Validation of two short questionnaires assessing physical activity in colorectal cancer patients

The present validation study was a sub-study of the ongoing CRC-NORDIET study, of which design and methods have been published elsewhere [38]. In brief, the aim of the CRC-NORDIET study is to investigate the effect of a diet similar to the Norwegian FBDG [3] on disease-free and overall survival among CRC patients post-diagnosis [38]. The risk factors shown to be related to CRC, i.e. diet and physical activity, are included in the Norwegian FBDG [40]. The CRC-NORDIET study is a prospective randomised controlled intervention trial, randomising 500 CRC patients into one of two study groups (i.e. 250 to diet intervention group and 250 to the control group). All patients are invited to the Study centre 3 times during the intensive 1-year intervention (i.e. at baseline 2–9 months post-surgery and at the two visits 6- and 12 months after baseline), and subsequently followed up for 14 years. Both study groups are offered equal recommendations on PA [38].

All patients from both study groups in the CRC-NORDIET study, who attended the follow-up at 6 months after baseline of intervention (i.e. 2–9 months post-surgery) from January 2014 to October 2015, were invited to take part in the present validation study. The patients were men and women aged 50–80 years old, with a confirmed CRC (ICD-10 C18–20), and staged I-III (i.e. locoregional disease without metastasis) according to the TNM staging system [41]. None of the patients included in the validation study underwent chemotherapy during the time-frame covered by the physical assessment methods used in the validation study (e.g. mean time from last chemotherapy injection to the start of the validation study (i.e 6 months after baseline) was 155 days among the 15% who received adjuvant treatment). During the 6-months visit, the patients completed the self-administered NORDIET-FFQ and HUNT-PAQ. In addition, they received the SenseWear Armband Mini (SWA), which was returned by mail to the CRC-NORDIET-study at the end of the test period of 7 days. Exclusion criteria for the present study were pacemaker implantation, not completed questionnaires or not wearing the SWA.

To characterize the subjects, anthropometric measurements (weight, height, and hip-and waist circumference) and physical tests (hand-grip strength and 30-s sit-to-stand test) were measured by the researchers of the CRC-NORDIET study during all visits at the Study centre, as previously described elsewhere [38]. In addition, education level and smoking status were self-reported by completion of questionnaires during the visits at the Study centre. Information about tumour location status was retrieved from medical records in cooperation with the hospital personnel. Energy expenditure (kJ/d) was estimated from the SWA.

The NORDIET-FFQ (available upon request to corresponding author), was designed to report both dietary intakes and PA in recent weeks (i.e. the last 1–2 months). The validity of the dietary items of the NORDIET-FFQ has been published elsewhere [42]. Completion of the NORDIET-FFQ as well as protocol for data handling followed the same procedures as described in detail in Henriksen et al. [42]. In brief, the NORDIET-FFQ was completed by the patients at the Study centre. The completed questionnaires were scanned by trained researchers and the image files translated into data files using the Cardiff Teleform 2006 Software (6.0) (Datascan). The last two questions in the NORDIET-FFQ asked for PA with two different intensities, MPA and VPA estimated in pre-defined intervals of frequency per week and duration in minutes. The explanatory texts for both PA questions included examples of typical activities: moderate intensity was exemplified by brisk walking, household chores or other activities resulting in slight breathlessness, and vigorous intensity was exemplified by running, cross-country skiing or other activities resulting in high breathlessness. The question about frequency contained different responses in times per week (time divided by seven) coded as follows: 0 = 0, 1 = 0.14, 2 = 0.29, 3 = 0.43, 4 = 0.57, 5 = 0.71, 6–7 = 0.93 and 8+ = 1.37 (added 20% to 8 and divided by 7). Moreover the responses of duration in minutes were coded as follows: 1–4 = 2, 5–9 = 7, 10–15 = 12.5, 16–20 = 18, 21–30 = 25.5, 31–45 = 38, 46–60 = 53 and 60 + =72 (added 20% to 60). Amounts in minutes of PA per day for each intensity were calculated by multiplying frequency (i.e. times per day) with duration (i.e. minutes each time). This resulted in variables of total-MPA and total-VPA, of which all minutes within each intensity were included (Additional files 1 and 2). Additionally, categories of 10-min bouts were computed and defined as ten or more consecutive minutes within each intensity level. This resulted in data on 10-min bouts of MPA and VPA.

The HUNT-PAQ was based on the questionnaire used in the HUNT 3-study [39]. Only the five questions about PA as described in Kurtze et al. [43] were used. The question about frequency contained the following responses: Never and Less than once a week, both coded as 0, Once a week coded as 1, 2–3 times a week coded as 2.5, and Almost every day coded as 7. The question about duration of activity contained the following responses: Less than 15 min coded as 12 (subtracted 20% from 15), 15–29 min coded as 22, 30–1 h coded as 45 and more than 1 h coded as 72 (added 20% to 60). The products of frequency and duration were weighed by intensity level, i.e. Low, Moderate or Vigorous coded as 1, 2 and 3, respectively. The low intensity level was not evaluated in the present study. Additionally, there was a question about daily sedentary time in hours on a usual day (not included sleeping at night-time). The questionnaire generated data on activities in bouts of 10 and more consecutive minutes for each intensity level, such as MPA and VPA.

The objective PA monitor SenseWear Armband Mini (SWA) (BodyMedia, Pittsburgh, Pennsylvania, USA) was used to record daily PA and energy expenditure during seven consecutive days [44]. A priori, we defined a valid day of recording if the wear time was ≥80% of a 24-h sampling period. The SWA has previously been validated against double-labelled water [45], indirect calorimetry [46] and other accelerometers [47] in adults and cancer patients. It monitors physiological data such as heat flux, galvanic skin response, 3-axis accelerometer and skin temperature. The SWA was pre-programmed by the researcher with the co-predictors such as weight, height, birth date, sex, smoking status (smoker/non-smoker) and whether the participant was left or right handed, and placed around the triceps muscle halfway on the upper non-dominant arm. The participants were instructed to continue their normal activity level while wearing the SWA. Water-based activities were not recorded by the SWA because the monitor is not waterproof. Participants were asked to remove the SWA when performing activities in water. All data were retrieved from the SWA to a computer with the SenseWear Professional Software Version 7.0 BodyMedia Inc. (Pittsburgh, Pennsylvania, USA).

Activity intensities were integrated into algorithms, providing estimates of energy expenditure expressed in metabolic equivalents (METs). The definition of 1 MET is the amount of oxygen consumed while sitting at rest and is equal to 3.5 ml O₂ per kg bodyweight per min [48]. Moderate and vigorous intensities were defined as 3–6 and > 6 METs, respectively, as calculated by Ainsworth and coworkers [49, 50]. Sedentary time was defined as all daily activities ≤1.5 METs, of which nighttime sleep was removed (a priori defined as from 12 midnight to 6.00 a.m.). All activities were calculated and expressed in minutes or hours per week (sedentary time).

The SWA records all intensities in 1-min intervals, which were translated into different categories of data such as total-MPA and total-VPA. Furthermore, the data were also computed into 10-min intervals, which were defined as ten or more consecutive minutes within the relevant intensity level. The bouts of 10-min were calculated for the two intensity levels, which gave data on bouts of 10-min for MPA and VPA.

The CRC patients were advice to follow the recommendations of moderate-to-vigorous intensity PA (MVPA), MPA and VPA. MVPA was defined as ‘MPA + (VPA*2)’ [51] and the cut-off points for fulfilling recommendations of MVPA and MPA were at least 150 min per week, and at least 75 min per week for VPA. The activities should be in bouts of 10 and more consecutive minutes.

Statistical analyses were performed by use of IBM SPSS Statistics, version 22. Results were considered significant with two-sided p-values below 0.05. Normal distribution was checked for all data by inspection of histograms, normal Q-Q-Plots and Kolmogorov-Smirnov test (p > 0.05).

All anthropometric measurements and other characteristics of the study population were normally distributed and are presented as means with standard deviations (SD). The categorical data are presented as frequency with percentages and compared by the Fischer exact test and Pearson chi-square test. As most of the estimates of PA from the SWA, NORDIET-FFQ and HUNT-PAQ were not normally distributed, they are presented as medians and 5th - and 95th percentile. Statistical significant differences in median activity between the two questionnaires compared to the SWA were tested with Wilcoxon Signed-Rank test for paired data. Bland-Altman plots with limits of agreements were used to explore the differences between the measurements from the two methods (i.e. questionnaire minus SWA) plotted against the average of the two measurements, for each individual subject, as well as to identify outliers [52, 53]. Systematic under- or over-reporting was tested by linear regression with MVPA and MPA from SWA as the independent variable and the difference between the questionnaires and SWA as the dependent variable. Ranking of individual time in PA and degree of association between the continuous variables from the two different methods were analysed by Spearman Rank Order Correlation (rho). The ability of the NORDIET-FFQ and the HUNT-PAQ to classify the individual’s activity intensity into the same category as the SWA was estimated by the use of sensitivity and specificity analysis. Sensitivity was defined as the number of subjects reported not to fulfil the MVPA with both questionnaires and SWA as a percentage of those who had reported not to fulfil the MVPA with the SWA. Specificity was defined as the number of subjects reported to fulfil the MVPA with both questionnaires and SWA as a percentage of those who had reported to fulfil the MVPA with the SWA.

In order to detect a Pearson correlation coefficient of 0.5 or higher between the test-method (i.e. questionnaires) and the reference method (i.e. SWA), a sample size of 38 men and 38 women was required to achieve a significance level of 5% and power of 90% [54]. With an expected consent rate of 90% and exclusion rate of maximum 5%, we aimed to invite 90 participants in order to include 76 participants to the present study.

Participants wore the SWA monitors for 97.9 ± 3.8% (mean, ± SD) of the time during 6.2 ± 0.8 days (mean, ± SD) of monitoring.

Median duration of PA estimated from the NORDIET-FFQ, HUNT-PAQ and the SWA is presented in Table 2. There was no significant difference between the NORDIET-FFQ and SWA for the measure for activity of moderate-to-vigorous intensity PA (MVPA), either for the total population (p = 0.897) or when analyzing sexes separately. The HUNT-PAQ, however, significantly measured MVPA differently on a group level compared to SWA (p < 0.001).

Mean differences in PA measures (i.e. questionnaire minus SWA) with corresponding limits of agreements between questionnaires and the SWA are shown in Bland Altman plots in Fig. 1. MVPA (Fig. 1a) was reported only 4% differently with the NORDIET-FFQ compared to the SWA (mean difference and limits of agreement − 12 ± 624 min/week). The under-estimation of MVPA (Fig. 1d) with the HUNT-PAQ was 58% compared to SWA (mean difference and limits of agreement − 162 ± 576 min/week). Moreover, the Bland Altman-plot for MVPA revealed an increase in the differences between both questionnaires and SWA with increased PA level. Additionally, the differences were randomly and evenly distributed above and below the mean difference for both questionnaires compared to SWA up to about 250 min/week (Fig. 1a and d). The slope of the linear regression was negative and significant for both questionnaires (β = − 0.79, p < 0.001(NORDIET-FFQ) and β = − 0.85, p < 0.001 (HUNT-PAQ)), indicating under-reporting at higher levels of PA (MVPA from SWA as independent variable). Removing of two outliers shown in the Bland Altman plot (Fig. 1a) did not have any effect on the limits of agreement or the linear regression (data not shown). Therefore, they were included in further analyses. The Spearman’s rho of the MVPA was insignificant and weak (rho = 0.08, p = 0.509) for the NORDIET-FFQ, indicating that the questionnaire was not able to rank individual time in MVPA. Likewise, in the HUNT-PAQ, ranking of individual time in MVPA was also poor (Spearman’s rho of 0.14, p = 0.238). The NORDIET-FFQ captured 63% individuals fulfilling the recommendation of MVPA (specificity), whereas only 29% of those in need of PA counselling (sensitivity) (Table 3). The HUNT-PAQ was better at capturing individuals not fulfilling the recommendation of MVPA (sensitivity of 71%), but worse in identifying those who did (specificity of 36%).

Time spent in activity of moderate intensity PA (MPA) did not differ significantly by sex between the NORDIET-FFQ and SWA, whereas in HUNT-PAQ time in MPA was measured significantly different from SWA on group level (Table 2).

MPA was under-reported in both questionnaires as shown by the mean differences and limits of agreement from the Bland-Altman plots of − 83 ± 512 min/week and − 153 ± 523 min/week from the NORDIET-FFQ and HUNT-PAQ, respectively (Fig. 1b and e). The differences of MPA were randomly and evenly distributed above and below the mean difference for both questionnaires compared to SWA up to about 250 and 200 min/week with the NORDIET-FFQ and HUNT-PAQ, respectively (Fig. 1b and e). Linear regression also revealed a significant systematic under-reporting at higher levels of PA (MPA from SWA as independent variable) in both questionnaires (β = − 0.78, p < 0.001(NORDIET-FFQ) and β = − 0.83, p < 0.001 (HUNT-PAQ)).

Ranking of individual time in MPA was fair among women only (Spearman’s rho = 0.37, p = 0.027) with the HUNT-PAQ, but weak and insignificant with the NORDIET-FFQ (Spearman’s rho = 0.01, p = 0.402). The HUNT-PAQ captured 74% of individuals not fulfilling the recommendation of MPA (sensitivity), but only 36% of those who did (specificity). Both sensitivity and specificity for MPA were low with the NORDIET-FFQ (Table 3).

Median time in activity at vigorous intensity PA (VPA) was reported significantly differently between both questionnaires and SWA. The Bland Altman plot for VPA revealed an over-reporting of 36 ± 176 min/week (mean difference ± limits of agreement) with the NORDIET-FFQ, which increased with increased activity (Fig. 1c). Since only one participant reported VPA with the HUNT-PAQ, data from this activity are not presented. Moreover, the NORDIET-FFQ identified 75% of the individuals not fulfilling VPA.

Amount of sedentary time was only measured in the HUNT-PAQ. Median time in sedentary intensity was significantly different between the questionnaire and SWA (Table 2). The Bland Altman plot revealed a high under-reporting of about 52% (6.5 h/day) compared to SWA, which decreased with increased sedentary time (Fig. 1f). However, the questionnaire was able to rank individuals according to sedentary time among women (r = 0.36. p = 0.035), but not among men or all participants in total.

Looking at each method separately, participants who reported to fulfil the MVPA of at least 150 min per week were 66% and 33% with the NORDIET-FFQ and HUNT-PAQ, respectively. However, only 55% of the participants actually met this recommendation according to the SWA (Table 4).