To do or not to do the other hip?

For non-ambulant children with bilateral cerebral palsy who have unilateral hip instability, should the other (stable) hip be routinely reconstructed along with the unstable side, or left alone to be addressed at some future time if it becomes unstable or symptomatic? That is the question. The crux of this debate hinges on the uncertainty about the probability of the stable hip becoming unstable in the future, and the comparative effectiveness of the prophylactic operation with that of a later reconstruction of the second hip if it becomes necessary. Effectiveness ought to be defined in terms of outcomes meaningful to the children in question and their carers, and/or costs associated with these strategies. The question is not readily answerable by a randomized trial because the fact that the outcomes of interest occur a long time (several years) later. Park et al.1 report the results of a decision analysis approach instead, which suggests that routine concurrent prophylactic reconstruction of the contralateral hip is the preferred strategy. Decision analysis is a powerful tool to compare the risk-benefit tradeoffs associated with advantages and countervailing disadvantages of competing strategies.2 This approach requires the construction of a model or decision tree that comprises two (or more) branches representing the competing strategies that are being compared. Each strategy is associated with a set of plausible consequences, both desirable and undesirable, the probabilities of which are derived from the literature or conjecture (sometimes called ‘expert opinon’). Undesirable effects (complications) in turn may require further intervention which will lead ultimately to some good or bad terminal outcome (Fig. 1). Each of these outcomes or health states can be quantitatively assigned a value, typically in the form of a utility or a cost. The utility of any given health state or end outcome, can vary from 0 (equivalent to death) to 1 (equivalent to perfect health). The expected utility of a particular strategy is computed as a product of the utilities associated with each of the terminal outcomes and the probability of having such an outcome. The branch with the higher expected utility becomes the preferred strategy.3–5 A medieval decision analysis? (Inspired by Dr Donald A Redelmeier). Fundamental to the credibility of any decision analysis is how well the decision tree represents and accounts for all possible important consequences that might ensue from the decision, as well as assigning sensible estimates of the likelihood (probabilities) of these events.6 In this instance, Park et al. constructed a simple decision tree of two strategies: (1) concurrent prophylactic reconstructive surgery for all; or (2) observation of the contralateral stable hip in all cases. They derived the probabilities of the various events from largely retrospective observational studies or through expert opinion. Fortunately, one of the strengths of a decision analysis is that it can be repeated by varying the probability of one (or more) variables in the model over a range of their plausible values. Such sensitivity analyses demonstrate the robustness of the model (one strategy remains the preferred one over a wide range of plausible values) and generate confidence in the recommendation derived from the model, or provide some threshold probability beyond which the alternative strategy becomes the favoured one. The authors conducted sensitivity analyses for two variables: the rate of future instability of the stable hip, and the rate of recurrent instability following prophylactic reconstruction of a stable hip. Concurrent prophylactic reconstruction of the stable hip was determined to be the preferred strategy, as long as the rate of future instability of the untreated stable hip exceeds 27% and the risk of recurrent instability of the reconstructed hip does not exceed 29%. The model is sensitive to the probability of instability of the untreated contralateral hip which has been reported to be between 4% and 75%. The wide variation, probably attributable to the heterogeneity of the population, with some children at higher risk than others, raises the question: should all children be treated the same? An alternative model could have compared the ‘treat all’, ‘observe all’ strategies to a third selective strategy of ‘treat some’, based on the presence (or absence) of certain risk factors that are known to be associated with hip instability (e.g. age <8y with many more years at risk for dislocating; or those that have an adduction contracture; or are more severely involved – Gross Motor Function Classification System Level V). One other weakness of this decision analysis is the validity of the authors’ arbitrary assignment of utility values to the terminal outcomes,7 which were confined to the single crude dimension of pain severity (‘mild’, ‘moderate’, and ‘severe’) without consideration of other outcomes such as ease of caregiving associated with hip contractures or costs. Utility values associated with these outcomes ought to be derived from patients or, in this case, from parents/carers of children with severe cerebral palsy. Although the authors do acknowledge these limitations, they failed to conduct a sensitivity analysis of these utility values which would have gone a long way to mitigate this problem, if the model was shown to have been stable over a wide range of plausible utility values. Notwithstanding these limitations, this study is an excellent example of how decision analysis can be used to address questions unlikely to be answered by trials. While this may not be the last word on this question, the authors have paved the way to repeat the decision analysis when more reliable estimates of the probabilities of outcomes associated with specific risk factors from population-based or other cohort studies become available, and with more valid outcomes and utility values in their model.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.