Results from this study indicate that Join2move is a plausible, feasible and acceptable program for patients with knee and/or hip OA. Although effectiveness was not proved due to the lack of power, results do indicate that Join2move has the potential to increase PA levels in patients with knee and/or hip OA. Participants reported higher levels of PA, particularly (and as expected) involving moderate activities like walking and cycling (200 minutes). In line with other research , walking was by far the most frequently selected activity. Our positive results correspond with a comparable face-to-face intervention, showing a moderate PA increase of 170 minutes . In the first three weeks, the increase was accompanied by more pain. Fortunately, after twelve weeks the pain scores declined towards baseline levels. Although the intervention focused on improving PA rather than on pain reduction, the increased pain was certainly a reason for concern. The precise cause of observed elevated pain scores is unclear. A possible explanation is the increased PA which may generate more muscle and joint pain. However, it is important to note that higher levels of pain are not associated with deterioration of OA [42, 43].
Providing an intervention does not automatically mean that patients will use it, particularly when it is self-directed, with minimum personal contact. Since the success of web-based interventions requires active participation, non-usage attrition has been pointed out as a common concern in the field of web-based education. In line with other studies, [21, 22, 44], the number of users gradually decreased during the nine-week program. Overall, 55% (n=11) of the participants completed at least 75% of the program. This exposure percentage corresponds with the study of Steele et al.  and can be rated as reasonably high for web-based interventions without human interference. The delivery of personal information on a weekly basis is a possible explanation for this relatively low non-usage attrition. In this respect, it was not possible for users to run the entire program at one time. Although we did not examine the specific strategies of engagement, the authors assume that the week-by-week basis provided an incentive to return to the website.
With respect to usability, the involvement of end-users was extremely valuable for identifying usability issues and system flaws. Along the way, we incorporated greater flexibility into the program. The implemented changes resulted in a less rigid version with more options tailored to the performance of the individual user.
The findings from this study need to be interpreted in light of the study's limitations. The small sample size, single group design and lack of long-term assessments limit conclusions of causality, long-term effects and generalizability. Furthermore, the potential presence of the so-called Hawthorne effect may have contributed to an overestimation of PA scores. This implies that observed PA changes may be partly the result of study participation. Besides the Hawthorne effect, self-reported PA measures may also contribute to an over-estimation of PA levels in this study. This may be a consequence of recall error, perceived social desirability and other biases. To obtain the best results, a combination of validated questionnaires with objective measures would be preferable in future studies. Another limitation concerns outside-care patients who lack computer skills or internet access. These groups are mostly excluded from web-based interventions. Unfortunately, this disadvantage applies also to the Join2move intervention. Typically, these patients are disproportionately less educated and have a lower income. Particularly with regard to these under-served populations, GPs should refer sedentary OA patients more frequently to a physical therapist or other health care provider. Further, it will be important to translate Join2move for other self-help platforms, such as videos, brochures and self-help books. A final limitation is that we only performed one Thinking Aloud test to detect and resolve usability issues. Unfortunately, we did not retest the redesigned intervention. In order to optimize usability for the implementation phase, a repetition of this procedure is advised.