Prognostic value of serum troponin levels in patients with pulmonary embolism: a systematic review and meta-analysis

Objective

Patients with pulmonary embolism (PE) often show elevated plasmatic troponin levels; this condition may be associated with increased mortality and other adverse outcomes. To clarify the possible association between serum troponin levels and clinical outcome in patients with PE we performed a systematic review of the literature.

Data source

The search strategy included electronic search of original studies in Medline, Embase, Cochrane Library, Pascal, Medion and ISI Proceedings, manual retrieval of recovered bibliographic references and personal communication with authors.

Study selection

We selected inceptive cohort studies relating baseline troponin levels and the occurrence of hospital mortality, unfavourable clinical course or echocardiographic signs of right ventricular dysfunction. The reviewers screened the articles independently for eligibility and resolved disagreements by consensus.

Extraction, quality assessment and results

Data about odds ratio (OR), sensitivity, specificity, likelihood ratios (LR) with confidence interval (95% CI), study population, type of test assay used (troponin T or troponin I), and methodological quality (using a modified score after the Lijmer and Altman checklist) were obtained independently in a duplicate form and disagreements resolved by consensus. Concordance between reviewers was assessed using the weighted kappa statistic. We explored methodological, statistical and clinical heterogeneity, and built a meta-regression model to explore threshold effects, constructed the Summary Receiver Operating Characteristic (SROC) curves and quantified the impact of study's characteristics on the magnitude of effects, using StatsDirect (v 2.3.7), SPSS (v 11.0.1) and Meta-Disc (v 1.1.0) software.

Results

We identified 13 studies, which included 1293 patients. Agreement for quality assessment previous to consensus for each pair of reviewers were 0.51 (95% CI 0.30, 0.72) and 0.74 (95% CI 0.58, 0.90), respectively. The quality of recovered studies was not optimal. The studies were relatively homogeneous for each outcome, and there was no evidence of threshold effect or publication bias. Overall weighted mortality was 14%. The presence of high troponin levels identified a subgroup of patients at high risk of: dying during the hospitalization (OR 6.20 [95% CI 4.04, 9.52]; area under SROC curve [AUC] 0.78 [CI 95% 0.73, 0.83]), showing echocardiographic signs of right ventricular dysfunction (OR 8.04 [95% CI 4.40, 16.69]; AUC 0.80 [95% CI 0,77, 0,82]), and suffering an unfavourable clinical course (OR 14.58 [95% CI 6.82, 31.16]; AUC 0.84 [95% CI 0.75, 0.94]). Normal troponin levels identified patients with favourable clinical course (LR 0.23 [95% CI 0.13, 0.41]).

Conclusions

Troponin levels are useful to stratify the risk in patients with PE. In the individual patient, normal troponin levels indicate a high probability of following a favourable clinical course.