The study population was recruited from two municipalities in Nord-Trøndelag County, Norway. The participants, aged 13–18 years old, were identified and randomly selected from four different schools in the included municipalities. An invitation was distributed by the teachers at the schools. Each participant received an information folder and signed a written consent. The parents of participants below 16 years also signed the consent.

Participation in the study was voluntary. A total of 200 adolescents were invited, 71 participated. With an estimated effect size of 0.5 and power of 80% (two-tailed alpha = .05), between the scores from the physical activity questionnaires and the objective measures, the study needed a sample of 58 participants

Seventy-one participants, 56.3% girls, completed the questionnaire and anthropometric measurements. Mean age was 14.9 years (girls 15.3 years, boys 14.4 years) (Table 2). Sixty-seven participants completed the VO_2peak measures (30 boys and 37 girls), while 62 (26 boys and 36 girls) completed all parts of the study.

Boys had a significantly higher VO_2peak compared to girls (Table 3), but there was no significant difference in VO_2peak between age groups. The PAL values for seven days differed significantly between age groups. Adolescents 13–15 year olds were more physically active than the 16–18 year olds. Internally in category "METs < 3", boys were physically active for fewer minutes than to girls. The age group 13–15 year had significantly fewer minutes registered at "METs < 3" compared to the 16–18 year olds, while in the "METs 3–6" (minutes) the 13–15 year olds had significantly more minutes registered than age group 16–18. There were no significant age and gender differences concerning answers on physical activity in any of the questionnaires (Table 3).

According to Landis and Koch divisions of agreement 30, the WHO HBSC questionnaire indicated a substantial overall reliability (frequency r = 0.73 and duration r = 0.71) (Table 4). Significant differences were found between girls and boys on the WHO HBSC frequency question (r = 0.87 and r = 0.59 respectively), and between age groups on the duration question (13–15 years r = 0.62 and 16–18 years r = 0.85).

The overall reliability of the IPAQ questionnaire varied for the different physical activity categories. The lowest correlation was found for walking (minutes per day) (r = 0.10), while the highest correlation was found for walking (days per week) (r = 0.62). The IPAQ walking (days) question showed statistically significant difference between genders (girls r = 0.53 and boys r = 0.77) and age groups (13–15 years r = 0.81 and 16–18 years r = 0.37) (Table 4).

For the total population, a statistically significant correlation was found between VO_2peak and the questions on both frequency (r = 0.39) and duration (r = 0.33) in the WHO HBSC questionnaire (Table 5). The correlation was also significant when the answers were divided into three categories (Table 5). Girls had a higher correlation between the WHO HBSC questionnaire and VO_2peak (r varied between 0.41 and 0.55) compared to boys (r varied between 0.21 and 0.31), and correlations were statistically significant in girls only.

The correlation coefficients of the WHO HBSC questions measured against the TEE and PAL was low (Table 5).

Vigorous activity (days per week) measured in the IPAQ and classified into three categories, was significantly correlated with VO_2peak (Table 5). Vigorous activity (minutes per day) and walking (minutes per day) in the IPAQ correlated negatively with VO_2peak, indicating that more minutes of both vigorous activity and walking was associated with a lower VO_2peak. There was, however, a significant correlation between the IPAQ expressed as walking (minutes per day) and VO_2peak for girls (r = -0.41).

The correlation coefficient between the IPAQ questions and PAL was significant for walking (minutes per day) including all (r = 0.43) and for boys when split by gender (r = 0.61). The IPAQ question on sitting (minutes per day) showed a significant negative correlation with PAL for boys (r = -0.68) and was significantly correlated with TEE in girls (r = 0.54). The other associations between the IPAQ questions and the ActiReg measures had a low correlation and were not significant (Table 5).

In general the reliability of WHO HBSC questionnaire was comparable to a study among Australian high school students 16. An interesting observation in our study is that the WHO HBSC questionnaire tended to be more reliable for girls. A reliability study by Treuth et al. 31 found no gender difference in the Fels physical activity questionnaire for children. Few studies have however, focused on possible gender differences. The gender differences in our study could be due to the fact that girls tend to be more precise in their answers. Girls probably are less competitive than boys concerning physical activity, and thus they may be more "honest" in their answers. Girls value different things and they do not need to emphasise themselves as very physically active. This may strengthen the reliability patterns for girls.

Our results also revealed a difference between age groups. The WHO HBSC questionnaire was more reliable for the oldest group. This is similar to what Treuth et al. found in their study 31. The lower reliability in the 13–15 year old could be due to a failure to interpret the questions correctly. Those 16–18 years probably had a better understanding of its contents, explaining the higher correlation in the oldest group. The reliability of the IPAQ was lower than that of the WHO HBSC questionnaire. This could be explained by the less structured format (open-ended questions) in the IPAQ. The lower test-retest reliability of the IPAQ could also be related to the reference period, because the questionnaire focuses on the last seven days, while physical activity may change considerably from one week to the next.

Concerning the IPAQ there were only minor differences between genders, except for the question about walking/days, where boys had a higher reliability than girls. The same was shown in the IPAQ, 12-country reliability and validity study among adults 17.

In our study, we used two objective methods to validate the questionnaires; physical activity (TEE and PAL) and physical fitness (VO_2peak). Physical activity is difficult to measure. Validating self-reported physical activity by questionnaire is therefore a great challenge. Different methods have been applied as validation criteria; accelerometer, pedometer, recall-logs, heart-rate monitoring and different energy expenditure methods 32. The doubly labelled water method (DLW), indirect calorimetry and direct observations are the most reliable and valid measurements. The DLW method has drawbacks like financial costs and limitations due to the laboratory test situation 32. Accelerometers have become increasingly popular as measurement tools for physical activity. However, inaccuracies, especially related to underestimation and inconsistency in the definition of what constitutes light, moderate, and vigorous activity have been reported 333435. The ActiReg, used in this study, has been validated against DLW and indirect calorimetry. Hustvedt et al. 24 found good agreement in moderate activity groups (moderately activity from 38 to 104 min per day) with a mean PAL of 1.70, which is comparable to our mean and therefore support the representativeness of our sample. The ActiReg has some limitations for PAL above 1.70. Applied on a population with low PAL (patients with chronic obstructive pulmonary disease), the ActiReg is found to be a valid tool to assess energy expenditure and distinguish between both the low intensity activity range and moderate to high intensity activity range of physical activity 36. ActiReg is also able to recognise activities such as sitting and different intensities of movements. We have analysed these separately, but these analyses showed no changes in results (data not shown)

Physical fitness has been related to total and cardiovascular mortality and heart disease. Therefore, cardiorespiratory fitness, measured as VO_2peak, has been preferred as the validation criterion for physical fitness in the last decades, and is considered the gold standard in the assessment of exercise tolerance 37.

There was a significant correlation between the WHO HBSC questionnaire and VO_2peak for all, except for boys when analysed stratified by gender. When each question in the WHO HBSC questionnaire was analysed separately, the frequency question had a higher correlation than the duration question. The same trend was evident when split by gender. A possible explanation for the differences in the dimensions (duration and frequency) is that the frequency question, which inquires days per week, estimated physical activity more precisely than the duration question, which requests hours per week of physical activity. Another explanation could be related to the fact that days per week are a rougher estimate than hours per week. The IPAQ had a low validity measured against VO_2peak, except for the question about vigorous physical activity during the last 7 days for all. This corresponds to previous research showing that vigorous activity is easier to recall than light activity 38. Craig et al. 17 reported a typical correlation coefficient for the IPAQ was 0.30 for validity. In our study, the IPAQ recoded in three categories had an acceptable correlation against VO_2peak for all (0.32) as well as separately for girls when split on gender. Nevertheless, each question separately was not a valid measure of physical fitness. The IPAQ seemed to be an acceptable instrument when the questions were compiled. This is important because physical activity is most often recoded and classified using a scoring protocol in epidemiological studies.

The validity for single IPAQ questions within "moderate activity", "walking" (days per week) and "sitting" was poor. "Walking", expressed as minutes per day, was negatively correlated to VO_2peak. This probably means that the girls reporting walking for small distances do not perform vigorous physical activity, and that those with a high intensity activity associated with high cardiorespiratory fitness are inclined not to report lower physical activity like walking. To improve cardiorespiratory physical fitness sedentary persons need an intensity 40 to 60% of maximal aerobic power, corresponds to being slightly out of breathing or sweating 39. In general an exercise intensity above 80% to 90% of the individuals' maximal aerobic power (vigorous activity; MET > 6) is recommended to increase VO_2peak 40.

The ActiReg registered all activity performed by the participants for seven consecutive days. It was surprising that only the IPAQ questions about walking and sitting expressed as minutes a day showed significantly negative correlations against the TEE and PAL. Adolescents who reported many walking minutes had a high PAL value. For the IPAQ measured as sitting (minutes per day), these findings indicated that the girls who reported many sitting minutes had a high TEE. In the boys, however, we found the opposite; namely that those who reported many sitting minutes had a lower TEE, which we would expect. The low correlations between IPAQ and TEE/PAL could be related to underreporting of vigorous and moderate activity. We also found this underreporting in the WHO HBSC questions and this could explain the low validity measured against the ActiReg. It is difficult to explain the lack of correlation between the questionnaires and the ActiReg. Based on our results one consideration is that the WHO HBSC questionnaire and the IPAQ have questions which are related to activities that increase cardiorespiratory fitness 1519, and therefore correlate better with VO_2peak than TEE and PAL. However, this large difference between the answers and ActiReg could also be caused by the difficulties in creating accurate questions, and this could be an indication that we should prefer objective methods to measure physical activity in youth 41. Another explanation might be that our participants were younger than the age group for which the IPAQ was designed, and thus might not fully understand the questions. Recall bias in questionnaires, especially among adolescents, may influence the retrospective response. Active adolescents tend to overestimate physical activity, whereas obese adolescents underestimate physical activity 42. These variations may result in weaker correlations, thus influencing the validity. Because regular physical activity over a long period leads to physical fitness, we would expect good correlation between answers in questionnaires on physical activity and both ActiReg and VO_2peak.

VO_2peak is a more stable measure than physical activity. Physical activity may change daily, and from one week to next week, while physical fitness does not change considerably in 2–3 weeks' time. A possible bias could be related to the reference period, and we therefore did not find correlations on the criterion physical activity measure.

In our study, the questionnaire answers tended to underestimate physical activity, compared to the ActiReg (Table 3). The individual variations and the underestimations could be the explanation on the low validity, compared with PAL and TEE, and illustrates the difficulty to capture the individual energy expenditure in questionnaires 43.

For an accurate validation, the strength of our study is the use of two objective measures to validate the questionnaires. The sample size in this study is an additional aspect. Our response rate of 35% is, however, rather low and might introduce a risk of an overrepresentation of those who are most physically active. Based on the participants' cardiorespiratory fitness and BMI, our population were, however quite comparable to those in other studies 4445, including the Young-HUNT study, including 90% of the population 13–19 years of age in Nord-Trøndelag County (data not shown). This indicates a low selection bias of our population.

Our findings of higher correlations with VO_2peak than TEE and PAL could be because adolescents report vigorous activity most precisely. Respondents with a high-energy expenditure may not necessarily have high VO_2peak. Adolescents, who perform vigorous physical activity and thereby have a high VO_2peak, may do little moderate activity and therefore have relatively lower total energy expenditure. Although physical activity and physical fitness are two different dimensions, they are linked and both are correlated to health and survival 13.