To the authors’ knowledge, this decision-tree model based on the IDSA guideline is the first to be created for Canada. However, its limitations and assumptions may limit its clinical utility. Overall, the decision model presented here may represent an oversimplification of the complexity of treatment of diabetic foot infections, and a model based on guidelines may differ from actual clinical practice.
Although clindamycin was the cost-effective agent in this academic model, the clinical success rates for this drug were taken from a small study (total n = 56). In addition, that study showed no significant difference in primary success rates between the 2 drugs studied (clindamycin and cephalexin). It must be remembered that the decision-tree model includes not only the primary success rates, but also the effects of subsequent drugs. Monte Carlo simulations confirmed clindamycin’s position, but there is still a strong possibility that the 2 drugs are very similar clinically (i.e., in practice) and that the cost- effectiveness result is an artifact. Unfortunately, the quality of the evidence for the oral antibiotics recommended by the IDSA guidelines is weak because of small sample sizes and because patients with diabetic foot infections were only a subgroup in some studies. In addition, some of the studies were done in the early 1990s, when the choice of antibiotics and resistance rates would have differed from current standards of practice. Larger and better-designed RCTs for the antibiotics recommended in the IDSA guidelines are needed to bridge these gaps and also to help clinicians to choose the most cost-effective agents for the treatment of mild diabetic foot infections.
Other limitations of this decision model included obtaining the clinical success rates from evaluable patients rather than from the intention-to-treat group. Also, clinical success (i.e., efficacy) rates were used, rather than cure rates, which might have overestimated the effectiveness rates of the antibiotics included in the model. Most of the clinical success rates used in this model were obtained from single RCTs of the antibiotic. In some cases, patients with diabetes constituted only a subset of patients in the RCT. Many of the trials were underpowered, and the clinical success rates reported could be overestimations of the true clinical success rates of the antibiotics included in the model. Only one large RCT involving diabetic patients was identified in the literature search, a comparison of ertapenem and piperacillin-tazobactam.
Low resistance rates were assumed, but current epidemiology of diabetic foot infections indicates that resistance rates are increasing, especially for community-acquired methicillin- resistant Staphylococcus aureus (MRSA). If MRSA is suspected, antibiotic regimens would have to include agents that are active against MRSA, such as linezolid or vancomcyin or, if localresistance patterns suggest susceptibility of community- acquired MRSA, rifampin, doxycycline, clindamycin, or sulfamethoxazole—trimethoprim.
Another concern is the serious adverse effects associated with clindamycin (e.g., pseudomembranous colitis), which could limit the use of this antibiotic; other antibiotics with more benign adverse effect profiles are available. Also, the cost of treating these adverse effects could offset the cost benefits of clindamycin in the decision model.
Finally, the costs presented here may represent an underestimation or an overestimation because overhead costs of hospital care were not factored in and because it was assumed that all patients treated in hospital received the full 14-day course of parenteral antibiotics. For patients whose condition is stable, home parenteral antibiotic therapy is a possibility; if this is considered, costs would be lower than projected with the current model. An ideal model should factor in the percentage of patients who would complete the course of therapy with parenteral or oral antibiotics on an outpatient basis, and this should be a consideration in future models.
Because the model was intended to examine treatment of mild infections only, and because moderate to severe infections in the current model represented only a subset analysis if oral antibiotic therapy failed, future work should include designing a model for moderate to severe diabetic foot infections in Canada. Other enhancements, including obtaining intention- to-treat values, would provide a better estimate of the true clinical success rates of the antibiotics included in this model.
Despite the limitations discussed above, the model did reveal that treatment costs associated with diabetic foot infections increase with treatment failure, as well as with increased rates of amputation and death. This finding concurs with the current literature and with recommendations for prompt and optimal management to reduce the incidence of infection- related morbidity and mortality and its associated costs.
In the decision-tree model developed in this study, clindamycin dominated other oral antibiotics recommended by the IDSA guidelines for treating mild diabetic foot infections, but this observation should be interpreted with caution. The evidence used to build the model was based on a small number of studies with small sample sizes. Therefore, more clinical studies evaluating oral antibiotics for treating mild diabetic foot infections are needed. Also, given the other limitations and assumptions of the model, the results should not be applied in isolation; rather, they should be combined with clinical experience and current standards of practice.