Twenty four studies met the inclusion criteria (n=10,710). Eleven studies had all the important predictors present in a significant proportion of the participants and 11 reported blinding of the interpreters of tests; only one was considered to have recruited an adequate sample size.
None of the items that related to clinical history could be relied upon to rule in or rule out LVSD: myocardial infarction (six studies, n=1,946); diabetes (two studies, n=717); hypertension (two studies, n=717); and male gender (two studies, n=1,471).
None of the items that related to symptoms could be relied upon to rule in or rule out LVSD: fatigue (two studies, n=1,079); dyspnoea (three studies, n=1,338); orthopnoea (three studies, n=1,338); and paroxysmal nocturnal dyspnoea (three studies, n=1,338).
Clinical signs:
Pooled positive likelihood ratio estimates for raised jugular venous pulsation (4.36, range 2.66 to 7.44; three studies), displaced apex beat (15.96, 95% CI 8.24 to 30.93; two studies) and a third heart sound (7.43, range 1.56 to 32.37; three studies) showed better diagnostic performance than other measures, but the numbers of studies in the analyses were small.
Pooled negative likelihood ratios were 0.88 (95% CI 0.83 to 0.91) for raised jugular venous pulsation, 0.58 (range 0.35 to 0.93) for displaced apex beat and 0.92 (range 0.77 to 0.96) for a third heart sound.
ECG was the most commonly evaluated diagnostic test (11 studies, n=3,570). The pooled positive likelihood ratio was 2.13 (95% CI 1.95 to 2.33) and negative likelihood ratio was 0.27 (range 0.06 to 0.76).
Brain natriuretic peptides (BNP) (nine studies) produced a pooled positive likelihood ratio of 1.90 (range 1.26 to 6.20) and negative likelihood ratio of 0.30 (range 0.02 to 0.80) when the cut-off nearest to 15pmol/L was used; there was significant heterogeneity between studies.
For clinical history, positive likelihood ratio ranged from 0.58 to 2.86 and negative likelihood ratio ranged from 0.68 to 1.30.
For symptoms, positive likelihood ratio ranged from 1.03 to 1.71 and negative likelihood ratio ranged from 0.50 to 0.98.
Most results were available in supplementary files available online (see URL For Additional Data; URLs accessed 31 July 2009).