RESULTS AND DISCUSSION
In this pilot study, 223 individuals were recruited and on average this cohort was 50 years old, which had been diagnosed 17 years ago, and of those who knew their HLA-B27 status 81% were positive (Table 1). At least half of participants could be considered to have active AS requiring management (52% attended a clinic at least twice a year and 38% were using anti-TNFα therapies). The majority of participants had comorbidities (75%), including chronic back pain, sciatica or osteoporosis (45%), cardiovascular problems (43%) and stomach or bowel problems (40%). The mean BASDAI, howRu score and IPAQ MET minutes reported over the 12-week period are summarised in Table 1.
Associations between demographic characteristics and BASDAI (Table 2) were generally as expected, with significantly elevated BASDAI scores observed in those with higher BMIs and those with more frequent rheumatology clinic visits or high numbers of comorbidities.23 Medication use did not correlate with levels of disease activity, except for patients using steroids who had the highest mean BASDAI (5.20 ± 2.15).
Physical activity, BASDAI and mental health
Higher physical activity levels were significantly associated with lower BASDAI even when adjusting for howRU (Table 3 and Figure 3). This supports previous studies where exercise or active lifestyles (even if formal exercise was minimal) have supported the management of AS.24 At least 1 hour of vigorous activity per week also lowered disease activity to below the clinical cutoff BASDAI 4 indicating the need for treatment review (mean BASDAI = 3.31 ± 2.07 versus 4.3 ± 2.17). Analysis of the individual components of BASDAI suggested that physical activity benefited pain, stiffness and tenderness, but not fatigue (Table 3). This contrasts with previous reports,25 where physical activity did improve fatigue levels. The association with peripheral and axial pain and activity was lost when adjusting the models for the howRU score, thus demonstrating the importance that mental health plays in the experience of pain.26
Health-related quality of life scores were very strongly inversely associated with BASDAI (Figure 4). A one-unit increase in the total howRu score was associated with a 0.56 (95%CI 0.50, 0.62, P < 0.001) reduction in the BASDAI score. Increased physical activity was associated with improved overall quality of life (Figure 5) and less pain, even when adjusted for BASDAI (Table 4). This fits with previous research where quality of life was found to very strongly associate with self-reported disease activity,27 and physical activity was noted to improve mental health.26 These findings may indicate that undertaking physical activity even during periods of higher disease activity is beneficial especially for quality of life.
Interestingly, physical activity did not affect the feeling of dependency (Table 4). This may be due to the longer term effects of living with AS, including concerns around the future,28 impacting an individual’s perception of their independence regardless of whether their view of how capable they are changes.
Types of physical activity
Qualitative data analysis on activity types resulted in the generation of a coding framework consisting of five themes, which applied across all activity intensities: exercise, recreation / hobbies, household (or work) activities, care activities, none and an ‘uncategorised’ for unsuitable listings. For each activity, intensity-level codes were generated within each theme and the range of formal exercise listed was extensive. The most frequently noted categories across moderate and vigorous levels were biking = running/walking > therapy > water-based sports > gym > land exercise classes > calisthenics > racquet games > ball games > contact sports (Table S2). Solo activities far outweighed team activities and this may reflect the variable nature of AS. Solo activities tend to require less planning and prior commitment, and therefore individuals can choose to partake depending on their daily wellness. Additionally, it may reflect feelings of confidence in abilities due to the noncompetitive nature of solo activities.28,29
Stratified analysis
Stratification of data by sex (Table 5) demonstrated that BASDAI in women was inversely associated with continuous IPAQ scores. However, this was not seen in men where more vigorous activity was required to see improvements in BASDAI. This may be partially explained by physical activity reducing symptoms that are more common in women, i.e. neck, knees and hip pain and tenderness.30 Physical activity benefitted health-related quality of life for men, with higher levels of physical activity improving howRu scores, but this association was not seen in women. This may be due to differences in the profile of AS symptoms typically seen in men and women.30
Associations between physical activity levels, BASDAI and quality of life were significant for individuals who were HLA-B27 positive (Table 5). No relationship was observed for HLA-B27 negative; however, numbers were low (28 participants). Additionally, in individuals with a disease duration of >10 years, physical activity significantly benefited quality of life, but not BASDAI. More recently, diagnosed participants doing an hour or more of vigorous activity per week associated with improved BASDAI, but negatively associated with quality of life. This may suggest that at early stages of the disease, physical activity negatively impacts the individual, possibly via the interaction of pain and mental health. Furthermore, individuals recording vigorous or moderated gardening (n = 98 versus no vigorous or moderate gardening n = 125) as one of their more frequent activities had higher average log IPAQ scores (7.91 ± 0.93 versus 7.59 ± 0.99) and lower disease activity (BASDAI = 3.59 ± 2.1 versus 4.18 ± 2.0).
Tool evaluation
Participant experience of using the AS Observer tool was generally positive, with the large majority (66 %) expressing intent to continue using it, either for the purposes of self- monitoring tool, active disease management (e.g. identifying triggers), to raise awareness of their own condition or as a motivator to undertake physical activity.
Withdrawal and completion rates
Only three participants chose to withdraw from the study before completion and these were all due to technical issues (including only having access to a phone, not a computer and failure of the site to load on particular browsers). The large majority of participants did not complete data entry every week with an average completion rate of 67%, i.e. 8 of 12 weeks (42% completed ≤ half, 55% completed 7–9 of 12 data entries, and only 3% completed 11 or 12). This fits with previous eHealth studies of web-based monitoring, with adherence rates ranging between 30% and 75%.31 All participants who answered exit questions appreciated the use of reminder emails; however, technical problems prevented some reminder emails being delivered in the last few weeks (reported by 20% of participants) and likely contributed to a drop in compliance. Not receiving reminder emails (including delivery to spam folders) was the main reason given by participants for skipping reporting (44%, Table 6) and this illustrates how important the inclusion of the reminder or prompt mechanism is for effective data collection in eHealth tools.
Participant experience
Exit questionnaires obtained from those withdrawing early (n = 1) and participants at the end of study (n = 92; 41%) were tabulated and coded for each question. Aside from participants who declined to answer (n = 33), the majority intended to continuing using the tool (yes/maybe = 66%). Participants generally found the historical score function useful, enabling monitoring of their condition overtime and helping with selfmanagement approaches.
Usability
Analysis of participant views on ease of tool use, likes and dislikes resulted in the generation of eight codes (Table 7). Participants liked the ability to see plots of their BASDAI scores and would have liked more information graphics (e.g. physical activity and averages for other users) to assist in monitoring their condition. There were split opinions among the participants regarding the tool layout, reflecting personal preference. A number of participants found recording physical activity difficult (Table 7). The use of an activity list as a memory prompt or a smartphone version allowing activity recording to occur at the time was proposed by participants. Additionally, participants would have liked a mechanism to record other influencing factors such as medication changes or periodic medicine use, weather, diet, reasons for no/low activity (e.g. injury/flare up), non-AS ill health, AS associated conditions (e.g. iritis/Crohns), alternative therapies (e.g. massage) and holidays.
Some participants reported technical problems with the website (e.g. compatibility issues and screen freezing during data input), with 28 enquiries to the technical support team. It was possible to resolve 11 of these enquiries, which predominantly related to problems caused when using certain browsers. Further specific information was needed and not obtained for seven enquiries, thus preventing resolution. Further seven enquiries could not be resolved, despite the use of browser mimic software, and due to the lack of an email address, it was not possible to follow up on three enquiries.
Comments received
Comments received via the project website fell into six main themes: statements of support (n = 17), eligibility enquiries (n = 8), comments regarding periodic unavailability (n = 3), technical problems (n = 4) and participation queries regarding confidentiality and time to complete (n = 2). Several comments were also received regarding the module content, which reflected the complex nature of AS and feedback received in the exit questionnaire, namely regarding recording of additional information, e.g. medication and diet, or that data recording did not reflect the complexity of AS (e.g. no functional measurements)or the detailed nuances of exercise (n = 3).
In this pilot study, participants were not asked to modify their physical activity levels; however, participants noted that tool use (including historical scores) helped them to engage with the role of physical activity in their condition. However, there was no evidence of an overall increase of physical activity during the study duration. Therefore, tool use may have enabled participants to engage with their activity and condition, but did not necessarily serve as an effective motivator for the majority.
Strengths and weaknesses
This study was successful in collecting research data of 223 participants, who were representative of AS populations, using an eHealth tool. Whilst not all individuals responded weekly over the 12-week-period sensitivity, analyses including only participants that had completed at least 6 weeks revealed generally similar findings (Table S3).
However, user-orientated platforms present a number of challenges to data collection. The challenges include selfassessment, which is subjective by nature and can therefore lack precision and allows for a range of interpretation and therefore no standardised approach to comprehending or responding to questions or inputting data,32 even when standardised collection tools (e.g. IPAQ) are utilised. Furthermore, there is always a risk of recruitment bias. In this pilot study, NASS was instrumental in recruitment and although other recruitment approaches were used, it is likely that this cohort was dominated by NASS members, who are likely to be of middle to high socioeconomic status (who are more likely to engage in e-health/technology-based research33), consist of potentially more severe AS cases and be particularly engaged with their condition. In this pilot study, we were unable to account for potential biases due to the socioeconomic status of the participants; however, socioeconomic status should not influence the observed associations between AS severity and exercise observed here, which utilised FE models to eliminate omitted variable bias. It should also be noted that due to study duration, it was not possible to account for seasonal impacts on physical activity or symptoms and that due to the nature of the data, the direction of associations, i.e. cause and effect, could not be established.
