OUP user menu

Journal of Crohn's and Colitis: 10 (10)


Laurence J. Egan, Ireland

Associate Editors

Shomron Ben-Horin, IsraelSilvio Danese, ItalyPeter Lakatos, HungaryMiles Parkes, UKJesús Rivera-Nieves, USABritta Siegmund, GermanyGijs van den Brink, NLSéverine Vermeire, Belgium


Published on behalf of

C-reactive protein (CRP), erythrocyte sedimentation rate (ESR) or both? A systematic evaluation in pediatric ulcerative colitis

Dan Turner, David R. Mack, Jeffrey Hyams, Neal LeLeiko, Anthony Otley, James Markowitz, Yair Kasirer, Aleixo Muise, Cynthia H. Seow, Mark S. Silverberg, Wallace Crandall, Anne M. Griffiths
DOI: http://dx.doi.org/10.1016/j.crohns.2011.05.003 423-429 First published online: 1 October 2011


Background: There has not been an extensive comparison of CRP and ESR in ulcerative colitis (UC), and thus, we aimed to explore their utility in UC.

Methods: Four previously enrolled cohorts of 451 children with UC were utilized, all including laboratory, clinical and endoscopic data. A longitudinal analysis was performed on prospectively collected data of 75 children. Disease activity was captured by both global assessment and pediatric UC activity index (PUCAI).

Results: The best thresholds to differentiate quiescent, mild, moderate and severe disease activity, were < 23, 23–29, 30–37, > 37 mm/h for ESR, and < 2.5, 2.5–5, 5.01–9, > 9 mg/L for CRP (area under the ROC curves 0.70–0.81). Correlation of endoscopic appearance with CRP and ESR were 0.55 and 0.41, respectively (P < 0.001). Both CRP and ESR may be completely normal in 34% and 5–10% of those with mild and moderate-severe disease activity, respectively. Elevated CRP in the presence of normal ESR or vice versa was noted in 32%, 38%, 30% and 17% of those with quiescent, mild, moderate and severe disease activity. Over time, the utility of CRP and ESR in reflecting disease activity remained stable in 70–80% of cases.

Conclusion: In ~ 2/3 of children, both CRP and ESR values reflect disease activity to a similar degree and in the remaining, either CRP or ESR may be sufficient, with slight superiority of CRP. CRP is more closely correlated with endoscopic appearance. When either CRP or ESR performs well for a given patient, this is likely to remain so over time. Therefore, it may not be justified to routinely test both ESR and CRP in monitoring disease activity.

  • C-reactive protein
  • Erythrocyte sedimentation rate
  • Ulcerative colitis
  • Pediatrics
  • Disease activity

1 Introduction

Short sigmoidoscopic examination is a relatively easy procedure by which to assess mucosal inflammation in ulcerative colitis (UC) but this is poorly accepted, especially in children. Thus, UC biomarkers of inflammation are routinely used to complement the clinical assessment of disease activity. The most frequently monitored markers are C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR). Hemoglobin, albumin, leukocyte and platelet count may also reflect activity of colitis but to a lesser degree.1

Despite the widespread use of ESR and CRP alone and together in clinical practice, their utility has seldom been compared, nor has the value of assessing both parameters been examined. Previous studies have reported only on the independent prevalence of normal and elevated ESR or CRP values in children with varying UC activity.2,3 We thus aimed to review the diagnostic utility of CRP and ESR in reflecting UC activity in four large pediatric datasets, in order to assess whether the combination of ESR and CRP better reflects disease activity than either alone.

We evaluated the psychometric properties of ESR and CRP, using discriminative and construct validity testing, based on both physician global assessment and the Pediatric Ulcerative Colitis Activity Index (PUCAI), a validated non-invasive multi-item measure with well established cut-off scores for quiescent, mild, moderate and severe disease activity.4 In a subset of patients, colonoscopic assessment of colitis was also performed.

2 Methods

Four datasets of pediatric UC that included concurrent measurement of CRP, ESR, PUCAI scores, and physician global assessment (PGA) of disease activity were combined in this study: 1) The first was a prospective multicenter cohort, in which 157 children with UC were enrolled for a study that developed and validated the PUCAI as a multi-item measure of disease activity.4 In this cohort the scoring of the PGA was made before the results of the blood tests were known, thereby ensuring independence of measurements; 2) The second cohort came from a prospective multicenter study of outcomes in 128 children with severe pediatric UC, admitted for intravenous corticosteroid treatment5; 3) The third was a retrospective cohort of 99 children with severe colitis studied with the aim of evaluating predictors of response to intravenous corticosteroids.6 For the second and third cohorts in which children were hospitalized with severe UC, we used the assessments performed after three days of corticosteroid therapy so that disease activity would be more heterogeneous; 4) The fourth cohort was a prospectively acquired cohort of 76 children with UC (2–18 years) undergoing complete colonoscopy at the Hospital for Sick Children in Toronto for any reason (not yet published). Prior to colonoscopy, disease activity was scored using the PUCAI and PGA on both a 5-point Likert scale (remission, mild, moderate, severe and fulminant disease) and a 100 mm visual analog scale. Activity of macroscopic colitis was scored independently by a pediatric gastroenterologist (other than the one collecting the clinical data) as 0–3 for each of the colon segments (ascending, transverse, descending, sigmoid and rectum) using Beattie's grading system of quiescent, mild, moderate or severe.7 A final score (0–15) was calculated by adding all subscores.4,8,9 Children with proctitis only, an infrequent UC phenotype in children, were excluded from all cohorts.

Seventy five children from the first pediatric cohort, who were followed longitudinally, were used to evaluate the stability of CRP and ESR over time.4 Change in disease activity between visits was established using a 7-point global rating of change (i.e. ‘−3’ significantly worse, ‘0’ no change, and ‘+3’ significantly better) scored prospectively. Improvement was defined as “at least moderate” change.

All studies were conducted following approval by the institutional ethics committees, and after obtaining informed consents from the participants.

2.1 Analysis

Since we had the raw data of all datasets, we were able to combine the data of the individual patients, thereby increasing the power and reducing variability. Subgroup analyses were made within each dataset to explore for overt variation in the results, but none was found. Data are presented as mean ± standard deviation or median (interquartile range) as appropriate for the data normality. Continuous variables were compared using Wilcoxon rank sum test and the Kruskal Wallis test, as appropriate for the number of contrasting variables.

In order to establish the cutoff values of CRP and ESR that best correspond to mild, moderate and severe disease states, the entire dataset was split into 2/3 derivation and 1/3 validation sets, using the automatic function in the SPSS software. First, tentative values were determined based on extensive literature search2,1018 modified by the data distribution in the derivation cohort (examined using serial scatter plots and ROC curves). The cutoff values were then validated in the validation set using the same methodology. Pearson's or Spearman's correlation coefficients were used to evaluate construct validity, as appropriate for the data normality. Correlation rho value of ‘0–0.25’ was considered as lack of correlation, ‘0.25–0.5’ fair correlation, ‘0.5–0.75’ moderate to good correlation, and ‘>0.75’ very good to excellent correlation. Agreement between the four categories of CRP and ESR values and the four categories of disease activity were analyzed using the Kappa statistics with quadratic weights.

To explore the contribution of ESR and CRP relative to each other, and to account for possible confounding variables, multivariable regression models were constructed using PUCAI, and separately PGA, as the dependent variable. Variables were retained in the model based on maximizing R2. Assumptions of the model were verified by residual plots.

All comparisons were made using two sided significance levels of P < 0.05. Statistical analyses were performed using SPSS V12.0.

3 Results

Of the total 460 children included in the four datasets, 9 (2%) were excluded due to missing of CRP and ESR data. Disease activity of the remaining 451 children was sufficiently distributed between remission and severe disease, and the other phenotypic and demographic characteristics were as expected from a representative pediatric UC cohort (Table 1).

View this table:
Table 1

Baseline characteristics. Medians (interquartile range), mean (± SD) or counts (proportions) are presented as appropriate for the data distribution.

Entire cohort (n = 451)Derivation cohort (n = 301)Validation cohort (n = 150)
Males209 (46%)152 (50%)59 (39%)
Age (years)12.5 ± 3.812.4 ± 3.812.7 ± 3.9
Range (years)1.7–181.7–181.9–18
Weight (kg)44.7 ± 17.544.7 ± 17.944.6 ± 16.7
Height (cm)150.2 ± 22149.8 ± 22151 ± 21.6
Disease duration (years)0.7 (0–3.3)0.82 (0–3.4)0.5 (0–2.9)
First attack178 (39%)116 (39%)88 (59%)
Exacerbation273 (41%)185 (61%)62 (41%)
Disease extent
Left sided78 (17%)54 (18%)24 (16%)
Extensive373 (83%)247 (82%)126 (84%)
Disease severity (PUCAI)a
Remission84 (19%)63 (21%)21 (14%)
Mild79 (18%)53 (18%)26 (17%)
Moderate188 (41%)121 (40%)67 (45%)
Severe100 (22%)64 (21%)36 (24%)

P > 0.05 for all comparisons between the two subgroups.

  • a As previously defined.4

All evaluated blood values were significantly different between categories of disease activity (Table 2). Serial ROC curves were constructed to explore the best cutoff of CRP and ESR to differentiate the different disease activity. Reflecting internal validity, the results were similar between the derivation and the validation sets (Table 3). The best cutoff values to discriminate the disease activity states were then calculated on the combined cohort (Fig. 1). On average, CRP had slightly higher discriminant validity than ESR, as reflected by higher area under the ROC curves. However, when exploring Fig. 1, ESR was superior in differentiating remission from mild disease, whereas CRP was superior in differentiating moderate from severe disease.

Figure 1

Discriminant validity and best cutoff values to differentiate disease activity states of ESR (Fig. 1a) and C-reactive protein (Fig. 1b); both reflecting the combined derivation and validation cohorts. Disease activity states were defined by the PUCAI, as previously established.4 For simplicity, CRP values were truncated at 60 mg/L and ESR at 120 mm/h. Sens/Spec; sensitivity/specificity; AUROC; area under the ROC curve.

View this table:
Table 2

Median lab values (IQR) between disease activity states measured by physician global assessment.

Remission17 (8–23)1.1 (0.1–3.5)44 (43–47)128 (121–136)4.1 (3.2–4.9)322 (264–387)
Mild26 (18–40)1.3 (3–7.3)39 (34–43)111 (88–126)6.2 (1.2–9.9)417 (324–558)
Moderate37 (21–59)7.6 (4.5–20)35 (31–40)107 (89–125)8 (5.3–11.6)460 (347–593)
Severe39 (25–51)12.6 (7–20)33 (27–38)104 (89–119)7.6 (5–10.6)487 (386–613)
Fulminant45 (26–62)25 (2.7–85)32 (24–37)101 (83–110)7.5 (4–10.7)482 (336–642)
P valuea< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001
  • a Kruskal Wallis test comparing test results across the categories of disease activity.

View this table:
Table 3

Optimal cutoff values of CRP and ESR between disease activity states measured by the PUCAI.

CutoffAUC (95% CI)Sens/SpecCutoffAUC (95% CI)Sens/Spec
CRPmg/LRemission< 30.78 (0.72–0.85)70/68%< 2.50.79 (0.71–0.88)71/77%
Mild3–4.30.78 (0.73–0.83)78/69%2.5–4.40.86 (0.80–0.92)81/78%
Severe> 70.76 (0.70–0.83)71/72%> 90.89 (0.83–0.94)86/91%
ESR mm/hRemission< 230.74 (0.67–0.81)71/70%< 210.83 (0.76–0.91)76/79%
Mild23–300.72 (0.66–0.78)72/67%21–300.80 (0.73–0.87)81/70%
Severe> 370.68 (0.61–0.75)69/69%> 360.72 (0.63–0.82)67/69%

For number of children in each disease activity category, refer to Table 1.

  • a Three comparisons were made: remission vs. all, remission + mild vs. moderate + severe and severe vs. all (thus no test utility is presented for the “moderate” group).

Showing similar slight superiority of CRP, agreement between the four disease activity states (as determined by PUCAI scores) and the four subcategories of CRP was 0.53 (95% CI 0.45–0.62) measured by the weighted kappa, and 0.45 (0.36–0.53) for ESR; both reflecting moderate agreement.19 The results were similar when disease activity was classified according to PGA rather than the PUCAI (data not shown).

Next, we explored the ability of ESR and CRP to act in a complementary way in the individual patients. In an attempt to perform these combinations in a systematic way, we constructed tables of sensitivity and specificity including all options of ESR and/or CRP categories, for differentiating those with quiescent vs. active disease, those with quiescent and mild disease vs. moderate and severe disease and finally those with severe disease activity vs. all others (e.g. CRP < 2.5 mg/L and ESR < 23 mm, CRP < 2.5 mg/L or ESR < 23, CRP < 2.5 mg/L and ESR < 29 etc.). This list of sensitivities and specificities for each disease activity category was used to construct a summary ROC (SROC) curve, reflecting the overall diagnostic ability of the combined tests to reflect disease activity (Fig. 2). The area under the SROC curve was slightly superior to the areas under the ROC curves obtained by ESR alone (Fig. 1a) but not to the corresponding CRP values (Fig. 1b). Since this means that on average combining the results of ESR and CRP is not superior to CRP alone, we next evaluated specific scenarios that may be of clinical importance.

Figure 2

Summary ROC curve (SROC) of the different combination cutoffs of ESR and/or CRP (see text).

The number of children with normal CRP and ESR values, stratified by disease activity is shown in Fig. 3. Both tests were in the normal range (i.e. CRP < 2.5 mg/L and ESR < 23 mm) in 46 of 84 (55%) children in remission, 27 of 79 (34%) children with mild disease, 16 of 188 (9%) with moderate disease, and 5 of 100 (5%) with severe disease (P < 0.001). Of 100 children with severe disease 75 (75%) had CRP > 9 mg/L, and 68 (68%) had ESR > 37 mm/h. Elevated CRP (≥ 2.5 mg/L) in the presence of normal ESR (< 23 mm) or vice versa was noted in 32%, 38%, 30% and 17% of those judged to have quiescent, mild, moderate and severe disease activity (P > 0.1), implying that in one third of patients one test is important in view of non-informative nature of the other. Only 12 (15%) of those with quiescent or mild disease had CRP > 9 mg/L. In contrast, 104/288 children (36%) with moderate and severe disease activity had CRP < 5 mg/L. Of those 104 children, 61 (59%) had elevated ESR (> 29 mm/h), reducing the number of children with active disease in the presence of normal or near-normal inflammatory tests to 15%.

Figure 3

Rate of normal ESR (i.e. < 23 mm/h) and normal CRP (< 2.5 mg/L) values, stratified by disease activity states (measured by the PUCAI).

3.1 Validity

Table 4 shows correlation of ESR and CRP with various constructs of disease activity, benchmarked by other standard laboratory tests. In agreement with the previous results, CRP was superior to ESR in reflecting disease activity. Perhaps most importantly, the correlation of CRP with colonoscopic appearance was 0.55 (P < 0.001) compared to 0.41 (P < 0.001) with ESR. Medians of both tests were significantly higher in children with extensive disease (n = 372, 82%) vs. left sided disease (n = 79, 18%): 6.2 (IQR 2–12.6) vs. 2.1 (0.5–5.7) for CRP (P < 0.001) and 33 (20–48) vs. 21 (13–38) for ESR (P = 0.001).

View this table:
Table 4

Spearman's correlation of common blood results with constructs of disease activity in pediatric ulcerative colitis (*P < 0.05).

Colonoscopic scorea0.41*0.55*− 0.56*0.48*− 0.26*0.34*0.27*
PGAb0.46*0.61*− 0.66*0.45*− 0.39*0.45*0.31*
Mayo score36a0.41*0.43*− 0.57*0.26−
PUCAI score40.46*0.61*− 0.68*0.50*− 0.43*0.46*0.35*
Lichtiger index370.39*0.34*− 0.48*0.28−
Average0.430.51− 0.590.39−

PGA, physicians' global assessment; PUCAI, pediatric ulcerative colitis activity index; ESR, erythrocyte sedimentation rate; CRP, C-reactive protein; Hb, hemoglobin.

  • a Available on the colonoscopy cohort only (n = 76); colonoscopic score was performed according to Beattie et al.7 in each colonic segment, as previously described.4

  • b Scored on a 100 mm visual analog scale.

3.2 Longitudinal analysis

Of the total 75 children in the longitudinal datasets, 56 (75%) and 54 (72%) had CRP and ESR values, respectively, within no more than one category deviation from the corresponding disease activity category, at the baseline visit (using the four categories of ESR and CRP values as in Fig. 1 and the four disease activity categories of remission, mild, moderate and severe disease activity based on PUCAI scores). For instance, CRP or ESR values in their lowest category were considered appropriate in the presence of mild disease activity, or remission. A total of 44/75 (59%) had both ESR and CRP within the appropriate category at baseline, 12/75 (16%) had only CRP in the appropriate category, 10/75 (13%) had only ESR in the appropriate category, and 9/75 (12%) had neither CRP nor ESR in the appropriate category.

At the follow-up visit, 46/56 (82%) still had appropriate CRP levels, compared with 42/54 (78%) for ESR. Only two of the 10 children with an appropriate ESR value in the presence of inappropriate CRP values at baseline reversed this finding at the follow-up visit (i.e. appropriate CRP with inappropriate ESR). Similar exchange between appropriate CRP to appropriate ESR was not observed in any of the 12 appropriate CRP values at baseline (0%).

Taken together this means that although both ESR and CRP were moderately stable over time (~ 70–80%), there was little shift between the utility of CRP and ESR in a given patient. When either CRP or ESR performs well within a given patient, this is likely to remain so over time.

Of the 29 children who were judged by the global rating of change to remain stable between the two visits, 28 (97%) and 20 (69%) had also stable CRP and ESR, respectively, within one category from baseline. Of the 31 improved children, 4 (13%) and 10 (32%) improved their CRP category or remained within the same category, respectively. The corresponding proportions for ESR were 1 (3%) and 16 (52%). Of the 15 patients who deteriorated clinically, 9 (60%) worsened their CRP category and 6 (40%) remained stable. The corresponding proportions for ESR were 7 (47%) and 6 (40%) patients.

3.3 Multivariate analysis

In agreement with the above results, the multivariate analyses indicated that CRP and ESR are complementary for reflecting disease activity on average (as opposed to within a given individual) with slight superiority of CRP.

Both ESR and CRP were highly significant in a model including only these two as the explanatory variables (both P < 0.0001). Removing CRP from the model decreased the adjusted R2 from 0.25 to 0.16; removing ESR from the model decreased the R2 to 0.17, suggesting that both tests were complementary to each other to some extent. Next, a model with all standard blood tests was constructed, including albumin (P < 0.001), CRP (P < 0.001), ESR (P = 0.035), platelets (P = 0.15), hemoglobin (P = 0.49) and absolute neutrophil count (ANC) (P = 0.001) (adjusted R2 = 0.51). Removing CRP from the model reduced R2 slightly to 0.48 but increased the standardized β value of ESR by ~ 50% (from 0.085 to 0.125) whereas removing ESR reduced R2 only to 0.50 and changed the standardized β value of CRP by only ~ 10% (from 0.17 to 0.187). No collinearity was noted in any of the models (all variance inflation factors (VIF) < 2.2), likely due to the large sample size.

4 Discussion

We have extensively analyzed the largest combined pediatric cohort to date, in order to evaluate whether both CRP and ESR are required to monitor disease activity in UC. We utilized several complementary analytic approaches including diagnostic utility tests, multivariate analysis, summary ROC curves, construct validation and longitudinal analysis.

Five main conclusions may be drawn from these analyses:

  1. Both CRP and ESR have only fair correlation with colonoscopic inflammation (and other constructs of disease activity), with a slight superiority to CRP.

  2. For the entire population studied on average, CRP is slightly superior to ESR in reflecting and differentiating disease activity states and monitoring change over time, especially at the more severe end of the spectrum.

  3. Once it is determined which study is associated with disease activity in a particular patient, for most patients (~ 85%), ESR or CRP alone is sufficient to accurately reflect disease activity.

  4. The accuracy of either CRP or ESR over time is stable in 70–80%, and rarely does the superiority of one test replace the other.

  5. In agreement with previous studies,20 CRP and ESR values were higher in the presence of extensive disease. It is therefore important to focus the analysis on children in whom UC is twice as often extensive as compared with adults.21,22

CRP is produced by hepatocytes in response to IL-6 but also IL-1β and TNFα.23 Our findings in UC are unlike in CD, wherein CRP seems to be clearly superior in performance to ESR.2,14,15,23 Several clinical trials of biologics in CD indicate the importance of high CRP values at baseline in predicting response to therapy.2426

Other studies have assessed blood markers to differentiate disease activity in UC. The combination of serum orosomucoid and ESR was found to be complementary in assessing disease activity in IBD children, while CRP was of no additional value.27 ESR correlated with physicians' global assessment of disease activity in 77 UC and CD patients.28 ESR, CRP, albumin, WBC, and platelets were significantly different between endoscopically quiescent or active UC, but CRP had the best combination of sensitivity (74%) and specificity (70%).10 CRP was correlated with endoscopic score in 21 UC patients, although less closely than did serum albumin and fecal alpha-1-antitrypsin.29 In contrast, Gomes et al. found no correlation between colonoscopic appearance and any laboratory values, including CRP and ESR.9 Furthermore, CRP (at a cutoff of 5 mg/L) and ESR (15 mm/hr) had poor sensitivity (69% for both) and specificity (62% and 65%, respectively) to differentiate UC patients in remission from those with active disease judged by the Mayo score.12 CRP and ESR were found to be good predictors of response to therapy in severe UC in some,5,6,3033 but not in all studies.34,35 The heterogeneity in results may reflect study design, including the presence of small cohorts and often retrospectively collected data, much different from our large combined prospectively acquired cohort. Nonetheless, it is clear from previous studies and from our results that CRP, ESR and other serum (or even fecal) biomarkers do not reflect closely enough mucosal inflammation and the search for an accurate non-invasive and inexpensive marker that can serve as a surrogate for mucosal healing continues.

This study is not without limitations. The longitudinal cohort had small sample size and CRP and ESR were measured in each site without global standardization. The choice of the third hospital day as the sampling day may have prioritized CRP given its higher responsiveness as compared with ESR, but the better apparent performance of CRP in our analysis rightly accounts for this advantage of CRP. Overall, however, the consistent results found here across several cohorts, using different analytic approaches and constructs of disease activity, strengthen our aforementioned conclusions which may be cautiously translated into clinical practice recommendations. Specifically, both CRP and ESR may be measured during the first few clinic visits of a newly diagnosed UC patient. Subsequently, only one test, ESR or CRP, may need to be measured, with choice based on the observed performance in the individual patient. On average, CRP is slightly superior to ESR in pediatric UC.

Conflict of interest statement

None of the authors have any conflict of interest to declare in relation to this manuscript. The study was not funded and has no financial sponsor.


DT designed the study, acquired data, performed the analysis and wrote the first draft manuscript; DRM, JH, NL, AO, JM, YK, AM, CHS, MSS, and WC acquired data, commented on the design and analysis and approved the final manuscript; AMG designed the study, acquired data and revised the manuscript.


View Abstract