-intensity activity and walking. These same countries seem to have relatively well developed facilities for recreational activity and a history of long-term promotion of exercise. Four of the countries with substantial rates of high physical activity had more than 30% of overall physical activity derived from walking (Canada, China, Colombia, Czech Republic) suggesting that countries with an infrastructure or culture that supports walking can achieve high levels of physical activity with lesser contribution from vigorous activity. However, substantial proportions from walking (Hong Kong SAR, Japan, Spain, Taiwan) and vigorous activity (Belgium, Brazil, Taiwan) were also found in countries with low overall physical activity prevalence rates (< 30% in the 'high' category), so there is no indication that an emphasis on promoting one domain of activity will lead to high levels of overall physical activity at the population level. One conclusion from these results is that different patterns of physical activity are associated with high prevalence estimates, so countries could tailor physical activity promotion strategies to local infrastructure, available programs, and culture.

There are a few European multi-country physical activity surveys that provide a context for interpreting present results. The 2002 Eurobarometer study used the IPAQ instrument [13] and identified low prevalence estimates for Belgium and Sweden, similarly to IPS and reported the highest levels of physical activity in the Netherlands and Germany (these two countries did not participate in IPS). However, earlier European-only studies had observed slightly different rankings [14, 23]. A Pan-European Union Survey on Consumer Attitudes to Physical Activity had found a similar low prevalence of activity in Belgium and Portugal, but reported a high prevalence rate in Sweden [23]. Another study used information from the 2002–3 WHO World Health Survey to collect IPAQ short-form physical activity data from 51 countries, mostly from population samples in developing countries [24]. Physical inactivity prevalence data from the four comparable countries were remarkably similar to the findings in this IPS study; for example, inactivity rates were low for China and the Czech Republic [around 10% in both the IPS and the World Health Survey], and high levels of inactivity reported in both surveys for Brazil [close to 30% inactive in both studies], and close to 25% for Spanish adults in both studies. The similarity of these estimates was noteworthy, despite different survey methods, suggesting that IPAQ data may be consistent, at least within country.

One interesting further comparison can be made from representative cross sectional data of adolescents in Europe and North America through the contemporaneous 2001–2002 Health Behaviour in School-aged children survey (HBSC) [25]. Looking at the group closest to the IPS age group (the 15 year old samples) the countries with the most active boys and girls included USA, Canada, Czech Republic and Lithuania in the upper quartile, and Belgium, Norway and Portugal in the lowest quartile. This distribution for physical activity among adolescents was similar to that observed in the IPS among adults from the same countries.

The findings from this study indicate that the majority of the population in most participating countries or regions appeared to undertake at least a moderate amount of physical activity when assessed using the multi-domain IPAQ. This suggests that most adults in these countries are obtaining some activity, yet the global problem of rising prevalence of obesity remains. Thus, it appears total physical activity in most countries remains insufficient to ensure energy balance and prevent obesity [26] or that the ratio of energy expenditure to dietary intake is unbalanced to maintain weight stability.

Strengths of IPAQ include its measurement of multiple domains, and the separate assessment of walking behaviour, compared to many current PA surveillance systems [27]. There are some limitations with IPAQ, including difficulties with respondents in distinguishing moderate and vigorous activities. It is also well recognised that self-reported measures can over-estimate physical activity [28], and the IPAQ may do this more than other surveys [29–31]. Although there is a benefit in IPAQ assessing multiple domains of physical activity as part of global surveillance, it could contribute to higher overall PA estimates than previous surveys that captured leisure-time activity alone [15]. There is also the possibility of differential measurement error using IPAQ, with some countries or population subgroups potentially giving relatively accurate estimates of their behaviour, while other populations may over-estimate or under-estimate their physical activity; this between-country variability appeared even greater in the World Health Survey, which was comprised of mostly developing countries [24]. The samples here were large-scale population samples, but eight of them were regional, not national samples, so these prevalence estimates cannot be generalised to the country level (for example, to all of China or Argentina). Based on education level (Table 1), these samples showed similar or slightly higher educational attainment to national or regional levels for those countries, suggesting comparability on this attribute.

Other methodological issues in the IPS included the variations in response rates across countries, and despite many efforts to standardize protocols, there were undoubtedly important differences in implementation. While all surveys were reviewed for consistency in the translation and cultural adaptation for participating countries, variations in how respondents understood the survey may have contributed to difficulties in interpreting findings of this international prevalence study. Though it is likely impossible to eliminate all challenges to the comparability of survey results across countries, subsequent surveillance studies could benefit from additional training of investigators, development of more rigorous translation and adaptation procedures, on-site inspections for quality control of data collection, centralized data entry system, and other methods for enhancing data quality. Furthermore, in low literacy populations, standardised showcards or illustrations should be used to depict types or intensity of different physical activities, but these cannot be used with telephone-based survey administration.

In addition to the abovementioned methodological issues, concerns remain with self report measures. Whilst they remain the most feasible and affordable instruments for global surveillance, objective population measures of physical activity, such as pedometers [32] or accelerometers [33, 34], may be beneficial to determine if differences across countries and between groups revealed in the present study represent true differences in physical activity behaviour.

The International Prevalence Study on Physical Activity provided data that allow 20 countries to be compared on physical activity behaviours for the first time. For some countries, these data represent their first large scale measurement of physical activity. Several countries have adopted the IPAQ and IPS methods as their national or regional surveillance system, and these data contribute to current WHO and European surveillance systems. A major contribution of IPS was to produce internationally comparable physical activity prevalence estimates and to demonstrate the feasibility of standardised data collection on physical activity using a common protocol. Though questions remain about the precision of the derived prevalence estimates, an important next step is the continued use of established instruments for the collection and monitoring of physical activity as part of ongoing non-communicable disease risk factor surveillance systems. Ongoing research will compare the IPAQ results with the Global Physical Activity Questionnaire (GPAQ), the World Health Organization's instrument also being used for international surveillance [21, 35], and compare both of these against objective measures.

The results from this international surveillance study on physical activity can provide useful baseline data, at least within countries, and studies could be repeated to ascertain population trends in physical activity. One important feature of a useful health monitoring system is the use of consistent measures over time [36]. Country-specific data and trends can be used to monitor efforts to promote physical activity and improve public health. Although surveillance data alone are not sufficient to motivate or guide the implementation of national policy, consistent trend data are an essential underpinning of the case for public health action.