Twenty five studies were included. Of those, 13 reported prognostic accuracy data. All except one study were prospective and retrospective cohort studies with moderate to good reporting. Reporting quality varied from 11 to 18 out of 22(STROBE) and from 12 to 15 out of 25 (STARD).
Prevalence of lower extremity amputation ranged from 6% to 78%. Sensitivity of the classification systems ranged from 37.6% with the Curative Health Services wound grade scale (CHS), to 100% using the Depth of the Ulcer, Extent of bacterial colonization, Phase of ulcer and Association aetiology classification system (DEPA) score (1 study). Specificity ranged from 31% with the Scottish Intercollegiate Guidelines Network (SIGN) classification system, to 88% using the Size (Area, Depth), Sepsis, Arteriopathy, Denervation system (S(AD)SAD) score. Three studies reported major lower extremity amputation rates separately. Of those, Lipsky scores had higher sensitivity and specificity, followed by Meggit-Wagner and SIGN. .
Meta-analysis was only possible for the accuracy of individual prognostic variables. Pooled analyses showed that diabetic peripheral neuropathy had the highest sensitivity (88%) and the lowest specificity (30%). On the other hand, gangrene had the lowest sensitivity (11%) and the highest specificity (95%). Positive likelihood ratios ranged from 1.22 (for diabetic peripheral neuropathy) to 5.50 (gangrene), negative likelihood ratios from 0.38 (diabetic peripheral neuropathy) to 0.91 (gangrene). Heterogeneity was generally high (Ι²>90%). Further results from the meta-analyses were reported.
Classification system reliability was evaluated in one study, and only four classification systems were validated by three or more studies. Further data on reliability and external validity were reported.