This individual patient data meta-analysis evaluated the independent value of the Ankle Brachial Index (ABI) to predict cardiovascular events. The authors concluded that use of ABI may improve the accuracy of cardiovascular risk prediction beyond the Framingham Risk Score. Recommendations for practice are offered. The authors conclusions reflect the evidence presented and appear to be derived reliably.

To evaluate the independent value of the Ankle Brachial Index (ABI) to predict cardiovascular events.

MEDLINE (1950 to 2008) and EMBASE (1980 to 2008) were searched to identify relevant studies for the review. The reference lists of articles, plus conference proceedings, were searched for additional papers. Search terms were reported. Further studies (including unpublished data) were sought from collaborators, cardiovascular epidemiologists and vascular physicians, and through correspondence with the Asia Pacific Cohort Studies Collaboration.

Population-based cohort studies with participants drawn from a general population were eligible for inclusion in the review. ABI was measured at baseline using a standardised technique. Participants were followed up to assess total and cardiovascular mortality. The meta-analysis is based on men and women who had no previous history of coronary heart disease (CHD). These were predominantly white, and were located in Australia, Belgium, Italy, Netherlands, Sweden, United Kingdom and United States. The mean age range was 47 to 78 years. Studies were selected by two collaborators; disagreements were resolved by discussion.

The authors stated that study data were agreed upon, but it is not clear to what extent data checking and validation took place.

Individual participant data (IPD) were obtained from the principal authors and lead investigators for 16 of 20 eligible studies. IPD included baseline ABI measurements and details of nonfatal and fatal events during the follow up period. A Framingham Risk Score (FRS) was calculated for each participant. Missing values were imputed for 3.9% of total values.

This was an IPD meta-analysis of population-based cohort studies. IPD collected were used to calculate Hazard Ratios (HR), risk estimates and 95% confidence intervals (CI). Kaplan-Meier estimates and standard errors for 10-year total mortality, cardiovascular mortality and major coronary events were extracted for each study and stratified for gender, FRS and ABI. HRs for 10 categories of ABI level compared to a reference category were calculated separately for men and women for the outcomes of total mortality (TM), cardiovascular mortality (CM) and major coronary events (MCE). It is unclear whether these were stratified by study. HRs for the same outcomes were calculated for men and women, stratified by study and both adjusted and unadjusted for FRS. Study HRs were combined using a random-effects model. Ten year outcome rates stratified by sex, FRS and ABI categories were calculated for each study and combined using a random-effects model. Area under receiver operating curves (ROC) was used to compare the predictive ability of ABI in addition to FRS with that of FRS alone. Heterogeneity was assessed using Χ² and Ι² tests.

The investigators of 16 (of 20 eligible) studies provided data for the analysis (24,955 men and 23,339 women). The median duration of follow up was between 3 and 16.7 years; more than half of the studies were followed up for more than 10 years. Baseline data for ABI and the 10-year incidence of CHD predicted by the FRS are given in the paper. There was considerable variation in annual rates of deaths between the studies.

During 480,325 person-years of follow up the HR for death (ABI compared to reference standard level) when plotted by ABI category showed a reverse J-shaped distribution. The HR or risk of death increased with each of the six decreasing categories of ABI below the standard level (1.11 to 1.20) and was also increased for the category >1.4. These results were observed for both men and women, and similar patterns of results were shown for cardiovascular mortality and major coronary events. (The number of participants varied considerably between categories and in some categories the numbers of events were few). The HRs for TM, CM and MCE showed an increased risk of death in the generally accepted "at risk" range (<0.90) compared to "normal" (1.10 to 1.40) ABI for both men and women. HRs remained high after adjusting for FRS in men and women. A low ABI (≤90) almost doubled the risk of 10-year total mortality, cardiovascular mortality and major coronary events compared with overall rates in all FRS categories.

The inclusion of the ABI in cardiovascular risk stratification alongside the FRS would mean that reclassification of the risk category and modified treatment recommendations would be necessary in approximately 19 per cent of men and 36 per cent of women.

Use of the ABI may improve the accuracy of cardiovascular risk prediction beyond the FRS.

The review question was clear and supported by detailed inclusion criteria. The search strategy was comprehensive and included steps to retrieve unpublished data. The review process included attempts to minimise further errors and bias in terms of study selection, and in the reported process for the collection of individual patient data. The chosen method of statistical pooling was appropriate. Although the included participants were drawn from a general population, no further detail was given on their characteristics and this may limit generalisability. The authors conclusions reflect the evidence presented and appear to be derived reliably.

Practice: The authors stated that the currently accepted high risk cut point of ABI ≤ 90 appears reasonable, but values from 0.91 to 1.10 and those exceeding 1.40 may also pose elevated cardiovascular risks beyond normal levels.

Research: The authors stated that a new risk equation combining ABI values with relevant Framingham risk variables should be validated. A cost-effectiveness analysis is also warranted to explore the effect of using ABI on clinical outcomes in the long term.

An educational grant for data processing and initial statistical analysis was provided by Sanofi-Aventis/BMS.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.