Article, Cardiology

Role of procalcitonin in the diagnosis of infective endocarditis: a meta-analysis

a b s t r a c t

Background: Infective endocarditis (IE) is a diagnostic challenge. We aimed to systemically summarize the current evidence on the diagnostic value of procalcitonin in identifying IE.

Methods: We searched EMBASE, MEDLINE, Cochrane database, and reference lists of relevant articles with no language restrictions through September 2012 and selected studies that reported the diagnostic performance of PCT alone or compare with other biomarkers to diagnose IE. We summarized test performance characteristics with the use of forest plots, hierarchical summary receiver operating characteristic curves, and bivariate random effects models.

Results: We found 6 qualifying studies that included 1006 episodes of suspected infection with 216 (21.5%) confirmed IE episodes from 5 countries. Bivariate pooled sensitivity, specificity, positive likelihood ratios, and negative likelihood ratios were 64% (95% confidence interval [CI], 52%-74%), 73% (95% CI 58%-84%), 2.35 (95% CI 1.40-3.95), and 0.50 (95% CI 0.35-0.70), respectively. Of the 5 studies examining C-reactive protein (CRP), the pooled sensitivity, specificity, positive likelihood ratios, and negative likelihood ratios were 75% (95% CI 62%-85%), 73% (95% CI 61%-82%), 2.81 (95% CI 1.70-4.65), and 0.34 (95% CI 0.19-0.60), respectively. The global measures of accuracy, Area Under the Receiver Operating Characteristic Curve and Diagnostic odds ratio (dOR), showed CRP (AUC 0.80, dOR 8.55) may have higher accuracy than PCT (AUC 0.71, dOR 4.67) in diagnosing IE. Conclusions: Current evidence does not support the routine use of serum PCT or CRP to rule in or rule out IE in patients suspected to have IE.

(C) 2013


Infective endocarditis (IE) is a life-threatening disease with an incidence varying between 30 and 100 cases per million persons, and a mortality rate of up to 40% [1-4]. By definition, IE is an infection of the endocardial surface of the heart, usually the heart valves, mural endocardium, or a septal defect [3]. IE may cause intraCardiac effects such as severe valvular insufficiency, or congestive heart failure, and extracardiac effects such as disseminated infected emboli and various immunological phenomena [3-5]. Since the first systematic analysis by Sir William Osler in 1885, the clinical features of IE have undergone great changes in developed countries, from a disease commonly

? Conflict of interest: None declared.

?? Funding: None declared.

* Corresponding authors. C.-C. Lee is to be contacted at Douliou, Yunlin County 640, Taiwan. Tel.: +886 5532391×2326; fax: +886 55341452. J.-Y. Wu, Taipei City 11267,

Taiwan. Tel.: +886 33281200×2505; fax: +886 33287715.

E-mail addresses: [email protected] (C.-C. Lee), [email protected] (J.-Y. Wu).

affecting patients with rheumatic valvular heart disease to a disease affecting Injection drug users, elderly persons with valvular sclerosis, patients with cardiovascular prostheses, hospitalized patients with indwelling catheters, and hemodialysis patients [1,2,6]. Mortality and morbidity continue to remain high, despite advances in medical and surgical treatment [3,4].

In clinical practice, IE still poses a diagnostic challenge for clinicians because of the various clinical presentations. Current diagnosis is largely based on the modified Duke’s criteria that integrates clinical findings, microbiological study results, and imaging findings. However, the typical clinical manifestations may be masked by the indiscriminate use of antimicrobial agents, or by underlying conditions in elderly individuals or immunosuppressed persons. Therefore, these criteria have been reported to over- or underdiag- nose IE in patients with subAcute disease or atypical presentations [3,5]. Given that the mortality for IE is as high as 12 to 30% within the first year of diagnosis, and that patients may benefit from the early administration of appropriate antibiotics, early diagnosis is manda- tory [4]. A biomarker with high sensitivity and specificity will greatly improve the diagnostic rate, and thereby influence outcomes.

0735-6757/$ – see front matter (C) 2013

Fig. 1. Flow chart of study identification and inclusion.

procalcitonin is a precursor of calcitonin that is constitutively secreted by C cells of the thyroid gland and K cells of the lung [7]. In healthy individuals, PCT is normally undetectable (b 0.01 ng/mL). When stimulated by endotoxin, PCT is rapidly produced by parenchy- mal tissue throughout the body [8,9]. Unlike CRP, PCT does not respond to sterile inflammation or viral infection [10]. This distinctive characteristic makes PCT a valuable diagnostic marker with broadening range of clinical indications, including the early diagnosis of IE. Despite the early encouraging report from Mueller et al [11], subsequent reports have shown mixed results [12-16].

The aim of this study is to systematically and quantitatively evaluate the role of PCT in the early diagnosis of IE by performing a meta-analysis of published reports.


Systemic meta-analysis guideline adherence

We adhered to the standard methods and procedures for systematic reviews and meta-analyses of diagnostic tests [17,18].

Literature search strategy

We performed a search on PubMed without language restrictions using the keyword ‘procalcitonin’ crossed with ‘endocarditis,’ ‘infective endocarditis,’ ‘Infectious endocarditis,’ ‘endocarditis lenta,’ and ‘sub- acute bacterial endocarditis.’ Our search was limited to human studies published from inception to March 2012. The last update was performed on September 2012. A similar search strategy and similar search terms were used to search in EMBASE. PubMed and EMBASE searches were conducted independently by 2 authors. To ensure a comprehensive acquisition of literature, independent supplemental manual searches were performed on the reference lists of relevant articles and the Cochrane database. Medical Subject Heading (MeSH) and EMbase TREE tool (EMTREE) were used to guide the choice of appropriate search terms. Any inconsistencies between the 2 investigators in article inclusion and exclusion were resolved by consensus.

Study selection and data extraction

We included studies that met all of the following inclusion criteria:

(1) evaluation of PCT alone or compared with other laboratory markers, such as CRP, to diagnose IE; and (2) sufficient data to reconstruct a 2 x 2 contingency table for meta-analysis. Two authors independently assessed all titles/abstracts to determine whether inclusion criteria were satisfied. Full-text articles were retrieved if any of the reviewers considered the abstracts suitable. The 2 reviewers then independently assessed the full text of the retrieved studies for their suitability for inclusion. Discrepancies between the 2 reviewers were resolved by having an additional reviewer assess the full articles. The decision about whether to include any article was made by consensus. The 2 original reviewers independently extracted data from each included study. Among the predefined variables collected were year of publication, study design (prospective or retrospective, cross-sectional or case-control), number of enrolled patients, age group (adults or children), setting (emergency department or ward), study population (general, drug abuser, immunocompromised), reference diagnostic standard for IE, timing of the PCT measurement,

Table 1

Characteristics of the 6 included studies that used biomarkers to assess

Author, year, country

Prevalence (N)

Mean age



testing systems


(PCT, ng/mL; CRP, mg/L)

Outcome(s) definition

Causative microorganisms


Sensitivity Specificity


Sensitivity, Specificity

Kocazeybek B 2003,

0.2 (50)




PCT = 0.19

Duke Criteria

S. sanguis, E. faecalis, Candida

84.0 %

100 %

Turkey [12]

CRP = 10.6

albicans, E. coli, A. baumannii

88.0 %

72.0 %

Mueller C 2004,





PCT= 2.3

Duke Criteria

S. aureus, S. aureus plus

81.0 %

62.5 %

Switzerland [11]


and the modified

Duke Criteria

streptococci, Viridans

streptococci, S. pyogenes,

85.0 %

71.0 %

S. pneumoniae

Watkin RW 2007,





PCT= 0.5

Duke Criteria

S. aureus, CNS, Viridans

46.0 %

81.5 %

UK [13]



streptococci , Enterococcus spp.


74.0 %

Cuculi F 2008,





PCT= 0.64,

Modified Duke Criteria

S. aureus, S. viridans, enterococcus,

52.0 %

52.0 %

Switzerland [14]


coagulase negative bacteraemia

56.0 %

54.0 %

Jereb M 2009,





PCT= 0.5

Duke Criteria

S. aureus, S. coagulase negative,


67.0 %

Slovenia [15]


E. faecalis, V. streptococci



Knudsen JB 2010,





PCT= 0.12

Duke Criteria

S. viridans, S. aureus, coagulase-

67.0 %


Denmark [16]

negative staphylococci, enterococci,

47.0 %

S. pneumoniae

IL-6, interleukin-6; LPSBP, lipopolysaccharide binding protein; CNS: coagulase negative staphylococci; TNF-?, tumor necrosis factor alpha.

Fig. 2. QUADAS criteria for included studies.

markers other than PCT, cutoff of the tested markers, and study results, including sensitivity and specificity. Grading of quality was performed according to the Quality Assessment of Diagnostic- Accuracy Studies (QUADAS) instrument [19], which is an established, evidence-based tool for systematic reviews of diagnostic studies. When multiple pairs of sensitivity or specificity were reported in 1 study, we consistently used the data with the highest Youden index (sensitivity + specificity – 1) for meta-analysis. This tool uses a list of 14 questions, which are answered as “yes,” “no,” or “unclear,” to examine the potential for bias in a study.

Data preparation and statistical analysis

Because pooling sensitivity and specificity separately can produce biased estimates of test accuracy, we preferred to generate pooled estimates when both sensitivity and specificity were reported in a study. We used the random effects bivariate random effects regression model for diagnostic meta-analysis to obtain pooled estimates of the sensitivity and specificity of PCT [20]. Zero cells were handled using a 0.5 continuity correction. The bivariate approach assumes a bivariate distribution for the logit-transformed sensitivity and specificity. The bivariate model estimates and adjusts for the negative correlation between the sensitivity and specificity of the index test due to the threshold effect [20]. The random effect model takes into consideration the between-study variation. We also generated hierarchical summary receiver operating characteristic (HSROC) curves as a way to summarize the global test performance from different diagnostic studies [21]. HSROC curves differ from traditional ROC curves in allowing random effects by each individual study. The HSROC curves generated the curve restricted by the observed range of sensitivity and specificity from the included studies. It does not extrapolate beyond the available data. Both bivariate model and HSROC methods are supported by the Cochrane Diagnostic test accuracy Working Methods group.

Multiple sources of heterogeneity frequently exist in diagnostic

studies. In addition to visual assessment with use of the forest plots, we calculated the inconsistency index (I [2] statistics) to estimate between-study variation that cannot be explained by the within- study variation. We also performed a diagnostic odds ratio (OR) meta- analysis and plotted the forest plots for the diagnostic OR. To test for

possible publication bias, we used a linear regression of log ORs on the inverse root of effective sample sizes to test funnel plot asymmetry [22]. We defined a priori the clinical and study design characteristics as potential relevant covariates: cutoff value and the PCT testing systEMS used. Statistical analyses were conducted using STATA 11.0 (Stata Corp, College Station, TX), notably with the user-written “midas” and “metandi” programs for stata. All statistical tests were 2 sided, and statistical significance was defined as a P b .05.


In total, 171 studies (excluding duplicates) were identified using the search strategy outlined earlier (Fig. 1). After the first round screening of title and abstracts, 156 nonrelevant studies, case reports, or reviews were excluded. Fifteen potential relevant studies were retrieved for full text evaluation, of which 6 further studies were excluded for varying reasons, leaving 6 that met the inclusion criteria. These studies included 1006 episodes of suspected infection with 216 (21.5%) confirmed IE episodes. Only 2 of the 6 studies reported data using the standard PCT cutoff value (0.5 ng/mL). Table 1 presents a summary of the characteristics of the included studies and patients. The number of subjects with IE in each study was 15 to 147, and their mean/median age was 42 to 60.2 years. Four studies used Duke Criteria, and the rest used modified Duke Criteria as the reference diagnostic standards. Common bacteria isolated were Staphylococcus aureus, Viridans streptococci, Enterococcus faecalis, coagulase-negative streptococci, Acinetobacter baumannii, Streptococcus sanguis, Escher- ichia coli, and Streptococcus pneumoniae, and Candida albicans was a common nonbacterial isolate. Measurements of serum PCT levels were performed by 3 different types of assays: semi-quantitative PCT-Q in 1 study, immunoluminometric LUMI test in 2 studies, and TRACE (time resolved amplified cryptate emission) technology Kryptor test in 3 studies. Five studies also examined the value of CRP in diagnosing IE. We used the QUADAS tool for study quality assessment. Fig. 2 provides an overall impression of the methodological quality of the studies. All Blood draws were taken in close proximity to the confirmation diagnosis. All patients were verified by the same reference standard in all studies. None of the included studies explained withdrawals or reported uninterpretable results, and in only one study the physicians were blinded to the index test while

verifying outcomes by reference standards. We could not exclude the possibility of incorporation bias.

Diagnostic accuracy indices

Results of the meta-analysis indicated that PCT testing has a low degree of accuracy for diagnosing IE. We constructed summary ROCs for both PCT and CRP (Fig. 3). PCT had an area under the ROC of 0.77 (95% confidence interval [CI] 0.73-0.80). Pooled sensitivity and specificity estimates of PCT were 0.64 (95% CI 0.52-0.74) and 0.73 (95% CI 0.58-0.84), respectively (Fig. 4). The positive likelihood ratio (LR+, 2.35; 95% CI 1.40-3.95) of the PCT test was not sufficiently high to be used as a rule-in test, and the high negative likelihood ratio (LR-, 0.64; 95% CI 0.32-0.73) could not reduce the pretest probability to a level that would enable Systemic Infection to be safely ruled out. CRP had an AUC of 0.80 (95% CI 0.77-0.84). The pooled sensitivity and specificity estimates of CRP were 0.75 (95% CI 0.62-0.85) and 0.73 (95% CI 0.61-0.82), respectively (Fig. 5). The positive likelihood ratio (LR+, 2.81; 95% CI 1.70-4.65) of CRP testing was not sufficient to support CRP as a rule-in test, but the negative likelihood ratio (LR-, 0.34; 95% CI 0.19-0.60) has moderate rule-out value. Overall, CRP has a higher discriminative capability than PCT in differentiating IE from other causes of systemic inflammatory response syndrome. The diagnostic OR for CRP was 8.55 (95%CI 2.6-28.1), while the diagnostic OR for PCT was 4.67(95% CI 2.0-11.0). We observed a substantial degree of heterogeneity for PCT (I2 = 79.2%; 95% CI 54.5-90.5) and CRP (I2 = 77.8%; 95% CI 45.8-90.7). No significant evidence of potential publication bias was noted by Egger’s test for funnel plot asymmetry (Table 2).

Subgroup analysis

We performed subgroup analysis by restricting analysis to the 3 studies using the high sensitive measurement tool for PCT (Kryptor PCT system, BRHAMS, Berlin, Germany). The specificity did increase, but the sensitivity deceased substantially. Overall, the accuracy was not enhanced (AUC 0.74, diagnostic OR 4.57). We used Galbraith plots to explore sources of heterogeneity. The Galbraith plots showed that studies by Kocazeybek et al and Mueller et al in early years were potential sources of heterogeneity. These 2 studies tend to show higher accuracy for PCT to diagnose IE. Subgroup analysis removing these 2 studies showed improved heterogeneity (I2 = 60.2%; 95% CI 0.0-86.71) with further declined accuracy. Notably, the pooled sensitivity decreased from 0.64 to 0.55 (95% CI 0.44-0.66) and the

AUC from 0.71 to 062 (95% CI 0.57-0.66).


IE is generally thought to originate from bacterial seeding to the turbulence-damaged endothelial surface of the heart. The epidemiol- ogy of IE has changed over the past decade. Increased numbers of elderly, chronically ill, and immunosuppressed patients has made clinical diagnosis more challenging because these patients are often afebrile and unable to mount an adequate immune response to exhibit the classic stigmata of IE caused by autoimmune reaction and peripheral embolization [1,5]. Thus, there is a need for a sensitive diagnostic aid before culture results are available. Advanced imaging modalities such as transesophageal echocardiography or multi-slice computed tomography have developed into useful tools for meeting this clinical challenge, but they are expensive, time consuming, and are not universally available in all levels of hospitals. Thus, a rapid, economical, and accessible biomarker assay may aid in establishing the diagnosis of IE earlier such that appropriate antibiotic treatment can be provided.

PCT have been recently shown to have moderate to high accuracy in diagnosing various systemic bacterial infection, but several report

have shown inconsistent results on the accuracy of PCT in the diagnosis of IE [12-16]. Our study was designed to evaluate the diagnostic performance of PCT for the diagnosis of IE. We found that serum PCT assays, either high or low sensitive assays, have little value as a screening tool in patients who are clinical suspected of having IE. The specificity or negative likelihood ratio of PCT is also unacceptably poor to be used as a rule-out tool. Our data showed CRP has a superior accuracy, notably in specificity, over PCT for diagnosis of IE.

Previous meta-analyses addressing the value of serum PCT have suggested that PCT is of moderate value (AUC 0.84, 95% CI 0.75-0.9; sensitivity 0.76, 95% CI 0.66-0.84; specificity 0.70, 95% CI 0.60-0.79) in distinguishing sepsis from other noninfectious causes of systemic

Fig. 3. Receiver operating curve analysis summary receiver operating characteristic (SROC) curve (solid line) of PCT (A) and CRP (B) for patients with infective endocarditis and the bivariate summary estimate (solid square), together with the corresponding 95%confidence ellipse (inner dashed line) and 95% prediction ellipse (outer dotted line). The symbol size for each study is proportional to the study size.

Fig. 4. Forest plots for (A) sensitivity and (B) specificity for studies using PCT to detect among patients with infective endocarditis.

inflammation syndrome [23]. Our meta-analysis showed PCT may have even lower accuracy in diagnosing IE than in diagnosing of other sources of sepsis. The overall positive likelihood ratio after excluding 2 potential outlier studies was only 2.22 (95% CI 1.36-3.65), not sufficiently high to be used as a reliable rule-in tool [24]. In a virtual population with a 20% prevalence (pretest probability) of IE, a positive likelihood ratio of 2.22 translates into a positive predictive value (posttest probability) of 36%. In other words, only 1 in 3 patients with positive PCT test results can be expected to have confirmed IE. Similarly, a negative likelihood ratio of 0.60 translates into 13% posttest probability. In other words, only 1 in 8 patients with negative PCT results may turn out to have IE. Compared to PCT, the posttest probability for patients with positive and negative CRP test result is 41% and 8%, respectively, still not sufficiently accurate as a major diagnostic tool for IE. IE is a disorder with a high mortality rate if not treated promptly. Thus, the decision to exclude a patient with possible IE from further imaging based on a negative serum PCT or CRP value might be hazardous. We do not, however, think that PCT and CRP are of no diagnostic value in the setting of IE. Instead, we recommend that further studies are needed to examine whether PCT has additional diagnostic or prognostic value to conventional clinical variables.

High heterogeneity was observed for the overall accuracy estimated for PCT. We tried to explore the potential source for heterogeneity and found studies by Mueller et al and Kocazeybek et al were potential outliers on Galbraith plot [25]. The 2 studies reported significantly higher accuracy than other studies. Mueller et al reported an optimum PCT cutoff value of 2.3 ng/mL, which is significantly higher than those in other studies (range 0.12-0.64 ng/mL) [11]. Kocazeybek et al adopted a case-control design comparing IE to non- infected and non-IE infected controls [12]. It has been shown that the case-control design tends to over-estimate the accuracy of a diagnostic test [17]. In addition, 3 different assay kits used in the 6 included studies may also have contributed to the between-study inconsistency. The 3 PCT assays used were Kryptor (BRHAMS, Berlin, Germany), LUMItest (BRHAMS), and the PCT-Q assay systems (BRHAMS). The PCT-Q assay system is a semi-quantitative bedside assay and the LUMItest PCT test uses an immunoluminometric method with a reported functional sensitivity (20% CV) of 0.5 ng/ mL. Values b 0.5 ng/mL, which is still of clinical importance in patients with IE, may lack precision. Only 3 studies used the Kryptor PCT assay with higher precision (detection limit of 0.02 ng/mL; functional sensitivity of 0.06 ng/mL) [26]. Subgroup analysis on studies using the

Fig. 5. Forest plots for (A) sensitivity and (B) specificity for studies using CRP to detect among patients with infective endocarditis.

high sensitive kit, however, did not show significant improvement in the diagnostic accuracy. Further studies are still needed to verify this result.

Our study has some potential limitations. First, although a single threshold of 0.5 ng/mL is usually used to define a positive serum PCT finding, we do not have enough data to provide a pooled summary accuracy estimate on this cutoff value. Second, because of the limited number of included studies, we did not perform meta-regression analyses to further investigate the statistical heterogeneity.


In conclusion, current evidence does not support the routine use of serum PCT or CRP as a Biochemical test to rule in or rule out IE in patients who are suspected to have IE. The high sensitive PCT test does not appear to help clinicians to screen out more patients who need Advanced diagnostic imaging. Considering the small sample size and suboptimal case-control design of the currently available studies, a sufficiently powered cohort design prospective study is needed to

Table 2

Summary of subgroup analysis of the 6 included studies


Studies (n)

Sensitivity (95% CI)

Specificity (95% CI)

Likelihood ratio+

Likelihood ratio-

AUROC (95% CI)

Diagnostic OR (95% CI)

I2 (95% CI)

Publication bias (Egger’s test p)

PCT [11-16]








79.2 (54.5-90.5)


PCT excluding








60.2 (0.0-86.71)


potential outlier

Kryptor [11,14,16]


0.67 (0.60-0.74)








CRP [11-15]










conclusively address the usefulness of serum PCT as a diagnostic aid in IE.


We thank Pei-Shan Hsieh from Medical Wisdom Inc for her help in creating tables and graphics.


  1. Tleyjeh IM, Abdel-Latif A, Rahbi H, et al. A systematic review of population-based studies of infective endocarditis. Chest 2007;132(3):1025-35.
  2. Coburn B, Morris AM, Tomlinson G, Detsky AS. Does this adult patient with suspected bacteremia require blood cultures? JAMA 2012;308(5):502-11.
  3. Pierce D, Calkins BC, Thornton K. Infectious endocarditis: diagnosis and treatment. Am Fam Physician 2012;85(10):981-6.
  4. Thuny F, Grisoli D, Collart F, Habib G, Raoult D. Management of infective endocarditis: challenges and perspectives. Lancet 2012;379(9819):965-75.
  5. Timsit JF, Dubois Y, Minet C, et al. New challenges in the diagnosis, management, and prevention of central venous catheter-related infections. Semin Respir Crit Care Med 2011;32(2):139-50.
  6. Fitzsimmons K, Bamber AI, Smalley HB. Infective endocarditis: changing aetiology of disease. Br J Biomed Sci 2010;67(1):35-41.
  7. Maruna P, Nedelnikova K, Gurlich R. Physiology and genetics of procalcitonin. Physiol Res 2000;49(Suppl 1):S57-61.
  8. Brunkhorst FM, Heinz U, Forycki ZF. Kinetics of procalcitonin in iatrogenic sepsis. Intensive Care Med 1998;24(8):888-9.
  9. Dandona P, Nix D, Wilson MF, et al. Procalcitonin increase after endotoxin injection in normal subjects. J Clin Endocrinol Metab 1994;79(6):1605-8.
  10. Assicot M, Gendrel D, Carsin H, Raymond J, Guilbaud J, Bohuon C. High serum procalcitonin concentrations in patients with sepsis and infection. Lancet 1993; 341(8844):515-8.
  11. Mueller C, Huber P, Laifer G, Mueller B, Perruchoud AP. Procalcitonin and the early diagnosis of infective endocarditis.Circulation 2004;109(14):1707-10 [Epub 2004 Apr 1705].
  12. Kocazeybek B, Kucukoglu S, Oner YA. Procalcitonin and C-reactive protein in infective endocarditis: correlation with etiology and prognosis. Chemotherapy 2003;49(1-2):76-84.
  13. Watkin RW, Harper LV, Vernallis AB, et al. Pro-inflammatory cytokines IL6, TNF- alpha, IL1beta, procalcitonin, lipopolysaccharide binding protein and C-reactive protein in infective endocarditis. J Infect 2007;55(3):220-5.
  14. Cuculi. Serum procalcitonin has the potential to identify Staphylococcus aureus endocarditis. Eur J Clin Microbiol Infect Dis 2008;27(11):1145-9 [Epub 2008 Jun 1143].
  15. Jereb M, Kotar T, Jurca T, Lejko Zupanc T. Usefulness of procalcitonin for diagnosis of infective endocarditis. Intern Emerg Med 2009;4(3):221-6.
  16. Knudsen JB, Fuursted K, Petersen E, et al. Procalcitonin in 759 patients clinically suspected of infective endocarditis. Am J Med 2010;123(12):1121-7.
  17. Leeflang MM, Deeks JJ, Gatsonis C, Bossuyt PM. Cochrane Diagnostic Test Accuracy Working G. Systematic reviews of diagnostic test accuracy. Ann Intern Med 2008; 149(12):889-97.
  18. Gatsonis C, Paliwal P. Meta-analysis of diagnostic and screening test accuracy evaluations: methodologic primer. AJR Am J Roentgenol 2006;187(2):271-81.
  19. Whiting P, Rutjes AW, Reitsma JB, Bossuyt PM, Kleijnen J. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3:25.
  20. Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol 2005;58(10):982-90.
  21. Harbord RM, Whiting P, Sterne JA, et al. An empirical comparison of methods for meta-analysis of diagnostic accuracy showed hierarchical models are necessary. J Clin Epidemiol 2008;61(11):1095-103.
  22. Song F, Khan KS, Dinnes J, Sutton AJ. Asymmetric funnel plots and publication bias in meta-analyses of diagnostic accuracy. Int J Epidemiol 2002;31(1):88-95.
  23. Simon L, Gauvin F, Amre DK, Saint-Louis P, Lacroix J. Serum procalcitonin and C-reactive protein levels as markers of bacterial infection: a systematic review and meta-analysis. Clin Infect Dis 2004;39(2):206-17.
  24. Halkin A, Reichman J, Schwaber M, Paltiel O, Brezis M. Likelihood ratios: getting diagnostic testing into perspective. QJM 1998;91(4):247-58.
  25. Galbraith RF. A note on graphical presentation of estimated odds ratios from several clinical trials. Stat Med 1988;7(8):889-94.
  26. Hausfater P, Brochet C, Freund Y, Charles V, Bernard M. Procalcitonin measurement in routine emergency medicine practice: comparison between two immunoassays. Clin Chem Lab Med 2010;48(4):501-4.

Leave a Reply

Your email address will not be published. Required fields are marked *