A Pragmatic Perspective on Evaluating Ankle Arthritis Treatments: The Challenges of Loss to Follow-up and Its Impact on Study Validity

Commentary Glazebrook and colleagues, a group that has been at the forefront of comparing treatments for ankle osteoarthritis (OA), compared total ankle replacement (TAR) and ankle arthrodesis (AA) with respect to postoperative clinical and patient-reported outcomes. At a minimum 10-year follow-up, 211 subjects (157 TARs; 54 AAs) were included from 4 centers. Four implant types were included, 3 of which were mobile-bearing designs (2 of which were never available in the U.S.). Their data did not demonstrate a difference in pain and disability outcomes between the groups, but there was a higher risk of reoperation in the TAR group. These data are consistent with the group’s prior 5.5-year results published in 20141. TAR has been utilized in the U.S. since the 1970s. Utilization dropped dramatically in the 1980s and 1990s when reports demonstrated higher revision rates and less success than hip or knee arthroplasty2. Unlike hip and knee OA, ankle OA can be effectively treated with arthrodesis. The ankle joint is anatomically and mechanically very different from the hip and knee joints: the foot is at a 90° angle to the tibia, sagittal plane motion is dramatically less, and the size of the implants is limited. These factors, combined with the relatively limited understanding of the mechanics and deformity of the hindfoot during that early era and the relative infrequency of TAR, led to suboptimal results. A dramatic increase in the utilization of TAR has occurred in the last 2 decades, despite the absence of high-quality evidence showing that TAR is superior to AA. So why are patients and surgeons choosing TAR? Other prospective studies have reported that, compared with AA, TAR is associated with better self-reported outcomes at 2, 3, and 4 years3; more rapid improvement4; and equivalent reoperation rates5. A recent randomized controlled trial (RCT) that included 281 subjects demonstrated that those who underwent TAR had numerically greater improvement in their Manchester-Oxford Foot Questionnaire scores at 52 weeks postoperatively compared with those who underwent AA; however, the difference between the groups was not significant6. Additionally, a meta-analysis cited by Glazebrook et al. that aggregated data from 21 studies, including 18,000 patients, showed equivalent risks of reoperation between the TAR and AA groups as well as greater patient-reported improvement in the TAR group7. The study by Glazebrook et al. provides useful information for patients and providers who want a realistic assessment of the likelihood of reoperation from an unbiased source. The investigators should be congratulated for their strong efforts to report the long-term follow-up comparisons of these 2 treatments. However, the study is limited by several pragmatic methodological challenges. The most important is the considerable loss to follow-up in the analysis cohort. The study lacks a STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) diagram, which is often required in the reporting of observational studies to allow the reader to have a more transparent understanding of the losses to follow-up in order to determine whether there may be potential bias as a result. The authors note that the original cohort included 381 participants. The follow-up analysis included 211 of these participants (55%). High loss to follow-up is not uncommon in surgical studies. Maintaining long-term contact with patients is a significant challenge; however, loss to follow-up is still a major threat to the validity of the results and should be handled with care. Methods to account for such losses may include sensitivity analyses that involve multiple imputation techniques using prior available data for patients who are missing from later follow-ups. In lieu of such statistical techniques, since the authors had such a rich dataset, a reasonable approach could have been to compare the earlier outcome scores, where data were available, between patients who did and did not reach their 10-year follow-up in order to determine if there were differences between the 2 treatment groups in the full cohort. Based on the reported data, it is difficult to know if the patients who were not accounted for may have had better or worse outcomes. In addition, the Ankle Osteoarthritis Scale (AOS) was the predominant outcome measure at the time this registry was initiated. It is a derivative of the Foot Function Index (FFI), which was validated in a small group of subjects with systematic inflammatory arthritis. Both the FFI and AOS have a large ceiling effect and a reported high minimal clinically important difference. The group under study was an active cohort of patients, and therefore the AOS may not have had adequate sensitivity to detect important comparative functional differences. In addition, the authors state that “The indications for TAR were not the same as those for AA.” The AA group had a mean age >7 years younger than the TAR group, and inflammatory arthritis was >8 times more prevalent in the TAR group. It is difficult to determine the degree to which inflammatory arthritis had an ongoing impact on the outcome measures or pain. A strong take-home message should be that we do not have enough data to help guide a treatment algorithm. RCTs are very expensive, and participants in the U.S. are reluctant to accept randomization. Accounting for long-term outcomes is a significant challenge. Our best hope is that the American Joint Replacement Registry (AJRR) and American Academy of Orthopaedic Surgeons Registry Program begin to include data from TARs so that sufficient data can be generated to guide care.