To Operate or Not to Operate? That Is (Still) the Question

COMMENTARY

It need hardly be said that the decision whether to operate is the most critical step in treating a patient with a torn anterior cruciate ligament (ACL). So why are there only a handful of studies to support an evidence-based approach to this important decision? One reason may be that such studies are so difficult to do well. Randomization is expensive, and it is a considerable undertaking to recruit and retain a representative sample of subjects about whom valid, generalizable conclusions might be drawn. However, without randomization, treatment allocations and outcomes can be confounded by so many sources of bias, on the part of investigators, treating physicians, and subjects alike, that it can be very difficult to sort out fact from fallacy, or valid inference from statistical artifact.

Grindem et al. assessed several clinically relevant outcomes in 143 patients who were identified and recruited within three months after an ACL injury. The outcomes that were studied included patient-reported knee function as measured with the International Knee Documentation Committee (IKDC) 2000 score, isokinetic strength compared with that of the uninjured knee, level of sports participation, and knee reinjuries. All patients underwent an initial period of rehabilitation and education, and this was followed by self-assignment into the operative arm or the nonoperative arm. Strengths of the study design include prospective enrollment, documentation of patients’ motivation for choosing one arm or the other, and use of monthly surveys to document sports participation level, knee reinjuries, and compliance with recommended activity restrictions. These monthly follow-up surveys are particularly interesting in a young and active population that is considered to be at high risk for reinjury and for whom final decisions about success or failure usually can be made within, rather than after, two years following presentation. Interestingly, the monthly surveys demonstrated a high rate of noncompliance with activity restrictions in both treatment groups.

The authors should be commended for their efforts to limit the potential confounding influences of independent variables on final outcomes due to lack of randomization. The propensity score is a statistical tool that has been promoted to adjust for bias in nonrandomized treatment allocations. I confess that I have almost no experience with that tool, and I suspect that this is true for most readers of The Journal. In theory it sounds like an interesting technique, but I have nagging doubts about its use in evaluating treatment of ACL deficiency. For example, activity level is a risk factor but also an outcome; it is both an independent and a dependent variable. What troubles me most about this study is that the authors appear to have been distracted by their focus on a complicated and little used statistical tool to control nonrandom selection and thus failed to use the more standard technique of intention-to-treat analysis to account for crossovers from the nonoperative arm. It is an open question whether the propensity-score adjustment made much difference, as in the end we readers need to interpret the data for ourselves in order to draw what insights we can from the work as it is presented.

One thing that seems apparent is, in terms of subjective knee function, the forty-three patients who completed the nonoperative arm of the study compared well with those who underwent an operation. This finding is consistent with those of prior studies that demonstrated the potential for good function following nonoperative treatment1–6. It is worth noting that nonoperative treatment is not the same as neglect. The entire study cohort was treated with a rehabilitation program that has been shown to be effective for the selection of candidates to undergo a trial of nonoperative care7,8. This is not meant as a criticism of the study, but it does suggest that the nonoperative group was carefully screened. In addition, the rate of concomitant surgery (for meniscal tears) was 32% in the operative group whereas no surgical procedures related to the index injury were done in the nonoperative group, suggesting that patients with additional injuries were more likely to select surgery at the outset.

Of the sixty-four patients who initially chose nonoperative treatment, twenty-one later crossed over to the operative group at an average of 12.7 months into the study. Postoperative outcomes for these patients were included with those for the operative group, but the authors did not provide any data on preoperative knee function or level of sports participation for this subsample that crossed over. Comparisons between the patients who underwent early and those who underwent late reconstruction are detailed in an appendix, where it appears that four of the twenty-one sustained a knee injury prior to the reconstruction. Had these injuries been included in the reinjury data for the nonsurgically treated knees, reinjury rates for the knees that had not (yet) undergone an operation would have been higher.

It is unfortunate that the study was designed and powered only to compare outcomes between the two main groups, as comparisons between the late-reconstruction and early-reconstruction groups may consequently lend themselves to a potentially false inference that no harm was caused by the delay in reconstruction. That is a central concern of clinicians managing these patients, and this study sheds little light on it. In fact, it is impossible to draw any conclusions at all about this important question. All outcomes for the late-reconstruction group are postoperative data, and we have no information on whether they functioned at a lower level preoperatively than either the early-reconstruction group or the patients in the nonoperative group who completed the study without crossing over.

Despite these shortfalls, the study by Grindem et al. highlights the potential for good outcomes with nonoperative care following ACL injury. Roughly a third of the entire study population experienced good function and low reinjury rates. These good outcomes were due to careful screening, followed by a tailored rehabilitation program and close follow-up monitoring for at least two years. The price paid, in terms of patients who crossed over presumably because of reinjury, unacceptable function, or limitations imposed by the treatment protocol, is difficult to glean within the limits of the study design, but it merits further study. In the end, we serve patients best if we can offer at least some of them a real alternative to reconstruction. However, there remains the challenge of returning patients to their desired level of activity while avoiding a substantial reinjury resulting from an unreconstructed ACL.