bs_bs_banner
REVIEW
Diagnostic value of the soluble triggering receptor expressed on myeloid cells-1 in lower respiratory tract infections: A meta-analysis WU YE, YANJIE HU, RUIFENG ZHANG AND KEJING YING Department of Respiratory Diseases, Sir Run Run Shaw Hospital, Zhejiang University, Hangzhou, China
ABSTRACT The soluble triggering receptor expressed on myeloid cells-1 (sTREM-1) is a promising diagnostic marker for many types of infections. A bivariate meta-analysis was performed to evaluate its diagnostic value for lower respiratory tract infections (LRTI). We searched PubMed, Cochrane Library and Web of Science (from January 1966 to August 2013) for all trials assessing diagnostic value of sTREM-1 for LRTI. The pooled sensitivity, specificity, positive likelihood ratio(PLR), negative likelihood ratio(NLR), diagnostic odds ratio (DOR), the area under summary receiver operator characteristic (SROC) curve and the Q* were calculated. Thirteen studies with 1138 patients were included in our metaanalysis. The pooled sensitivity and specificity of sTREM-1 for diagnosis of LRTI was 0.84 and 0.77. The PLR, NLR and DOR were 3.6, 0.21 and 17. The area under SROC curve was 0.88 and the Q* was 0.82. The univariate meta-regression analysis demonstrated that the assay method for sTREM-1 significantly affected sensitivity for LRTI. The Q* of sTREM-1 for diagnosis of community-acquired LRTI was 0.82, and the area under SROC curve was 0.88. The Q* of sTREM-1 in diagnosis of hospital-acquired LRTI was 0.83, and the area under SROC curve was 0.90. The Q* of sTREM-1 for distinguishing culture-positive LRTI from culturenegative diseases was 0.79, and the area under SROC curve was 0.86. Current evidence suggests that sTREM-1 is an accurate marker of LRTI. The overall diagnostic value of sTREM-1 for LRTI, communityacquired LRTI and hospital-acquired LRTI is similar. Key words: diagnosis, diagnostic test, lower respiratory tract infection, meta-analysis, soluble triggering receptor expressed on myeloid cells-1.
Abbreviations: BALF, bronchoalveolar lavage fluid; CI, confidence interval; DOR, diagnostic odds ratio; ELISA, enzyme-linked immunosorbent assay; FN, false negative; FP, false positive;
Correspondence: Kejing Ying, Department of Respiratory Diseases, Sir Run Run Shaw Hospital, Zhejiang University, 3 East Qingchun Road, Hangzhou 310016, China. Email: yingkj55 @126.com Received 16 October 2013; invited to revise 25 November 2013; revised 30 November 2013; accepted 31 December 2013 (Associate Editor: Marcos Restrepo). © 2014 Asian Pacific Society of Respirology
ICU, intensive care unit; IL, interleukin; LRTI, lower respiratory tract infections; NLR, negative likelihood ratio; PLR, positive likelihood ratio; QUADAS, quality assessment for studies of diagnostic accuracy; SROC, summary receiver operator characteristic; STARD, standards for reporting of diagnostic accuracy; sTREM-1, soluble triggering receptor expressed on myeloid cells-1; TN, true negative; TP, true positive; TREM-1, triggering receptor expressed on myeloid cells-1; VAP, ventilator-associated pneumonia.
INTRODUCTION The triggering receptor expressed on myeloid cells-1 (TREM-1) is expressed on neutrophils and monocytes, and plays an important role in systemic infections. Upregulation of TREM-1 seems to be a crucial step during bacterial and fungal infections.1,2 TREM-1 is an important signalling receptor, magnifies the inflammatory reaction by elevating pro-inflammatory cytokines and blocking interleukin (IL)-10 synthesis. The upregulation of the TREM-1 leads to elevation of the soluble form of TREM-1 (sTREM-1) in body fluids, including serum, bronchoalveolar lavage fluid (BALF) and cerebrospinal fluid.2 Respiratory tract infections are still the most frequent reason for patients to seek medical service and remain the main source of sepsis.3 Identifying a true ‘gold standard’ for diagnosis of respiratory tract infections is often problematic. The use of blood and sputum cultures have limitations, including the duration of time required to obtain positive cultures, excluding contamination of specimens and identifying colonization.3 Massive research has been performed to discover new diagnostic and prognostic biomarkers. Some biomarkers, including C-reactive protein, procalcitonin, sTREM-1, IL-1b, plasminogen activation inhibitor-1, soluble urokinase-type plasminogen activator receptor and presepsin, appear promising.2,4 Recently, many studies show that sTREM-1 has usefulness for differential diagnosis of lower respiratory tract infections (LRTI).5–7 Therefore, we carried out a bivariate meta-analysis to estimate the diagnostic value of sTREM-1 for LRTI and explored whether its diagnostic values were different for community-acquired and hospitalacquired LRTI. Respirology (2014) 19, 501–507 doi: 10.1111/resp.12270
502
METHODS Search strategy We retrieved Pubmed, Cochrane Library and Web of Science (from January 1966 to August 2013) for all studies examining diagnostic value of sTREM-1 in LRTI. The following keywords or Medical Subject Headings were used: ‘sTREM-1’,‘TREM1 protein, human’, ‘soluble triggering receptor expressed on myeloid cells-1’, ‘respiratory tract infections’, ‘respiratory infection’ and ‘pneumonia’. The references of searched articles and relevant reviews were also reviewed. Study selection We included all studies that fulfilled the following criteria: (i) papers limited to human subjects; (ii) trials evaluating diagnostic accuracy of sTREM-1 for LRTI; (iii) studies provided enough data to construct the 2 × 2 contingency table and calculate true positive (TP), false positive (FP), false negative (FN) and true negative (TN); and (iv) sample source for sTREM-1 was serum or BALF. Data extraction Two reviewers (W.Y. and Y.H.) independently reviewed the trials, screened the citations, selected eligible studies and extracted the relevant information. If disagreement occurred, two reviewers discussed and resolved by consensus. The information extracted from included studies contained the following details: (i) family name of the first author, published year and country of the trial; (ii) category and prevalence of infections; (iii) the number and age of study population; (iv) sample source; (v) TP, FP, FN, TN; (vi) assay method for sTREM-1; and (vii) cut-off value. Quality assessment The methodological qualities of the trails were evaluated using the QUADAS (quality assessment for studies of diagnostic accuracy)8 and the STARD (standards for reporting of diagnostic accuracy).9 We attributed 1 score for each ‘yes’ (high quality), 0 score for each ‘no’ (low quality) and 0.5 score for each ‘unclear’. Statistical analysis We performed the analyses using the statistical software (Stata version 12.0, Stata Corporation, College Station, TX, USA). Bivariate random effects regression model was used to calculate the pooled specificity, sensitivity, diagnostic odds ratio (DOR), positive likelihood ratios (PLR) and negative likelihood ratios (NLR). The DOR of a test is the ratio of the odds of positivity in disease relative to the odds of positivity in the non-diseased: (sensitivity/(1 − sensitivity))/ ((1 − specificity)/specificity).10 The DOR is a single overall indicator of diagnostic accuracy.10 The summary receiver operator characteristic (SROC) curve was constructed with sensitivity on the vertical axis and 1–specificity on the horizontal axis. We also Respirology (2014) 19, 501–507
W Ye et al. 121 citations identified from Pubmed, Cochrane Library and Web of Science
20 relevant references listed in the searched papers and review
141 citations identified 116 citations excluded on the basis of title and abstract 25 selected for detailed full-text review 12 excluded: 8 did not evaluate diagnostic accuracy 2 assessed pleural effusion 2 could not generate 2 × 2 tables
13 included for the meta-analysis Figure 1 A flow chart of study identification, inclusion and exclusion.
calculated the area under SROC curve and the Q*, which were used to demonstrate the overall assessment of the test accuracy. The Q* is the point value on the SROC curve where sensitivity equals specificity. The univariate meta-regression analysis was performed to find the sources of heterogeneity for included trials. The Deek’s funnel plot was used to examine publication bias. Subgroup analysis was also performed to calculate the sensitivity, specificity, PLR, NLR, DOR, the area under SROC curve and the Q* to assess diagnostic accuracy of sTREM-1 in community-acquired and hospital-acquired LRTI, respectively.
RESULTS Study characteristics Our initial search yielded 121 citations and 20 relevant references of searched articles, and reviews were added into our initial search results. One hundred sixteen citations were excluded after reviewing the title and abstract. Twelve articles were excluded on the basis of the full-text review: eight articles did not evaluate diagnostic accuracy, two of them only assessed pleural fluid sTREM-111,12 and two studies could not generate 2 × 2.13,14 Finally, our meta-analysis included 13 studies with 1138 patients (Fig. 1).5–7,15–24 Table 1 showed the detail information of the 13 studies included. The prevalence of LRTI ranged from 0.117 to 0.762. The diagnostic accuracy of sTREM-1 was assessed for both community-acquired and hospital-acquired LRTI in one of the thirteen included studies.15 Partial information of the article18 has been reported in another study.15 Our meta-analysis © 2014 Asian Pacific Society of Respirology
© 2014 Asian Pacific Society of Respirology
France. USA Korea
USA USA Greece Spain Spain China USA
Gibot et al.18 Horonenko et al.19 Huh et al.20
El Solh et al.21 Anand et al.22 Porfyridis et al.23 Ruiz-González et al.5 Ramirez et al.24 Xie et al.6 Palazzo et al.7
50 24 80
>18 >18 >18 75 40 68 124 21 224 45
148 27 212
>18 Adult NA
NA Adult >16 >18 NA >18 >18
Patients (n)
Age (y)
0.507 0.475 0.500 0.117 0.762 0.430 0.442
0.620 0.583 0.363
0.568 0.296 0.415
Prevalence CAP/VAP VAP Pneumonia and Anthonisen type 1 COPD exacerbations, VAP VAP Bacterial and fungal pneumonia Pulmonary aspiration VAP CAP CAP HAP/VAP HAP VAP
Types of infection
BALF BALF Serum Serum Mini-BALF Serum BALF
Mini-BALF BALF BALF
Mini-BALF BALF Serum
Sample source
ELISA ELISA ELISA ELISA ELISA ELISA ELISA
Immunoblot technique ELISA ELISA
Immunoblot technique ELISA Immunoblot technique
Assay method
250 200 19.53 120 900 19 204
5 7 184
5 200 163 400
Cut-off (pg/mL)
25 8 28 10 13 86 15
29 14 25
82 6 71
TP
3 2 13 29 1 18 20
2 9 5
6 3 43
FP
13 11 6 3 3 19 4
2 0 4
2 2 17
FN
34 19 21 82 4 101 6
17 1 46
58 16 81
TN
BALF, bronchoalveolar lavage fluid; CAP, community-acquired pneumonia; COPD, chronic obstructive pulmonary disease; ELISA, enzyme-linked immunosorbent assay; FN, false negative; FP, false positive; HAP, hospital-acquired pneumonia; NA, not available; TN, true negative; TP, true positive; VAP, ventilator-associated pneumonia.
France The Netherlands Singapore
Country
Gibot et al. 15 Determann et al.16 Phua et al.17
Study
Table 1 Characteristics of studies included in the meta-analysis
Diagnostic value of the sTREM-1 in LRTI
503
Respirology (2014) 19, 501–507
504
W Ye et al.
Figure 2 Methodological quality assessment of the included studies. ( ) High quality; ( ) unclear; ( ) low quality.
contains some types of LRTI, including communityacquired pneumonia, hospital acquired pneumonia, ventilator-associated pneumonia (VAP), chronic obstructive pulmonary disease exacerbation and pulmonary aspiration. Five included studies with 432 patients assessed the potential value of sTREM-1 for distinguishing culture-positive LRTI from culturenegative diseases.5,7,15,21,22 Mini-BALF was obtained for three trials, BALF was gathered in six studies and serum was collected in the other four studies. The sTREM-1 values was measured by the immunoblot technique in three studies and by enzyme-linked immunosorbent assay (ELISA) method in the other 10 studies. Among the included studies, the cut-off concentration of sTREM-1 differed greatly, ranging from 5 pg/mL to 163 400 pg/mL. The result (163 400 pg/ mL) was extremely high in one of the studies.17
Quality assessment of the studies The methodological quality of the included trials is shown in Figure 2. In our meta-analysis, the overall QUADAS scores of all studies were above 10, which suggested all included studies were of good quality. Diagnostic accuracy for LRTI Figure 3 showed the forest plot for the pooled sensitivity and specificity of sTREM-1 in the diagnosis of LRTI. The pooled sensitivity of sTREM-1 used for diagnosis of LRTI was 0.84 (95% confidence interval (CI): 0.75–0.90) and the pooled specificity was 0.77 (95% CI: 0.62–0.87). The PLR was 3.6 (95% CI: 2.1–6.3), the NLR was 0.21 (95% CI: 0.13–0.34) and the DOR was 17 (95% CI: 8–40). The SROC curve reveals the trade-off between sensitivity and specificity, and provides an overall summary of test accuracy. Figure 4 showed that the Q* of sTREM-1 for diagnosis of LRTI was 0.82, and the area under the SROC curve was 0.88(95% CI: 0.85–0.90), which indicated a median level of overall diagnostic value of sTREM-1 for LRTI. Respirology (2014) 19, 501–507
The pretest probability of LRTI in our study was 44%. The Fagan’s nomogram indicated that the sTREM-1 detection would raise the post-test probability to 74% and decrease the post-test probability to 14% (Fig. S1 available in the Supporting Information).
Heterogeneity assessment and univariate meta-regression analysis The chi-square test could be used to examine whether the statistically significant heterogeneity existed among included the studies. The I2 for the pooled sensitivity, specificity, PLR, NLR and DOR was 75.2%, 88.6%, 90.26%, 75.39% and 100%, which showed substantial heterogeneity among studies. Therefore, univariate meta-regression analysis was performed to find the sources of potential heterogeneity among included trails. We included the following variables: prevalence of LRTI (≥50% vs
REVIEW
Diagnostic value of the soluble triggering receptor expressed on myeloid cells-1 in lower respiratory tract infections: A meta-analysis WU YE, YANJIE HU, RUIFENG ZHANG AND KEJING YING Department of Respiratory Diseases, Sir Run Run Shaw Hospital, Zhejiang University, Hangzhou, China
ABSTRACT The soluble triggering receptor expressed on myeloid cells-1 (sTREM-1) is a promising diagnostic marker for many types of infections. A bivariate meta-analysis was performed to evaluate its diagnostic value for lower respiratory tract infections (LRTI). We searched PubMed, Cochrane Library and Web of Science (from January 1966 to August 2013) for all trials assessing diagnostic value of sTREM-1 for LRTI. The pooled sensitivity, specificity, positive likelihood ratio(PLR), negative likelihood ratio(NLR), diagnostic odds ratio (DOR), the area under summary receiver operator characteristic (SROC) curve and the Q* were calculated. Thirteen studies with 1138 patients were included in our metaanalysis. The pooled sensitivity and specificity of sTREM-1 for diagnosis of LRTI was 0.84 and 0.77. The PLR, NLR and DOR were 3.6, 0.21 and 17. The area under SROC curve was 0.88 and the Q* was 0.82. The univariate meta-regression analysis demonstrated that the assay method for sTREM-1 significantly affected sensitivity for LRTI. The Q* of sTREM-1 for diagnosis of community-acquired LRTI was 0.82, and the area under SROC curve was 0.88. The Q* of sTREM-1 in diagnosis of hospital-acquired LRTI was 0.83, and the area under SROC curve was 0.90. The Q* of sTREM-1 for distinguishing culture-positive LRTI from culturenegative diseases was 0.79, and the area under SROC curve was 0.86. Current evidence suggests that sTREM-1 is an accurate marker of LRTI. The overall diagnostic value of sTREM-1 for LRTI, communityacquired LRTI and hospital-acquired LRTI is similar. Key words: diagnosis, diagnostic test, lower respiratory tract infection, meta-analysis, soluble triggering receptor expressed on myeloid cells-1.
Abbreviations: BALF, bronchoalveolar lavage fluid; CI, confidence interval; DOR, diagnostic odds ratio; ELISA, enzyme-linked immunosorbent assay; FN, false negative; FP, false positive;
Correspondence: Kejing Ying, Department of Respiratory Diseases, Sir Run Run Shaw Hospital, Zhejiang University, 3 East Qingchun Road, Hangzhou 310016, China. Email: yingkj55 @126.com Received 16 October 2013; invited to revise 25 November 2013; revised 30 November 2013; accepted 31 December 2013 (Associate Editor: Marcos Restrepo). © 2014 Asian Pacific Society of Respirology
ICU, intensive care unit; IL, interleukin; LRTI, lower respiratory tract infections; NLR, negative likelihood ratio; PLR, positive likelihood ratio; QUADAS, quality assessment for studies of diagnostic accuracy; SROC, summary receiver operator characteristic; STARD, standards for reporting of diagnostic accuracy; sTREM-1, soluble triggering receptor expressed on myeloid cells-1; TN, true negative; TP, true positive; TREM-1, triggering receptor expressed on myeloid cells-1; VAP, ventilator-associated pneumonia.
INTRODUCTION The triggering receptor expressed on myeloid cells-1 (TREM-1) is expressed on neutrophils and monocytes, and plays an important role in systemic infections. Upregulation of TREM-1 seems to be a crucial step during bacterial and fungal infections.1,2 TREM-1 is an important signalling receptor, magnifies the inflammatory reaction by elevating pro-inflammatory cytokines and blocking interleukin (IL)-10 synthesis. The upregulation of the TREM-1 leads to elevation of the soluble form of TREM-1 (sTREM-1) in body fluids, including serum, bronchoalveolar lavage fluid (BALF) and cerebrospinal fluid.2 Respiratory tract infections are still the most frequent reason for patients to seek medical service and remain the main source of sepsis.3 Identifying a true ‘gold standard’ for diagnosis of respiratory tract infections is often problematic. The use of blood and sputum cultures have limitations, including the duration of time required to obtain positive cultures, excluding contamination of specimens and identifying colonization.3 Massive research has been performed to discover new diagnostic and prognostic biomarkers. Some biomarkers, including C-reactive protein, procalcitonin, sTREM-1, IL-1b, plasminogen activation inhibitor-1, soluble urokinase-type plasminogen activator receptor and presepsin, appear promising.2,4 Recently, many studies show that sTREM-1 has usefulness for differential diagnosis of lower respiratory tract infections (LRTI).5–7 Therefore, we carried out a bivariate meta-analysis to estimate the diagnostic value of sTREM-1 for LRTI and explored whether its diagnostic values were different for community-acquired and hospitalacquired LRTI. Respirology (2014) 19, 501–507 doi: 10.1111/resp.12270
502
METHODS Search strategy We retrieved Pubmed, Cochrane Library and Web of Science (from January 1966 to August 2013) for all studies examining diagnostic value of sTREM-1 in LRTI. The following keywords or Medical Subject Headings were used: ‘sTREM-1’,‘TREM1 protein, human’, ‘soluble triggering receptor expressed on myeloid cells-1’, ‘respiratory tract infections’, ‘respiratory infection’ and ‘pneumonia’. The references of searched articles and relevant reviews were also reviewed. Study selection We included all studies that fulfilled the following criteria: (i) papers limited to human subjects; (ii) trials evaluating diagnostic accuracy of sTREM-1 for LRTI; (iii) studies provided enough data to construct the 2 × 2 contingency table and calculate true positive (TP), false positive (FP), false negative (FN) and true negative (TN); and (iv) sample source for sTREM-1 was serum or BALF. Data extraction Two reviewers (W.Y. and Y.H.) independently reviewed the trials, screened the citations, selected eligible studies and extracted the relevant information. If disagreement occurred, two reviewers discussed and resolved by consensus. The information extracted from included studies contained the following details: (i) family name of the first author, published year and country of the trial; (ii) category and prevalence of infections; (iii) the number and age of study population; (iv) sample source; (v) TP, FP, FN, TN; (vi) assay method for sTREM-1; and (vii) cut-off value. Quality assessment The methodological qualities of the trails were evaluated using the QUADAS (quality assessment for studies of diagnostic accuracy)8 and the STARD (standards for reporting of diagnostic accuracy).9 We attributed 1 score for each ‘yes’ (high quality), 0 score for each ‘no’ (low quality) and 0.5 score for each ‘unclear’. Statistical analysis We performed the analyses using the statistical software (Stata version 12.0, Stata Corporation, College Station, TX, USA). Bivariate random effects regression model was used to calculate the pooled specificity, sensitivity, diagnostic odds ratio (DOR), positive likelihood ratios (PLR) and negative likelihood ratios (NLR). The DOR of a test is the ratio of the odds of positivity in disease relative to the odds of positivity in the non-diseased: (sensitivity/(1 − sensitivity))/ ((1 − specificity)/specificity).10 The DOR is a single overall indicator of diagnostic accuracy.10 The summary receiver operator characteristic (SROC) curve was constructed with sensitivity on the vertical axis and 1–specificity on the horizontal axis. We also Respirology (2014) 19, 501–507
W Ye et al. 121 citations identified from Pubmed, Cochrane Library and Web of Science
20 relevant references listed in the searched papers and review
141 citations identified 116 citations excluded on the basis of title and abstract 25 selected for detailed full-text review 12 excluded: 8 did not evaluate diagnostic accuracy 2 assessed pleural effusion 2 could not generate 2 × 2 tables
13 included for the meta-analysis Figure 1 A flow chart of study identification, inclusion and exclusion.
calculated the area under SROC curve and the Q*, which were used to demonstrate the overall assessment of the test accuracy. The Q* is the point value on the SROC curve where sensitivity equals specificity. The univariate meta-regression analysis was performed to find the sources of heterogeneity for included trials. The Deek’s funnel plot was used to examine publication bias. Subgroup analysis was also performed to calculate the sensitivity, specificity, PLR, NLR, DOR, the area under SROC curve and the Q* to assess diagnostic accuracy of sTREM-1 in community-acquired and hospital-acquired LRTI, respectively.
RESULTS Study characteristics Our initial search yielded 121 citations and 20 relevant references of searched articles, and reviews were added into our initial search results. One hundred sixteen citations were excluded after reviewing the title and abstract. Twelve articles were excluded on the basis of the full-text review: eight articles did not evaluate diagnostic accuracy, two of them only assessed pleural fluid sTREM-111,12 and two studies could not generate 2 × 2.13,14 Finally, our meta-analysis included 13 studies with 1138 patients (Fig. 1).5–7,15–24 Table 1 showed the detail information of the 13 studies included. The prevalence of LRTI ranged from 0.117 to 0.762. The diagnostic accuracy of sTREM-1 was assessed for both community-acquired and hospital-acquired LRTI in one of the thirteen included studies.15 Partial information of the article18 has been reported in another study.15 Our meta-analysis © 2014 Asian Pacific Society of Respirology
© 2014 Asian Pacific Society of Respirology
France. USA Korea
USA USA Greece Spain Spain China USA
Gibot et al.18 Horonenko et al.19 Huh et al.20
El Solh et al.21 Anand et al.22 Porfyridis et al.23 Ruiz-González et al.5 Ramirez et al.24 Xie et al.6 Palazzo et al.7
50 24 80
>18 >18 >18 75 40 68 124 21 224 45
148 27 212
>18 Adult NA
NA Adult >16 >18 NA >18 >18
Patients (n)
Age (y)
0.507 0.475 0.500 0.117 0.762 0.430 0.442
0.620 0.583 0.363
0.568 0.296 0.415
Prevalence CAP/VAP VAP Pneumonia and Anthonisen type 1 COPD exacerbations, VAP VAP Bacterial and fungal pneumonia Pulmonary aspiration VAP CAP CAP HAP/VAP HAP VAP
Types of infection
BALF BALF Serum Serum Mini-BALF Serum BALF
Mini-BALF BALF BALF
Mini-BALF BALF Serum
Sample source
ELISA ELISA ELISA ELISA ELISA ELISA ELISA
Immunoblot technique ELISA ELISA
Immunoblot technique ELISA Immunoblot technique
Assay method
250 200 19.53 120 900 19 204
5 7 184
5 200 163 400
Cut-off (pg/mL)
25 8 28 10 13 86 15
29 14 25
82 6 71
TP
3 2 13 29 1 18 20
2 9 5
6 3 43
FP
13 11 6 3 3 19 4
2 0 4
2 2 17
FN
34 19 21 82 4 101 6
17 1 46
58 16 81
TN
BALF, bronchoalveolar lavage fluid; CAP, community-acquired pneumonia; COPD, chronic obstructive pulmonary disease; ELISA, enzyme-linked immunosorbent assay; FN, false negative; FP, false positive; HAP, hospital-acquired pneumonia; NA, not available; TN, true negative; TP, true positive; VAP, ventilator-associated pneumonia.
France The Netherlands Singapore
Country
Gibot et al. 15 Determann et al.16 Phua et al.17
Study
Table 1 Characteristics of studies included in the meta-analysis
Diagnostic value of the sTREM-1 in LRTI
503
Respirology (2014) 19, 501–507
504
W Ye et al.
Figure 2 Methodological quality assessment of the included studies. ( ) High quality; ( ) unclear; ( ) low quality.
contains some types of LRTI, including communityacquired pneumonia, hospital acquired pneumonia, ventilator-associated pneumonia (VAP), chronic obstructive pulmonary disease exacerbation and pulmonary aspiration. Five included studies with 432 patients assessed the potential value of sTREM-1 for distinguishing culture-positive LRTI from culturenegative diseases.5,7,15,21,22 Mini-BALF was obtained for three trials, BALF was gathered in six studies and serum was collected in the other four studies. The sTREM-1 values was measured by the immunoblot technique in three studies and by enzyme-linked immunosorbent assay (ELISA) method in the other 10 studies. Among the included studies, the cut-off concentration of sTREM-1 differed greatly, ranging from 5 pg/mL to 163 400 pg/mL. The result (163 400 pg/ mL) was extremely high in one of the studies.17
Quality assessment of the studies The methodological quality of the included trials is shown in Figure 2. In our meta-analysis, the overall QUADAS scores of all studies were above 10, which suggested all included studies were of good quality. Diagnostic accuracy for LRTI Figure 3 showed the forest plot for the pooled sensitivity and specificity of sTREM-1 in the diagnosis of LRTI. The pooled sensitivity of sTREM-1 used for diagnosis of LRTI was 0.84 (95% confidence interval (CI): 0.75–0.90) and the pooled specificity was 0.77 (95% CI: 0.62–0.87). The PLR was 3.6 (95% CI: 2.1–6.3), the NLR was 0.21 (95% CI: 0.13–0.34) and the DOR was 17 (95% CI: 8–40). The SROC curve reveals the trade-off between sensitivity and specificity, and provides an overall summary of test accuracy. Figure 4 showed that the Q* of sTREM-1 for diagnosis of LRTI was 0.82, and the area under the SROC curve was 0.88(95% CI: 0.85–0.90), which indicated a median level of overall diagnostic value of sTREM-1 for LRTI. Respirology (2014) 19, 501–507
The pretest probability of LRTI in our study was 44%. The Fagan’s nomogram indicated that the sTREM-1 detection would raise the post-test probability to 74% and decrease the post-test probability to 14% (Fig. S1 available in the Supporting Information).
Heterogeneity assessment and univariate meta-regression analysis The chi-square test could be used to examine whether the statistically significant heterogeneity existed among included the studies. The I2 for the pooled sensitivity, specificity, PLR, NLR and DOR was 75.2%, 88.6%, 90.26%, 75.39% and 100%, which showed substantial heterogeneity among studies. Therefore, univariate meta-regression analysis was performed to find the sources of potential heterogeneity among included trails. We included the following variables: prevalence of LRTI (≥50% vs
