ELISA versus PCR for diagnosis of chronic Chagas disease: systematic review and meta-analysis

Abstract

Background: Most current guidelines recommend two serological tests to diagnose chronic Chagas disease. When serological tests are persistently inconclusive, some guidelines recommend molecular tests. The aim of this investigation was to review chronic Chagas disease diagnosis literature and to summarize results of ELISA and PCR performance.

Methods: A systematic review was conducted searching remote databases (MEDLINE, LILACS, EMBASE, SCOPUS and ISIWeb) and full texts bibliography for relevant abstracts. In addition, manufacturers of commercial tests were contacted. Original investigations were eligible if they estimated sensitivity and specificity, or reliability -or if their calculation was possible - of ELISA or PCR tests, for chronic Chagas disease.

Results: Heterogeneity was high within each test (ELISA and PCR) and threshold effect was detected only in a particular subgroup. Reference standard blinding partially explained heterogeneity in ELISA studies, and pooled sensitivity and specificity were 97.7% [96.7%-98.5%] and 96.3% [94.6%-97.6%] respectively. Commercial ELISA with recombinant antigens studied in phase three investigations partially explained heterogeneity, and pooled sensitivity and specificity were 99.3% [97.9%-99.9%] and 97.5% [88.5%-99.5%] respectively. ELISA's reliability was seldom studied but was considered acceptable. PCR heterogeneity was not explained, but a threshold effect was detected in three groups created by using guanidine and boiling the sample before DNA extraction. PCR sensitivity is likely to be between 50% and 90%, while its specificity is close to 100%. PCR reliability was never studied.

Conclusions: Both conventional and recombinant based ELISA give useful information, however there are commercial tests without technical reports and therefore were not included in this review. Physicians need to have access to technical reports to understand if these serological tests are similar to those included in this review and therefore correctly order and interpret test results. Currently, PCR should not be used in clinical practice for chronic Chagas disease diagnosis and there is no PCR test commercially available for this purpose. Tests limitations and directions for future research are discussed.

Figure 1

Flowchart of abstracts and full texts evaluation and data analysis plan. Several full texts evaluated two or more tests; therefore the sum of tests with data extracted is not equal to the total of full text included.

Figure 2

Regular ELISA for Chagas disease summary ROC for tests with blinded reference standard evaluation. Dots - crude estimates; filled black squares - shrunken estimates from bivariate model; filled red circle - bivariate model summary estimate; black solid line - SROC curve.

Figure 3

Forest plots of sensitivity (A), specificity (B) and DOR (C) for phase 3 studies investigating commercial ELISA tests. Events - identified as with Chagas disease by the test; OR - (diagnostic) odds ratio; proportion - sensitivity (or specificity) individual point estimate; Total - number of subject within each sample (with or without Chagas disease); W - weights.

Figure 4

Forest plots of sensitivity (A), specificity (B) and DOR (C) for phase 3 studies investigating commercial ELISA with recombinant antigens. Events - identified as with Chagas disease by the test; OR - (diagnostic) odds ratio; Proportion - sensitivity (or specificity) individual point estimate; Total - number of subject within each sample (with or without Chagas disease); W - weights.

Figure 5

PCR for Chagas disease HSROC - subgroups of blood samples boiled before processed and stored with guanidine. Big blue upside down triangle - summary estimate of the group stored with guanidine and boiled; Big green diamond - summary estimate of the group stored with guanidine and not boiled; Big red triangle - summary estimate of the group not stored with guanidine and not boiled; Blue small squares and blue small upside down triangles - shrunken estimates from bivariate model and crude estimates (respectively) from group stored with guanidine and not boiled; blue solid line - SROC curve from group stored with guanidine and boiled; dot - crude estimates; small green squares and small green diamonds - shrunken estimates from bivariate model and crude estimates (respectively) from group stored with guanidine and not boiled; solid green line - SROC curve from group stored with guanidine and not boiled; small red squares and small red triangles - shrunken estimates from bivariate model and crude estimates (respectively) from group not stored with guanidine and not boiled; red solid line - SROC curve from group not stored with guanidine and not boiled; NA - not assigned or missing.