Secondary Logo

Journal Logo

Original Articles: Gastroenterology

Faecal Calprotectin in Suspected Paediatric Inflammatory Bowel Disease

Degraeuwe, Pieter L.J.*; Beld, Monique P.A.*; Ashorn, Merja; Canani, Roberto Berni; Day, Andrew S.§; Diamanti, Antonella; Fagerberg, Ulrika L.#; Henderson, Paul**; Kolho, Kaija-Leena‡‡; Van de Vijver, Els§§; van Rheenen, Patrick F.||||; Wilson, David C.**; Kessels, Alfons G.H.¶¶

Author Information
Journal of Pediatric Gastroenterology and Nutrition: March 2015 - Volume 60 - Issue 3 - p 339-346
doi: 10.1097/MPG.0000000000000615

Abstract

As many as 25% of all inflammatory bowel disease (IBD) cases present in childhood. Given the impact of the disease and its therapy, a reliable and timely diagnosis is mandatory (1). The reference standard, endoscopic evaluation with biopsies (2), is, however, invasive, not without risks, and also costly.

Faecal calprotectin (FC) has been investigated as a surrogate marker of neutrophil influx into the bowel lumen and a noninvasive diagnostic test for IBD (3,4).

van Rheenen et al (5) published a study-level meta-analysis on the diagnostic accuracy of FC for IBD in both children and adults. Based on the calculated summary estimates for specificity, sensitivity, and likelihood ratios (LRs), they concluded that the test is a useful screening tool for identifying those patients most likely needing endoscopic evaluation. A key problem is, however, that this meta-analysis was based on aggregate data and did not account for different cutoff levels or different patient characteristics (6).

By contrast, individual patient data (IPD) meta-analysis allows using the original test results as continuous data rather than as dichotomous classification data. In addition, the effect of patient characteristics on test accuracy can be evaluated and quantified (6,7).

We undertook the present study to update the study-level meta-analysis on the diagnostic performance of FC for paediatric IBD and to complement it with an individual-level meta-analysis. We also sought to develop a prediction model for IBD based on calprotectin level and readily available patient variables.

METHODS

The research was conducted in accordance with the Code of Conduct for Medical Research of the Dutch Federation of Biomedical Scientific Societies (8). Identification, selection, and appraisal of the relevant studies were carried out independently, by 2 reviewers (P.L.J.D. and M.P.A.B.). Disagreements were resolved through discussion.

Search Methods for Identification of Studies

A systematic search was performed initially from inception to December 2010 in MEDLINE (Ovid), EMBASE (Ovid), the Database of Abstracts of Reviews of Effects (DARE) of the Centre for Reviews and Dissemination (9), and the MEDION database of the University of Maastricht (10). The electronic search was last updated on April 1, 2012. We refrained from using a diagnostic search filter (11,12) or language restrictions. Details of the search are given in an online-only document (eText 1, https://links.lww.com/MPG/A396). Duplicate articles identified in both Medline and Embase were manually deleted. The reference lists of selected studies were checked for further relevant studies.

Study Selection

Only cohort studies evaluating the diagnostic performance of FC concentration in paediatric patients suspected of having IBD were considered for review. Other study designs, such as case-control, are indeed prone to spectrum bias (13). In addition, the following inclusion criteria were applied: FC measurement and reference standard available for all of the paediatric participants (or a follow-up period long enough to exclude IBD), and sufficient data to calculate 2 × 2 tables.

Data Abstraction

The following data were extracted from each study on a predesigned form: the spectrum of the studied population (indication for testing, age, and sex), details of the index and reference test, and counts in the 2 × 2 table.

IPD Sets

For the IPD meta-analysis, we contacted the authors of the selected studies and invited them to share the raw, deidentified study data (FC level, age, sex, and final diagnosis). The raw data were checked for internal consistency against the summary results published in the original article. Some small discrepancies were found, discussed with the authors, and ascertainable divergences were corrected.

Quality Assessment

The quality of the included studies was assessed using the revised QUADAS tool (14). Because the calprotectin assay is an objective measurement, 3 of 14 items were omitted: blinded interpretation of the index test, availability of clinical data, and reporting of uninterpretable results. Two reviewers independently answered the 11 remaining questions in the affirmative, in the negative, or as being unclear.

Data Synthesis and Statistical Analysis

Literature-Based Meta-Analysis

The aggregate data meta-analysis was performed in Stata/SE version 11.2 (StataCorp, College Station, TX) using the Midas command (15). Accordingly, summary statistics for all of the diagnostic performance indices were calculated within the bivariate mixed-effects logistic regression modelling framework. Between-study heterogeneity of the results was assessed graphically by using forest plots of the diagnostic odds ratios (ORs), and statistically by using the χ2 test of homogeneity and the inconsistency index (I2) (16). An I2 value >50% was taken to indicate significant heterogeneity. The potential for publication bias was estimated by using a Deeks’ funnel plot. As recommended, P <0.1 was considered statistically significant (17). A hierarchical summary receiver operating characteristic graph with 95% confidence interval (CI) region and 95% prediction region was constructed.

Individual Patient-Based Meta-Analysis

Although we largely prefer the analytical approach using logistic regression, we also evaluated the diagnostic performance of FC based on the merged individual data. The diagnostic performance was calculated through the MedCalc software, version 11.5.01. (MedCalc Software, Mariakerke, Belgium).

Logistic Regression Analysis: Predicted Probability

The contribution of calprotectin concentration and age as continuous variables, and sex and study as categorical variables, to the diagnosis of IBD was explored using stepwise forward (LR) binary logistic regression analysis. The logit and logistic commands in Stata/SE 10.1 were used. An entry probability for each variable was set at 0.05. A clinical prediction rule was derived from the final regression model (see eText 1, https://links.lww.com/MPG/A396, for calculation details).

RESULTS

Description of Studies

Our search (Fig. 1) returned 161 citations. After removal of 36 duplicates and because 104 of the abstracts were considered not pertinent, 23 records were retrieved as full texts. An additional 4 studies were excluded because of the inappropriateness of the population studied. Insufficient information was available for constructing a 2 × 2 table from 3 studies. Eventually, 8 cohort studies fulfilled the inclusion criteria (18–25). A ninth relevant population-based cohort study was discovered in a recently published diagnostic meta-analysis (5). The original study (26) does not provide data to construct a 2 × 2 table, but the meta-analysis does.

FIGURE 1
FIGURE 1:
Flowchart showing the search for and selection of papers evaluating FC in children with suspected IBD. The final search was carried out on April 1, 2012. FC = faecal calprotectin; IBD = inflammatory bowel disease.

The authors of 8 cohort studies (18–25) were willing to share a dataset with IPD (age, sex, calprotectin concentration, and final diagnosis). Age and sex were not available from the Diamanti et al study owing to a computer crash. The authors of the Ashorn et al article (18) provided data on 31 additional patients recruited after the publication of their study results. In the most recently published study (25), not all of the patients underwent endoscopy, but a sufficiently extended follow-up period (disease-free period of 6 months) should minimise the risk of verification bias (27). To completely eliminating this type of bias, we excluded the 43 patients without histologically confirmed IBD from the IPD meta-analysis. The Norwegian group (26) justified their refusal by arguing that the data will again be used for their own follow-up study. None of the authors are aware of missed (un)published diagnostic accuracy studies fulfilling the selection criteria.

Characteristics of Included and Excluded Studies

All 9 included cohort studies (18–26) were undertaken in referral centres for paediatric gastroenterology and involved 853 patients from the toddler age group to young adults. Table 1 summarises the characteristics of the study population, a description of the applied index and reference tests, and the findings. In 1 study (19) 4 patients were excluded because infectious gastroenteritis was the final diagnosis. Upon enquiry, none of the authors were aware of similar patients in whom readily available diagnostic investigation could have prevented unjustified study entry and unnecessary false-positive or true-negative cases. Another study (24) contained 16 patients whose calprotectin results were expressed as greater or less than a numerical value. These test results were substituted by a value midway between the reported numerical value and a higher or lower value present in the dataset. This procedure is unlikely to influence the test accuracy parameters because 14 of the 16 adjusted values were situated in the highest or lowest quintile. The bar graph in eFigure 1 (https://links.lww.com/MPG/A395) is a representation of the quality of the 9 selected cohort studies (18–26).

TABLE 1
TABLE 1:
Characteristics and 2 × 2 data of all included studies

Summary Results for All Included Studies

Literature-Based Meta-Analysis

A summary of test accuracy estimates is shown in Table 2. The Cochran Q test and the I2 values are indicative for substantial heterogeneity (28). Figure 2 shows a forest plot of the diagnostic OR and the pooled estimate for the 9 cohort studies (18–26). The average prevalence of IBD in the cohort studies was 0.54. Accordingly, the posttest probability of a positive calprotectin test could be estimated to be 0.79 (95% CI 0.72–0.85). The posttest probability of a negative calprotectin test is 0.05 (0.01–0.14). Figure 3 shows the hierarchical summary receiver operating characteristic graph with 95% CI region and 95% prediction region.

TABLE 2
TABLE 2:
Numerical results of the literature-based meta-analysis (FC for the diagnosis of IBD)
FIGURE 2
FIGURE 2:
Forest plot of diagnostic OR of each individual cohort study, pooled odds ratio, Cochran-Q test heterogeneity and I 2 statistic for inconsistency. CI = confidence interval; OR = odds ratio.
FIGURE 3
FIGURE 3:
Summary ROC space of sensitivity and specificity for FC in the diagnosis of paediatric IBD. The circles depict the observed bivariate pairs of sensitivity and specificity of the 9 diagnostic cohort studies (18–26). The blue solid line is the summary ROC curve. The diamond is the bivariate summary point. The red dashed line triangle is the bivariate boundary of the 95% CI region for the bivariate summary point, and the green dotted line encloses the 95% prediction region. The study numbers correspond to those reported in Table 1. AUC = area under the ROC curve; CI = confidence interval; FC = faecal calprotectin; IBD = inflammatory bowel disease; SROC = summary receiver operating characteristic.

Publication bias was assessed by Deeks’ funnel plot asymmetry test for small study effect/publication bias. The nonsignificant slope (P = 0.62) indicates that no significant bias was found.

Individual Patient-Based Meta-Analysis

IPD on final diagnosis and FC were collated from 742 children from 8 studies (18–25). The studies were significantly different with respect to age (P = 0.004, 1-way analysis of variance [ANOVA] test), mean FC concentration (P < 0.001, 1-way ANOVA test), and area under the receiver operating characteristic (ROC) curve (AUC; P = 0.001, 1-way ANOVA). The pretest probability ranged from 0.51 (95% CI 0.39–0.63) to as high as 0.84 (0.75–0.90).

In the pooled dataset of all of the children, the “optimal” cutoff value (the value with the highest accuracy − minimal false-negative and false-positive results) for FC was 212 μg/g corresponding with a sensitivity of 0.90 (95% CI 0.87–0.93), a specificity of 0.85 (0.81–0.88), a positive LR of 5.99 (4.6–7.8), and a negative LR of 0.11 (0.09–0.20). The AUC of FC for the diagnosis of paediatric IBD was 0.94 (95% CI 0.92–0.95).

Logistic Regression Analysis: Predicted Probability

The results of the logistic regression analysis (eText 1, https://links.lww.com/MPG/A396) disclosed that calprotectin concentration, age, and study centre were independent predictors of IBD. The influence of study centre (LR test significant) confirms heterogeneity across the studies.

Outside the 8 study centres, a regression equation containing a term related to the study centre is not of use; therefore, the impact of study centre on the diagnosis was omitted from the final regression model: logit(P) = S = −3.294 + 0.004 × FC + 0.175 × AGE, where the explanatory variable FC is the FC concentration in micrograms per gram and AGE is the age in years. The predictivity of this simplified model was evaluated by ROC curve analysis. The estimated AUC for the final model was calculated to be 0.92 (95% CI 0.89–0.94). The logistic model using calprotectin concentration and age predicts IBD correctly in 85.5% (466/545) of children (sensitivity 0.81; 95% CI 0.76–0.85), specificity 0.92 (0.88–0.95), positive predictive value 0.93 (0.89–0.96), and negative predictive value 0.73 (0.72–0.83).

The probability of having IBD is determined by the equation: P = exp(S)/(1 + exp(S)). We programmed an Excel spreadsheet (eFile1, https://links.lww.com/MPG/A397) computing the probability (with CI) of having IBD for each combination of FC, age, and disease prevalence. Figure 4 illustrates that assuming a disease prevalence of 56%, an FC of 700 μg/g in a 6-year-old child corresponds to an IBD probability of 64% (95% CI 52–71). This prediction lacks precision and the posttest probability is hardly improved compared with the average pretest probability. In a 17-year-old adolescent the same test result makes the diagnosis quite probable, but ruling out IBD at this age and this pretest probability is practically impossible.

FIGURE 4
FIGURE 4:
Plots of the predicted probabilities as a function of FC concentration, age, and disease prevalence. This figure illustrates that age has an important additional value for calprotectin testing for the diagnosis of IBD. CI = confidence interval; FC = faecal calprotectin; IBD = inflammatory bowel disease; LL = lower limit of 95% CI; UL = upper limit of 95% CI.

DISCUSSION

This systematic review has provided an updated aggregate data meta-analysis confirming the high diagnostic accuracy of FC for the detection of IBD in referral centres for paediatric gastroenterology. In addition, a meta-analysis using individual participant data enabled us to develop an algorithm predicting that the probability of having IBD depends on the FC level and the age of the child. This was previously not recognised.

The literature-based meta-analysis indeed confirms that FC has an excellent overall sensitivity of 0.97 (95% CI 0.92–0.99) and a modest specificity of 0.71 (0.59–0.80) for diagnosing paediatric IBD. A first observation is that, owing to the relatively small number of pooled investigated patients, the imprecision of the predictive values is considerable.

In addition, although we used recommended, robust state-of-the-art statistical methodology (29), the estimation of summary points does not hold true if the included studies have used different threshold values (30). For the clinician, it is not evident at which cutoff value the calculated summary estimates of the test accuracy measures apply. Therefore, we recalculated the diagnostic performance of FC using merged IPD. The “optimal” cutoff for the whole group equals 212 μg/g. Taking into account the pretest probability and the LRs, the clinician can now better interpret a test result and discuss the predictive value with the patient and his or her parents. We should, however, realise that the predictive values or LRs are applicable to the whole group of patients presenting with a test result above or below the threshold and that dichotomising carries with it a loss of information for the individual child (31). This drawback can be overcome by using logistic regression analysis, a technique that allows taking into consideration patient-level covariates as well.

The contribution of age in the prediction of IBD turned out to be significant. FC concentrations have been shown to be higher in preschool children, especially infants, than in older children (32,33). This age dependency does not seem to play an important role in our logistic regression model because there is no significant correlation between age quartile and FC in non-IBD subjects (r = −0.06, P = 0.34). In contrast, the prevalence of IBD is significantly increased with higher age quartile (χ2 test, P < 0.0001), suggesting that the age factor corrects for the age-dependent prevalence.

Our literature search revealed some of the well-known shortcomings in the quality and reporting of diagnostic research (27). We had to exclude nearly half of the paediatric accuracy studies because they used a case-control design known for introducing spectrum bias (13). None of the studies prespecified a target value for sensitivity, specificity, predictive accuracy, and a minimal acceptable lower confidence limit, enabling sample size calculation (34,35). Admittedly, neither did we define a target region within which the summary estimates of our meta-analysis should fall. We suggest that the absence of a predefined target makes authors undeservedly enthusiastic about the diagnostic value of an index test.

Our meta-analysis differs in some aspects from the study of van Rheenen et al (5). They included 1 study that we excluded (36), whereas we added 3 new studies totalling 404 participants (20,22,25). Golden (personal communication, March 18, 2010) discouraged us from using her group's data (36) because Magne Fagerhol's original assay, measuring calprotectin as milligrams per litre assay buffer, does not correlate well with the new commercial assays expressing the results as micrograms per gram faeces.

The major difference between the results predicted by van Rheenen et al (5) and this article is the IPD meta-analysis. As already known, the collection of IPD is time-consuming and difficult. A high level of persuasiveness was needed to collect the raw study data from 8 of the 9 cohort studies. Consequently, we could use easily available patient characteristics and the original continuous calprotectin data instead of the dichotomised study results in a logistic regression analysis. It is also noteworthy that we detected and could correct small discrepancies between the raw data and the published data.

By publishing all of the available raw data (eTable 1, https://links.lww.com/MPG/A398), we comply with the appropriate and growing request to share complete data, allowing re-analysis by the reader and sequential meta-analysis if new studies appear. There is indeed a growing awareness that sharing data is an ethical obligation (37–42).

Compared with the systematic review of van Rheenen et al (5), the key strength of our study is the IPD meta-analysis. Although a literature-based meta-analysis provides a good overall impression of the diagnostic accuracy of the test, pooling of studies with different diagnostic thresholds (50–160 μg/g faeces) precludes the calculation of the predictive value at a self-selected threshold or the patient's result. Only IPD permits prediction based on numerical test results and adjustment for patient characteristics. Our prediction tool calculates an individualised disease risk with accompanying confidence range taking into account the FC concentration and the patient's age. In a paediatric gastroenterology referral centre, ruling out IBD is difficult in older children, whereas in younger patients a higher FC concentration is needed to obtain a posttest probability larger than the prevalence.

Our search methodology aimed to avoid language and citation bias, but we cannot exclude publication or reporting bias in our meta-analysis. Publication bias is probably even more of a problem for diagnostic and prognostic than for therapeutic studies (43,44). The effects of reporting bias in therapeutic meta-analyses have been shown to be substantial (45).

Our IPD meta-analysis may also experience availability bias. Data from 2 large studies are incomplete or unavailable. In this context, it is noteworthy that the study of Perminow et al (26) was not intended to be a diagnostic accuracy study and showed the lowest diagnostic OR.

We also recognise that the number of patients in the meta-analysis is limited. Therefore, the precision of the diagnostic accuracy measures and our predictive algorithm leaves room for improvement. Furthermore, the final logistic regression equation contains only 2 independent variables. This likely represents an oversimplification of a complex clinical diagnostic process, including information collected during history taking and physical examination. The user of the algorithm should be aware of this. Other laboratory test results, such as C-reactive protein, haemoglobin, and iron indicators, could improve diagnostic precision but were not available for this study. Clinical suspicion or “gut feeling” may also be of additional diagnostic value.

Finally, the between-study heterogeneity is of concern and not fully understood. Although the selection criteria are more or less the same (suspected IBD), between-study differences were shown for age and FC concentration. The prevalence of IBD varied from 51% to 84%, and the false-positive rate ranged from 0% to 29% suggesting dissimilar study entry criteria. We have also noticed that there was a negative, just significant linear correlation (r = −0.71, P = 0.049) between prevalence and the false-positive rate.

Differences in clinical laboratory measurement procedures for FC may be another source of heterogeneity (46). We know that even small analytic biases can indeed shift the laboratory values and lead to diagnostic misclassification (47,48).

Implications for Clinical Practice and Future Research

Despite the absence of validation, and given the mentioned concern about study heterogeneity and (reporting) bias, our predictive algorithm is presently the best available tool for predicting IBD in the individual child. As such it deserves a valid place in the decision-making process.

We nevertheless advocate a large prospective multicentre study to improve and refine the IBD screening tool. More data are needed to validate the prediction algorithm on a different dataset to improve the precision of the tool. Every effort should be made to standardise and harmonise calprotectin measurements, and the predictive value of other clinical variables and (faecal) biomarkers (49,50) should be investigated simultaneously.

CONCLUSIONS

Using an IPD meta-analysis and through regression modelling, we identified FC concentration and age as independently associated with the diagnosis of IBD. We developed a prediction rule that enables the practicing paediatric gastroenterologists to numerically interpret the FC value together with the patient's age. As such, this rule can be a valuable adjunct to the diagnostic armamentarium making physicians and patients/families better equipped to make personalised decisions.

REFERENCES

1. Timmer A, Behrens R, Buderus S, et al. Childhood onset inflammatory bowel disease: predictors of delayed diagnosis from the CEDATA German-Language Pediatric Inflammatory Bowel Disease Registry. J Pediatr 2011; 158:467–473.
2. Levine A, Koletzko S, Turner D, et al. ESPGHAN Revised Porto criteria for the diagnosis of inflammatory bowel disease in children and adolescents. J Pediatr Gastroenterol Nutr 2014; 58:795–806.
3. Konikoff MR, Denson LA. Role of fecal calprotectin as a biomarker of intestinal inflammation in inflammatory bowel disease. Inflamm Bowel Dis 2006; 12:524–534.
4. von Roon AC, Karamountzos L, Purkayastha S, et al. Diagnostic precision of fecal calprotectin for inflammatory bowel disease and colorectal malignancy. Am J Gastroenterol 2007; 102:803–813.
5. van Rheenen PF, Van de Vijver E, Fidler V. Faecal calprotectin for screening of patients with suspected inflammatory bowel disease: diagnostic meta-analysis. BMJ 2010; 341:c3369.
6. Riley RD, Dodd SR, Craig JV, et al. Meta-analysis of diagnostic test studies using individual patient data and aggregate data. Stat Med 2008; 27:6111–6136.
7. Lyman GH, Kuderer NM. The strengths and limitations of meta-analyses based on aggregate data. BMC Med Res Methodol 2005; 5:14.
8. Code of Conduct for Medical Research of the Dutch Federation of Biomedical Scientific Societies. http://www.federa.org/sites/default/files/bijlagen/coreon/code_of_conduct_for_medical_research_1.pdf. Accessed October 10, 2012.
9. The University of York Centre for Reviews and Dissemination. http://www.crd.york.ac.uk/crdweb. Accessed April 1, 2012.
10. The Medion Database. www.mediondatabase.nl. Accessed April 1, 2012.
11. Doust JA, Pietrzak E, Sanders S, et al. Identifying studies for systematic reviews of diagnostic tests was difficult due to the poor sensitivity and precision of methodologic filters and the lack of information in the abstract. J Clin Epidemiol 2005; 58:444–449.
12. Leeflang MM, Scholten RJ, Rutjes AW, et al. Use of methodological search filters to identify diagnostic accuracy studies can lead to the omission of relevant studies. J Clin Epidemiol 2006; 59:234–240.
13. Lijmer JG, Mol BW, Heisterkamp S, et al. Empirical evidence of design-related bias in studies of diagnostic tests. JAMA 1999; 282:1061–1066.
14. Whiting PF, Weswood ME, Rutjes AW, et al. Evaluation of QUADAS, a tool for the quality assessment of diagnostic accuracy studies. BMC Med Res Methodol 2006; 6:9.
15. MIDAS: Stata module for meta-analytical integration of diagnostic test accuracy studies [computer program]. Boston College Department of Economics; 2009/02/05/.
16. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ 2003; 327:557–560.
17. Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol 2005; 58:882–893.
18. Ashorn S, Honkanen T, Kolho KL, et al. Fecal calprotectin levels and serological responses to microbial antigens among children and adolescents with inflammatory bowel disease. Inflamm Bowel Dis 2009; 15:199–205.
19. Berni Canani RB, de Horatio LT, Terrin G, et al. Combined use of noninvasive tests is useful in the initial diagnostic approach to a child with suspected inflammatory bowel disease. J Pediatr Gastroenterol Nutr 2006; 42:9–15.
20. Diamanti A, Panetta F, Basso MS, et al. Diagnostic work-up of inflammatory bowel disease in children: the role of calprotectin assay. Inflamm Bowel Dis 2010; 16:1926–1930.
21. Fagerberg UL, Loof L, Myrdal U, et al. Colorectal inflammation is well predicted by fecal calprotectin in children with gastrointestinal symptoms. J Pediatr Gastroenterol Nutr 2005; 40:450–455.
22. Henderson P, Casey A, Lawrence SJ, et al. The diagnostic accuracy of fecal calprotectin during the investigation of suspected pediatric inflammatory bowel disease. Am J Gastroenterol 2012; 107:941–949.
23. Kolho KL, Raivio T, Lindahl H, et al. Fecal calprotectin remains high during glucocorticoid therapy in children with inflammatory bowel disease. Scand J Gastroenterol 2006; 41:720–725.
24. Sidler MA, Leach ST, Day AS. Fecal S100A12 and fecal calprotectin as noninvasive markers for inflammatory bowel disease in children. Inflamm Bowel Dis 2008; 14:359–366.
25. Van De Vijver E, Schreuder AB, Muller-Kobold AC, et al. Safely ruling out IBD in children and teenagers without referral for endoscopy. Arch Dis Child 2012; 97:1014–1018.
26. Perminow G, Brackmann S, Lyckander LG, et al. A characterization in childhood inflammatory bowel disease, a new population-based inception cohort from South-Eastern Norway, 2005–2007, showing increased incidence in Crohn's disease. Scand J Gastroenterol 2009; 44:446–456.
27. Bossuyt PM, Reitsma JB, Bruns DE, et al. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD Initiative. Ann Intern Med 2003; 138:40–44.
28. The Cochrane Collaboration; 2011; Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Intervention.
29. Willis BH, Quigley M. Uptake of newer methodological developments and the deployment of meta-analysis in diagnostic test research: a systematic review. BMC Med Res Methodol 2011; 11:27.
30. Leeflang MM, Deeks JJ, Gatsonis C, et al. Systematic reviews of diagnostic test accuracy. Ann Intern Med 2008; 149:889–897.
31. Altman DG, Royston P. The cost of dichotomising continuous variables. BMJ 2006; 332:1080.
32. Olafsdottir E, Aksnes L, Fluge G, et al. Faecal calprotectin levels in infants with infantile colic, healthy infants, children with inflammatory bowel disease, children with recurrent abdominal pain and healthy children. Acta Paediatr 2002; 91:45–50.
33. Rugtveit J, Fagerhol MK. Age-dependent variations in fecal calprotectin concentrations in children. J Pediatr Gastroenterol Nutr 2002; 34:323–324.
34. Bachmann LM, Puhan MA, ter Riet G, et al. Sample sizes of studies on diagnostic accuracy: literature survey. BMJ 2006; 332:1127–1129.
35. Flahault A, Cadilhac M, Thomas G. Sample size calculation should be performed for design accuracy in diagnostic test studies. J Clin Epidemiol 2005; 58:859–862.
36. Bunn SK, Bisset WM, Main MJ, et al. Fecal calprotectin: validation as a noninvasive measure of bowel inflammation in childhood inflammatory bowel disease. J Pediatr Gastroenterol Nutr 2001; 33:14–22.
37. Anderson BJ, Merry AF. Data sharing for pharmacokinetic studies. Paediatr Anaesth 2009; 19:1005–1010.
38. Fischer BA, Zigmond MJ. The essential nature of sharing in science. Sci Eng Ethics 2010; 16:783–799.
39. Hrynaszkiewicz I, Norton ML, Vickers AJ, et al. Preparing raw clinical data for publication: guidance for journal editors, authors, and peer reviewers. BMJ 2010; 340:c181.
40. Terry SF, Terry PF. Power to the people: participant ownership of clinical trial data. Sci Transl Med 2011; 3:69cm3.
41. Vickers AJ. Making raw data more widely available. BMJ 2011; 342:d2323.
42. Walport M, Brest P. Sharing research data to improve public health. Lancet 2011; 377:537–539.
43. Irwig L, Macaskill P, Glasziou P, et al. Meta-analytic methods for diagnostic test accuracy. J Clin Epidemiol 1995; 48:119–130.
44. Rifai N, Altman DG, Bossuyt PM. Reporting bias in diagnostic and prognostic studies: time for action. Clin Chem 2008; 54:1101–1103.
45. Hart B, Lundh A, Bero L. Effect of reporting bias on meta-analyses of drug trials: reanalysis of meta-analyses. BMJ 2012; 344:d7202.
46. Greg Miller W, Myers GL, Lou Gantzer M, et al. Roadmap for harmonization of clinical laboratory measurement procedures. Clin Chem 2011; 57:1108–1117.
47. Klee GG. Establishment of outcome-related analytic performance goals. Clin Chem 2010; 56:714–722.
48. Klee GG, Schryver PG, Kisabeth RM. Analytic bias specifications based on the analysis of effects on performance of medical guidelines. Scand J Clin Lab Invest 1999; 59:509–512.
49. Lewis JD. The utility of biomarkers in the diagnosis and therapy of inflammatory bowel disease. Gastroenterology 2011; 140:1817.e2–1826.e2.
50. van Schaik FDM, Oldenburg B, Hart AR, et al. Serological markers predict inflammatory bowel disease years before the diagnosis. Gut 2013; 62:683–688.
Keywords:

adolescent; child; infant; inflammatory bowel diseases; leukocyte L1 antigen complex; meta-analysis

Supplemental Digital Content

© 2015 by European Society for Pediatric Gastroenterology, Hepatology, and Nutrition and North American Society for Pediatric Gastroenterology,