Validity of the Physical Activity Questionnaires IPAQ-SF and GPAQ for Cancer Survivors: Insights from a Spanish Cohort

Abstract Regular physical activity (PA) decreases mortality risk in survivors of breast and colorectal cancer. Such impacts of exercise have prompted initiatives designed both to promote and adequately monitor PA in cancer survivors. This study examines the validity of 2 widely used self-report methods for PA determination, the International Physical Activity Questionnaire short version (IPAQ-SF) and Global Physical Activity Questionnaire (GPAQ). Both instruments were compared with the triaxial accelerometry (Actigraph) method as an objective reference standard. Study participants were 204 cancer survivors (both sexes, aged 18–79 years). Compared with accelerometry, both questionnaires significantly overestimated PA levels (across all intensities) and underestimated physical inactivity levels. No differences were detected between the 2 questionnaires except for a shorter inactivity time estimated by GPAQ (p=0.001). The Bland and Altman method confirmed that both questionnaires overestimated all PA levels. Receiver operating characteristic (ROC) analysis classified IPAQ and GPAQ as fair and poor predictors, respectively, of the proportions of survivors fulfilling international PA recommendations (≥150 min·week−1 of moderate-vigorous PA). IPAQ-SF showed a higher sensitivity but lower specificity than GPAQ. Our data do not support the use of IPAQ-SF or GPAQ to determine PA or inactivity levels in cancer survivors.


▼
The possible link between levels of physical activity (PA) and cancer is receiving increasing attention. So far, research has shown a rising incidence of both cancer and physical inactivity, and the latter has been held accountable for 10 + % of the disease burden of breast and colon cancer, both of which are highly prevalent in the developed world [27]. It has been estimated that ~1/3 of adults worldwide engage in < 150 min · week − 1 of moderate-intensity PA such as brisk walking [17]. Higher levels of regular PA have been associated with: (i) a lower cancer risk in the general population [1,3,4,6,13,32,48,55,56] and (ii) a lower mortality risk among cancer survivors, particularly of breast and colorectal cancer [8,20,46]. Further, there are also data to support an overall dose-dependent benefit of PA, at least for the risk of breast cancer [26,30,31,55]. Thus, besides a need for effective PA assessment in cancer-related epidemiological research, adequate tools for this purpose would be useful to monitor cancer survivors or populations at can-cer risk [29,44]. In effect, PA can be objectively quantified through accelerometry [10] whereby minute-by-minute recordings of PA can be made. Notably, accelerometers were used to show that PA levels were low, i. e., well below minimum recommendations of 150 min · week − 1 , in US [35,36,50] or Canadian and Australian cohorts of cancer survivors [34]. At least 3-5 days of accelerometer monitoring is required to reliably estimate habitual PA [49] and it is generally accepted that the device should be worn for a minimum of 10 h per day [39]. To simplify data collection, the more conventionally used method of obtaining self-reported PA data through questionnaires may be used. This method is also inexpensive and generally well-accepted by participants, although the validity of obtained data is more questionable [41]. Of several questionnaires developed for PA assessment [51], 2 have been widely used: the International Physical Activity (IPAQ) and Global Physical Activity (IPAQ) questionnaires. The IPAQ was originally developed as an instrument for cross-national records of PA [9]. Today, a simpler, short form (IPAQ-SF) and a • ▶ tumor • ▶ self-reported physical activity • ▶ physical inactivity • ▶ accelerometer Abstract ▼ Regular physical activity (PA) decreases mortality risk in survivors of breast and colorectal cancer. Such impacts of exercise have prompted initiatives designed both to promote and adequately monitor PA in cancer survivors. This study examines the validity of 2 widely used selfreport methods for PA determination, the International Physical Activity Questionnaire short version (IPAQ-SF) and Global Physical Activity Questionnaire (GPAQ). Both instruments were compared with the triaxial accelerometry (Actigraph) method as an objective reference standard. Study participants were 204 cancer survivors (both sexes, aged 18-79 years). Compared with accelerometry, both questionnaires significantly overestimated PA levels (across all intensities) and underestimated physical inactivity levels. No differences were detected between the 2 questionnaires except for a shorter inactivity time estimated by GPAQ (p = 0.001). The Bland and Altman method confirmed that both questionnaires overestimated all PA levels. Receiver operating characteristic (ROC) analysis classified IPAQ and GPAQ as fair and poor predictors, respectively, of the proportions of survivors fulfilling international PA recommendations ( ≥ 150 min · week − 1 of moderate-vigorous PA). IPAQ-SF showed a higher sensitivity but lower specificity than GPAQ. Our data do not support the use of IPAQ-SF or GPAQ to determine PA or inactivity levels in cancer survivors. long form (IPAQ-LF) exist, both of which involve 7-day recall of PA. The short version is recommended for national monitoring [9]. Other than a recent study reporting that IPAQ-LF overestimated PA levels in women with breast cancer [21], no study has assessed the validity of IPAQ in cancer survivors. The other questionnaire, GPAQ, was developed by the World Health Organization for PA surveillance across countries and no validation data are available for cancer survivors. The present study was designed to determine the validity of IPAQ-SF and GPAQ in a Spanish population of cancer survivors using accelerometry as the reference standard.

▼
This cross-sectional study examined a cohort of cancer survivors undergoing follow-up at a large hospital in Madrid (Hospital Universitario de Fuenlabrada, Madrid, Spain). Accelerometry PA data for this cohort have been recently reported by our group [42]. The study protocol met the ethical standards of the journal [18] and adhered to the tenets of the Declaration of Helsinki and the statements in Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) [52,53]. The study received Institutional Review Board (Hospital Universitario de Fuenlabrada) approval. Written informed consent was obtained from all participants. As detailed elsewhere [42], cancer survivors (n = 204) were recruited from the hospital's Oncology Department between May 2011 and June 2012 if they met the following criteria: aged 18-79 years; able to walk independently and understand requirements for valid accelerometry; time after cancer diagnosis ≥ 1 year; time after last anti-cancer treatment (chemotherapy, radiotherapy or surgery) ≥ 3 months; no evidence of tumor recurrence or metastasis. All accelerometry data (see below) were analyzed by the same experienced observer (C.F.-L.). To check the reliability of these measurements, data from 18 randomly selected participants (9/ group) were analyzed by an external observer who was blinded to the results obtained (N.G.). A third researcher with expertise in accelerometry recordings compared the results obtained by the 2 independent observers (A.S.-L.). All subjects wore a triaxial Actigraph GT3X monitor device (Actigraph, Pensacola, FL, USA) for a minimum of 5 (and a maximum of 10) consecutive days. The validity of this device to quantify free-living PA, at least within frequencies (~2-4 Hz) common to most types of daily activities, has been previously shown by our group [45]. In each participant, monitoring was continued for a minimum of 5 days including 2 weekend days, and a minimum of 10 h of complete accelerometry data were recorded per day. For participants providing more than 5 consecutive days of recordings, only the data for the last 5 days including 2 weekend days were used [42]. The predefined epoch and sample rate were 15 s and 30 Hz, respectively. Data were analyzed using ActiLife5 LITE software (Actigraph, Pensacola, FL, USA). Outcome variables were expressed as average intensity (counts · min − 1 ) and counts were transformed into time (average min · day − 1 and total min · week − 1 ) engaged in either physical inactivity or light, moderate and vigorous PA using the following cut-offs [14]: inactivity < 100 counts · min − 1 ; light PA = 100-1 951 counts · min − 1 ; moderate PA = 1 952-5 724 counts · min − 1 [corresponding to 3-5.9 metabolic equivalents (METs), where 1 MET is equivalent to an oxygen consumption of 3.5 ml · kg − 1 · min − 1 ]; and vigorous PA ≥ 5 725 counts · min − 1 ( ≥ 6 METs). For any PA to be considered 'moderate PA', 10 consecutive minutes of observation had to exceed the moderate intensity cut-off, with allowance for a maximum of 2 observations falling below the cut-off during that period (8 out of 10 min had to be above the cut-off). All participants completed the Spanish versions of IPAQ-SF [38] and GPAQ upon return of the accelerometers (within the same day and with no predetermined or randomized order for the 2 questionnaires) [2].

Data analysis
Outliers in questionnaire and accelerometry data, i. e., data outside the inter-quartile range, were removed by constructing box and whisker plots. Since PA data did not follow a normal distribution, we used the Wilcoxon signed-rank to compare the differences between questionnaire and accelerometry data. Agreement between the 2 questionnaires (IPAQ-SF and GPAQ) vs. accelerometry was assessed with the Bland-Altman technique [5]. Association between the difference and magnitude of measurements (i. e., heteroscedasticity) was examined by regression analysis, entering the difference between the 2 methods (IPAQ-SF or GPAQ vs. accelerometry) as the dependent variable and the averaged value between them as the independent variable [45]. The χ 2 test was used to evaluate the difference between: i) the percentage of subjects who fulfilled minimum PA recommendations according to IPAQ-SF, GPAQ and accelerometry data and ii) the sensitivity and specificity of the IPAQ-SF and GPAQ. Finally, receiver operating characteristic (ROC) curves were constructed to determine the areas under the curve (AUC) and 95 % confidence intervals, specificity and sensitivity. According to AUCs, the 3 methods were classified as excellent (≥ 0.90), good (0.80 to 0.89), fair (0.70 to 0.79) or poor (< 0.70) predictors of the fulfillment of minimum PA recommendations. All statistical tests were performed using PASW software (v. 22.0 for MAC, Chicago). Table 1 shows the main clinical characteristics of study cohort.

▼ • ▶
Patients did not have any evidence of cancer and all had completed the treatment. Almost half (46 %) of the total sample had breast cancer and 26 % had colorectal cancer. Outliers were identified from IPAQ-SF, GPAQ and accelerometer data, resulting in 11 values from 204 ( • ▶ Fig. 1). Finally, complete accelerometer data were available for 177 cancer survivors whereas complete IPAQ-SF and GPAQ data were available for 183 participants. Levels of PA obtained through IPAQ-SF, GPAQ and accelerometry and χ 2 test for adherence to minimum PA recommendations are provided in • ▶ Table 2. Compared with the accelerometry data, both questionnaires significantly overestimated PA levels across all intensities and underestimated inactivity levels (p < 0.01). No differences were detected between the 2 questionnaires except for a shorter inactivity time estimated by GPAQ (p = 0.010). The aforementioned results did not change essentially when analyzing both sexes separately or the most frequent type of cancer in the present cohort (breast cancer) vs. the other type of tumors) (Supplementary File 1).  Table 4). IPAQ-SF showed a higher sensitivity (true positive rate) but lower specificity (true negative rate) than GPAQ.

▼
This study is the first to compare the performance of 2 widely used PA questionnaires (IPAQ-SF and GPAQ) using accelerometer data as reference to determine PA and inactivity levels in cancer survivors. General agreement between self-reported and accelerometer-measured PA levels was poor and correlation coefficients were lower than recommended [51]. Hence, according to our data, these questionnaires cannot be considered the best option to assess PA or inactivity levels in this subject population, as they tend to overestimate PA levels while underestimating inactivity time. A systematic review concluded that IPAQ-SF overestimated PA as measured by objective criterion by an average of 84 % [28]. Greatest differences between self-reported and accelerometry data were produced for vigorous PA and inactivity levels. Thus, these questionnaires are not even useful to identify individuals who do not meet minimum PA recommendations. Given the recently reported benefits of PA against cancer risk or mortality [44] and provocative data indicating a relationship between PA and lower cancer mortality [24,25,33], our findings have implications for the adequate quantification of PA in cancer survivors. Subjective methods (questionnaires) of PA determination have been used in the most influential studies leading to international recommendations, and these methods are considered useful for large-scale screening as they are inexpensive and easy to administer. However, questionnaires often seem to overestimate (or even underestimate) PA levels [43]. Self-reported PA data are susceptible to error due to misreporting because of a tendency to reflect social desirability or due to cognitive limitations related to recall or comprehension. In effect, self-reporting of PA can be particularly difficult in the elderly. Individuals also tend to overreport PA and underestimate sedentary pursuits such as watching television [23]. Sims et al. noted that people who had been specifically encouraged to exercise reported a greater volume of PA compared with PA levels inferred from heart rate data [47]. In a recent report, the authors argued that the use of questionnaires to estimate PA and inactivity levels could result in failure to detect real relationships with metabolic and vascular disease risk factors or underestimation of the strength of those relationships [8]. IPAQ was developed to evaluate an individual's level of PA in terms of getting around, leisure time activities and household tasks. The questionnaire targets the age range 15-65 and until further development and testing is undertaken, its use in older or younger persons is not recommended. IPAQ-SF was designed for surveillance studies in which time is limited and consists of 8 items to estimate time spent conducting moderate to vigorous PA and inactivity (time spent sitting). This short form assesses 4 PA intensity levels: 1) vigorous PA such as aerobics, 2) moderateintensity PA such as leisure-cycling, 3) walking and 4) sitting. The first comprehensive validation of IPAQ-SF was conducted across 12 countries, and correlations with data obtained using the uniaxial CSA model 7164 accelerometer were reported [9]. Spain was not included among these countries. However, the wide range of Spearman correlations reported raises concerns about the validity of the questionnaire for use across populations. Indeed, the single study using the expensive doubly labeled water technique revealed marked underestimation of questionnaire-derived energy expenditure at higher PA levels [37]. In a systematic review of 23 validation studies of the IPAQ-SF it was also observed that in most studies, this questionnaire only shows discrete correlation with objective measures of PA such as accelerometers [28]. Effectively, although some authors indicate a few exceptions (with vigorous PA and walking providing some acceptable correlations), none of the IPAQ-SF studies reviewed reached the minimal acceptable correlation standard of 0.5 recommended for objective measures. These authors make a call for additional well-designed studies. A recent metaanalysis showed overall convergent validity of IPAQ within each PA category; using a short form to estimate the amount of PA as a form of continuous measures was found acceptable if the primary interest of the study is not domain-specific measures [22]. showed poor sensitivity (71 %) and specificity (59 %) for predicting adherence to minimum PA recommendations. After its translation and adaptation for use in Latin American subjects [38,40], IPAQ-SF has also shown poor validity for assessing moderate to vigorous PA [40]. Finally analyzed data,i.e., subjects with complete data of IPAQ-SF, GPAQ and accelerometry Fig. 1 Flow diagram of participants and measurements. The validity of GPAQ determined here with respect to the reference accelerometry method was similar to that observed for IPAQ-SF, and this instrument also returned a poor negative predictive value [11]. In the present clinical setting, these 2 widely used questionnaires failed to identify people who were physically inactive. We would particularly need to identify these persons to target them for interventions promoting PA. Recent epidemiologic studies have reported on the health hazards of sedentary behavior especially sitting time. This behavior is perhaps one of the most difficult domains to assess through questionnaires as demonstrated by poor correlations with objectively measured sedentary time according to Helmerhorst [19] and confirmed by our data. A Norwegian study also detected large variations between self-reported (IPAQ-SF) and accelerometermeasured PA and sedentary time [12]. The higher volume of self-reported vs. accelerometer-measured vigorous-intensity PA detected here is consistent with the findings of others [12,15,41]. Thus, differences between the 2 types of data were enhanced for the higher activity levels, which is also in agreement with the results of other validation studies using different IPAQ versions [15,16,54].
The major strength of our study is that it is the largest to compare self-reported and accelerometry PA data in individuals with a history of cancer and the first to compare IPAQ-SF and GPAQ against accelerometer data in this type of population. Its main limitation is the small geographical area examined, reducing its external validity. A further limitation is that reliability was not assessed and that the order of questionnaire administration was not randomized. Accelerometry also has some inherent limitations including its inability to accurately assess the intensity of specific exercise modalities such as weight-lifting, cycling and swimming. The choice of somewhat arbitrary cut-offs to classify intensities could be viewed as another limitation of this method.
In conclusion, our data do not support the use of IPAQ-SF or GPAQ to determine PA or inactivity levels in cancer survivors. We propose the use of more objective methods to cover the need to accurately quantify PA levels and promote PA both in cancer survivors or other target populations and for research purposes.