It is clearly feasible to do certain types of randomised placebo controlled trials over the internet. We were able to solicit and authenticate applicants and to conduct the internet based trial with adherence and retention rates comparable to those of trials done in traditional settings. Furthermore, our participants reported high levels of satisfaction and willingness to participate in similar trials in the future.
The trial was efficient with respect to the application rates, direct data entry, short time to database lock, and minimal staffing levels. On the other hand, the time taken to obtain consent and medical records limited the speed of enrolment. These were two aspects of our method that were not internet based. Although the large number of applicants offset the delay, it could be improved by measures to circumvent the need for paper documents and medical records. We were also not able to tackle the issue of scalability in this single trial.
The decrement in numbers at each step of the application process may have reduced the external validity of our sample (). Although our participants seemed broadly similar to those in comparable traditional trials, the requirement for internet access probably resulted in a sample of higher educational level. External validity is also a common problem for traditional trials, owing to restrictive eligibility criteria and selection biases in hospital based settings, whereas the internet based approach allows participation from the home or workplace. In either case, the problem needs to be tackled by an adequate description of the setting and sample characteristics.15
In principle, considerations about generalisability should not affect the internal validity inherent in a well performed randomised controlled clinical trial. We used epidemiological methods for case confirmation and a self report assessment of disease severity validated for computer use.7,9
However, we relied on self reported pill counts that may be less reliable than those done by a research nurse. Future internet based trials could assess this aspect more robustly through the use of returned blister packs or other technologies. The fact that the WOMAC pain scores, and their variability, were similar to those in traditional trials provides further evidence for the construct validity of this method. Also, we detected regression to the mean by using this measure and obtained a (negative) result concordant with recent independent trials of this compound.16–18