- Top of page
Background and objective: Parapneumonic effusions (PPE) that require drainage are referred to as complicated parapneumonic effusions (CPPE). Following resolution of these effusions, residual pleural thickening (RPT) may persist. We hypothesize that the concentrations of CRP in pleural fluid (CRPpf) and serum (CRPser) can be used to identify CPPE and to predict RPT.
Methods: All patients with non-purulent PPE, who were admitted to two tertiary hospitals during a 30-month period, were enrolled in the study. Baseline CRPpf and CRPser levels were compared between patients with complicated or uncomplicated PPE, as well as between patients with or without RPT of >10 mm, 6 months after discharge from hospital. Cut-off values for identification of CPPE and prediction of RPT were determined by receiver operating characteristic curve analysis. Logistic regression analysis was performed to assess the association between CRP levels and RPT.
Results: Fifty-four patients were included in the study. Patients with CPPE (n = 23) had significantly higher levels of both CRPpf and CRPser than those with uncomplicated PPE. For identification of CPPE, a CRPpf level >78.5 mg/L and a CRPser level >83 mg/L gave 84% and 47% sensitivity, with 65% and 87% specificity, respectively. Classical criteria (pleural fluid pH <7.20, LDH >1000 IU/L, glucose <600 mg/L) were superior for this purpose. A combination of classical biomarkers with CRP levels using an ‘AND’ or ‘OR’ rule improved the positive and negative predictive values, respectively. CRPser was an independent predictor for development of RPT (adjusted OR 1.18). A CRPser level >150 mg/L had 91% specificity and 61% sensitivity for prediction of RPT.
Conclusions: This study demonstrated the value of CRPser for prediction of RPT in patients with PPE. Moreover, when used in combination with classical biomarkers, CRP levels may be a useful adjunct for decision-making in relation to treatment of patients with non-purulent PPE.
- Top of page
Any pleural effusion associated with bacterial pneumonia, lung abscess or infected bronchiectasis is designated as a parapneumonic effusion (PPE).1 Approximately 40% of patients hospitalized with pneumonia have an accompanying pleural effusion.1,2 Although most PPE resolve completely with antibiotic treatment alone, ∼10% of cases require pleural fluid drainage and are referred to as complicated PPE (CPPE).1 During the recovery phase, pleural fibrosis due to severe pleural inflammation, and manifesting as residual pleural thickening (RPT), develops in ∼14% of all patients with PPE.1,3
While empyema always requires drainage, the decision to drain non-purulent PPE is currently based on the anatomical (size and loculations), biochemical (pH <7.20, LDH >1000 IU/L and glucose <600 mg/L) and bacteriological (Gram stain and culture) characteristics of the effusion.4–6 Pleural fluid pH (pHpf) <7.20 is considered the most accurate criterion for drainage, although specificity is <100%, which means that a number of PPE with low pHpf may resolve without operational intervention.6,7 In addition, measurement of pHpf is highly dependent on the method of sample collection, and evaluation of pHpf is problematic in patients with systemic acidosis or Proteus infections.1,8 On the other hand, clinical features including purulent pleural fluid, increased severity of pneumonia, temperature ≥38°C and delayed resolution after diagnosis (>15 days) have shown to be associated with RPT.3,9
CRP is an inflammatory marker that is synthesized by hepatocytes in response to pro-inflammatory cytokines such as IL-6 and tumour necrosis factor-α.10 The concentrations of CRP in serum (CRPser) and pleural fluid (CRPpf) have shown to be useful for assessing the severity of pneumonia, discriminating infectious from malignant effusions and predicting the likelihood that patients with pneumonia will develop pleural infections.11–15 Furthermore, experimental pleural fibrosis has shown to result in increased CRPser levels in rabbits.16
The aim of this study was to compare CRPser and CRPpf levels with biomarkers currently used for the identification of non-purulent PPE requiring drainage, to assess the performance of CRPser and CRPpf in combination with biomarkers currently used for the same purpose, and to investigate their usefulness for prediction of RPT.
- Top of page
During the 30-month period of the study, 57 patients were enrolled; three patients (5%) died due to sepsis before pleural fluid drainage and were excluded, as the adverse outcome could be explained by the progression of pneumonia, regardless of pleural fluid characteristics. The mean age of the remaining 54 patients was 64 ± 11 years, with 38 (70%) being males. Results for pH, LDH, glucose, CRPpf and CRPser were missing for 5 (9%), 5 (9%), 7 (12%), 3 (5%) and 4 (7%) patients, respectively, due to technical reasons such as clotting of pleural fluid or inadequate quantity of specimen. The upper normal limit for serum LDH in our laboratory was 225 IU/L. Results for both CRPser and CRPpf were available for 47 patients; the correlation between CRPser and CRPpf in these patients was weak, although it was statistically significant (P = 0.006, r2 = 0.19).
Identification of patients with complicated parapneumonic effusion
Twenty of the 54 patients (37%) had a CPPE. The characteristics of the patients with complicated or uncomplicated PPE are shown in Table 1. Both CRPpf (P = 0.002) and CRPser (P = 0.02) levels were significantly higher in patients with CPPE. There was no significant difference in CRPpf and CRPser levels between patients with CPPE that required thoracoscopy/thoracotomy and those that were effectively treated with chest tubes (P = 0.42, data not shown).
Table 1. Characteristics of the patients with complicated or uncomplicated parapneumonic effusions
|Characteristic||n||CPPE (n = 20)||UCPPE (n = 34)||P value|
|Age, years†||54||60 ± 10||69 ± 11||0.01*|
|Males, n (%)||54||15 (75)||23 (68)||0.76|
|Comorbidities, n (%)||21||6 (30)||15 (44)||0.79|
| Cardiovascular disease||9||1 (5)||8 (24)||0.50|
| Neuropsychiatric disease||2||1 (4)||1 (3)||1.00|
| Gastrointestinal disease||6||1 (5)||5 (15)||0.39|
| Diabetes mellitus||4||2 (9)||2 (6)||1.00|
| Substance abuse||2||1 (4)||1 (3)||1.00|
|NCCpf × 109/L‡||46||12.25 (3.1–35.5)||6.7 (2.3–10.95)||0.07|
|Neutrophilspf, n‡||46||9588 (2848–29 695)||4690 (1240–8285)||0.03*|
|pHpf‡||46||6.81 (6.37–7.00)||7.39 (7.34–7.42)||<0.0001*|
|LDHpf, IU/L‡||46||2803 (1411–8600)||495 (366–624)||<0.0001*|
|Proteinpf, g/L†||42||44.6 ± 12.4||44.0 ± 10.6||0.86|
|Glucosepf, mg/L†||46||344.0 ± 436.2||1080 ± 491.4||<0.0001*|
|CRPpf, mg/L†||48||96.8 ± 53.3||51.8 ± 44.6||0.002*|
|CRPser, mg/L†||47||188.9 ± 104.0||127.3 ± 88.8||0.03*|
ROC curves were constructed to evaluate the accuracy of CRPpf and CRPser for identifying CPPE. CRPpf (area under curve (AUC): 0.73) and CRPser (AUC: 0.68) were both less accurate than all other biomarkers, while pHpf (AUC: 0.93) showed the best performance for that purpose (Table 2). With cut-off values of >78.5 mg/L for CRPpf and >83 mg/L for CRPser, giving 84% and 47% sensitivity, with 65% and 87% specificity, respectively, these parameters were found to be inferior to the classical parameters for identification of CPPE. The most accurate criterion for this purpose was pHpf <7.20, which gave 91% sensitivity and 93% specificity with a positive predictive value (PPV) of 91% and a negative predictive value (NPV) of 93%. The combination of classical criteria with CRP levels, using an ‘AND’ or ‘OR’ rule, improved the positive and negative predictive values, respectively (Table 3). A CRPpf level >78.5 mg/L with a pleural fluid glucose concentration <600 mg/L had a PPV of 100%, while a CRPser level >83 mg/L with pH <7.20 and a CRPser level >83 mg/L or a pleural fluid glucose concentration <400–600 mg/L gave a NPV of 100%.
Table 2. Receiver operating characteristic curve analysis of the accuracy of biomarkers for identification of complicated parapneumonic effusions
|Biomarker||Cut-off value||Sensitivity, % (95% CI)||Specificity, % (95% CI)||LR (+)||LR (−)||AUC (95% CI)|
|pHpf||7.248||90 (73–98)||95 (75–100)||17.93||0.11||0.93 (0.85–1.00)|
|LDHpf||828 IU/L||83 (64–94)||89 (67–99)||7.86||0.19||0.90 (0.80–0.99)|
|Glucosepf||660 mg/L||93 (76–99)||85 (62–97)||6.17||0.08||0.90 (0.78–1.00)|
|CRPpf||78.5 mg/L||84 (66–95)||65 (41–85)||2.40||0.25||0.74 (0.60–0.88)|
|CRPser||83 mg/L||47 (28–66)||85 (62–97)||3.11||0.62||0.69 (0.54–0.84)|
Table 3. Performance characteristics of classical biomarkers for the identification of complicated parapneumonic effusions when used individually or in combination with serum and pleural fluid levels of C-reactive protein
|Biomarker||Sensitivity, %||Specificity, %||LR (+)||LR (−)||PPV, %||NPV, %|
|pHpf <7.20||90 (68–99)||93 (77–99)||13.05 (4.01–54.79)||0.11 (0.03–0.33)||90 (68–99)||93 (77–99)|
|LDHpf >1000 IU/L||85 (62–97)||86 (68–96)||6.16 (2.57–14.04)||0.17 (0.05–0.45)||81 (58–95)||89 (72–98)|
|Glupf <600 mg/L||80 (56–94)||96 (81–100)||21.60 (4.00–430.5)||0.21 (0.15–0.43)||94 (71–100)||87 (69–96)|
|CRPpf >78.5 mg/L AND|
|pHpf <7.20||60 (36–81)||93 (77–99)||8.70 (2.25–53.89)||0.43 (0.32–0.71)||86 (57–98)||77 (60–90)|
|LDHpf >1000 IU/L||55 (32–77)||90 (73–98)||5.32 (1.65–22.23)||0.50 (0.35–0.81)||79 (49–95)||74 (57–88)|
|Glupf <600 mg/L||45 (23–68)||100 (87–100)||—||0.55 (0.55–0.79)||100 (66–100)||71 (54–85)|
|CRPser >83 mg/L AND|
|pHpf <7.20||80 (56–94)||97 (82–100)||23.20 (4.28–462.4)||0.21 (0.15–0.43)||94 (71–100)||88 (71–96)|
|LDHpf >1000 IU/L||75 (51–91)||90 (73–98)||7.25 (2.51–27.07)||0.28 (0.15–0.56)||83 (59–96)||84 (66–95)|
|Glupf <600 mg/L||65 (41–85)||100 (87–100)||—||0.35 (0.35–0.57)||100 (75–100)||79 (62–91)|
|CRPpf >78.5 mg/L OR|
|pHpf <7.20||95 (75–100)||83 (64–94)||5.50 (2.70–7.14)||0.06 (0.003–0.31)||79 (58–93)||96 (80–100)|
|LDHpf >1000 IU/L||95 (75–100)||79 (60–92)||4.59 (2.39–5.72)||0.06 (0.003–0.33)||76 (55–91)||96 (79–100)|
|Glupf <600 mg/L||95 (75–100)||81 (62–94)||5.13 (2.52–6.64)||0.06 (0.003–0.32)||79 (58–93)||96 (78–100)|
|CRPser >83 mg/L OR|
|pHpf <7.20||100 (83–100)||41 (24–61)||1.71 (1.21–1.71)||0.00 (0–0.52)||54 (37–71)||100 (74–100)|
|LDHpf >1000 IU/L||95 (75–100)||41 (24–61)||1.62 (1.11–1.80)||0.12 (0.006–0.74)||53 (35–70)||92 (64–100)|
|Glupf <600 mg/L||100 (83–100)||37 (19–58)||1.59 (1.14–1.59)||0.00 (0–0.60)||54 (37–71)||100 (69–100)|
Prediction of residual pleural thickening
Eleven of the 54 surviving patients (20%) were found to have RPT of >10 mm; the details of these patients are shown in Table 4. The requirement for drainage during the acute phase of the PPE occurred significantly more frequently among patients who finally developed RPT. The biomarkers that differed significantly between patients with or without RPT were CRPser and pleural fluid LDH, pH and glucose. To evaluate the accuracy of these parameters for predicting RPT, ROC curves were constructed. Analysis of these ROC curves showed that CRPser was the most accurate biomarker for predicting RPT (Table 5). The results of the univariate analysis for factors significantly associated with RPT and the final multivariate logistic regression model are shown in Table 6. CRPser and pleural fluid glucose were the only factors independently associated with RPT.
Table 4. Characteristics of the patients with or without residual pleural thickening
|Characteristic||n||With RPT (n = 11)||Without RPT (n = 43)||P value|
|Age, years†||54||65 ± 14||66 ± 11||0.32|
|Males, n (%)||54||9 (82)||31 (67)||0.35|
|Comorbidities, n (%)||24||2 (18)||22 (51)||0.09|
| Cardiovascular disease||11||1 (9)||10 (23)||0.43|
| Neuropsychiatric disease||2||1 (9)||1 (2)||0.37|
| Gastrointestinal disease||7||0 (0)||7 (16)||0.32|
| Diabetes mellitus||4||0 (0)||4 (9)||0.57|
| Substance abuse||2||0 (0)||2 (5)||1.00|
|Drainage, n (%)||20||8 (73)||12 (28)||0.01*|
| Thoracostomy||18||5 (45)||12 (28)||0.48|
| Thoracoscopy||1||1 (9)||0 (0)||0.20|
| Thoracotomy||1||1 (9)||0 (0)||0.20|
|NCC × 109/L‡||52||11.4 (2.05–39)||6.95 (2.45–13.25)||0.30|
|Neutrophils, n‡||52||7600 (1353–33 150)||5850 (1684–11 013)||0.34|
|pHpf,‡||52||6.96 (6.70–7.29)||7.37 (6.87–7.42)||0.02*|
|LDHpf, IU/L‡||52||2810 (1426–8600)||555 (410–26 640)||0.003*|
|Proteinpf, g/L†||46||47.3 ± 7.5||42.7 ± 13.0||0.37|
|Glucosepf, mg/L†||50||442.8 ± 434.1||858.0 ± 612.6||0.04*|
|CRPpf, mg/L†||54||93.5 ± 53.9||65.1 ± 51.1||0.11|
|CRPser, mg/L†||53||247.7 ± 95.7||127.7 ± 82.88||0.0002*|
Table 5. Receiver operating characteristic curve analysis of the accuracy of biomarkers for prediction of residual pleural thickening in patients with parapneumonic effusions
|Biomarker||Cut-off value||Sensitivity, %||Specificity, %||LR (+)||LR (−)||AUC|
|pHpf||7.36||53 (36–69)||91 (59–100)||5.79||0.52||0.74 (0.61–0.88)|
|LDHpf||687 IU/L||63 (46–78)||91 (59–100)||6.95||0.41||0.80 (0.67–0.93)|
|Glucosepf||570 mg/L||75 (58–88)||73 (39–94)||2.75||0.34||0.69 (0.51–0.86)|
|CRPpf||68 mg/L||68 (51–83)||73 (39–94)||2.51||0.44||0.66 (0.47–0.84)|
|CRPser||150 mg/L||61 (43–76)||91 (59–100)||6.66||0.43||0.82 (0.69–0.95)|
Table 6. Univariate and multivariate analyses for variables associated with residual pleural thickening in patients with parapneumonic effusions
|Variable||OR||95% CI||P value|
|Variable||Adjusted OR||95% CI||P value|
- Top of page
The purpose of this study was to evaluate CRPser and CRPpf as predictors of RPT in patients with PPE, and to assess the usefulness of these parameters for the identification of non-purulent PPE that require drainage. The results from 54 patients suggest that CRP is a prognostic marker in PPE. More specifically, CRPser was not only significantly higher in patients who finally developed RPT but also the most accurate of the biomarkers that were associated with RPT. Furthermore, CRPser was significantly higher in the patients with PPE that required drainage; the same was also true for CRPpf. Both CRPser and CRPpf predicted the requirement for drainage less accurately than classical biomarkers. However, the combination of classical biomarkers with CRP levels using an ‘AND’ or ‘OR’ rule improved the positive and negative predictive values, respectively.
The prevalence of RPT in this study (20%) was comparable with that reported in previous studies (14–17%).3,9 A CRPser level >150 mg/L gave 61% sensitivity and 91% specificity for predicting RPT, whereas CRPpf was not useful for this purpose. CRPser, but not CRPpf, has previously shown to be useful for prediction of RPT in patients with tuberculous pleural effusions.23 This is probably because serum levels directly reflect liver production of this protein. In contrast, pleural fluid levels are influenced by the permeability of the pleura, which is generally increased or decreased in the presence of pleural inflammation or fibrosis, respectively.24 This is most likely the reason for the weak correlation between CRPser and CRPpf and the fact that CRPpf was not a good predictor of RPT.
The requirement for drainage of PPE is unequivocally mandatory when pus is aspirated from the pleural cavity or when bacteria are isolated from non-purulent pleural fluid.1 Considering the low sensitivity (20–40%) of pleural fluid cultures, however, the decision whether to initiate drainage is difficult in the majority of patients with non-purulent PPE.25 Classical criteria supporting this decision include pleural fluid pH <7.20, LDH >1000 IU/L and glucose <600 mg/L.4,6,7 Although current guidelines suggest that pH <7.20 is the most accurate criterion for this purpose, a study by Jimenez et al. of 238 patients with PPE showed that some PPE with a pH of 7.20–7.37 required drainage, while 33% of the PPE with a pH of 7.00–7.20 resolved with antibiotic treatment alone.26 In the present study, a pH <7.20 showed the best performance, giving 91% sensitivity and 93% specificity, with a NPV of 93%. However, if the decision to initiate drainage was based solely on pH, approximately one in 10 patients would be subjected to unnecessary interventions, while about the same proportion would suffer sepsis due to delayed drainage of pleural fluid.
This study demonstrated the value of CRPser for the identification of non-purulent complicated PPE that require drainage. Previous studies have evaluated the use of CRPpf for the same purpose. In two studies of patients with non-purulent PPE, Porcel et al. reported that a CRPpf level >80 mg/L (AUC: 0.79 and 0.81) identified CPPE with a sensitivity of ∼70% and a specificity of ∼73%, respectively.14,15 Furthermore, Chen et al. reported that a CRPpf level >87 mg/L (AUC: 0.94) gave 80% sensitivity and 97% specificity for identification of CPPE.27 However, the much greater accuracy of CRPpf as reported in that study probably resulted from the inclusion of patients with purulent PPE, which is not clinically relevant.4,6
To compare CRP with classical criteria in the present study, the decision to initiate drainage was based on the attending physician's experience. For this purpose, a CRPser level >83 mg/L and a CRPpf level >78.5 mg/L gave 47% and 84% sensitivity, with 87% and 65% specificity, respectively; both these parameters were inferior to classical criteria. However, the combination of classical criteria with CRP, using an ‘AND’ or ‘OR’ rule improved the PPV and NPV, respectively. This combination of criteria may be useful in ambiguous cases, where the values for classical biomarkers are close to the recommended cut-off values. For example, in PPE with a pH >7.20 but <7.30, a CRPser level <83 mg/L would identify the effusions that would definitely resolve with antibiotic treatment alone (NPV 100%). On the other hand, a CRPser level >83 mg/L in PPE with pleural fluid glucose concentrations between 300 mg/L and 400 mg/L would definitely identify the effusions requiring drainage (PPV 100%).
This study had limitations. First, the small number of patients included in the study may have limited its power. Second, the fact that the decision to initiate drainage was based on the attending physician's experience may have biased the results. As mentioned previously, there is currently no ‘gold standard’ to aid the decision whether to initiate drainage in patients with non-purulent PPE. Moreover, the possibility that the attending physicians might have taken pleural fluid pH and glucose concentrations into account when assessing the requirement for drainage may have further biased the findings from this study population (incorporation bias). Third, it could be argued that detection of RPT on CXR should not be used as an outcome, as it does not guarantee that the patient has a restrictive lung function abnormality. Pulmonary function tests would be the ideal approach to evaluating functional impairment at 6 months, but these were not performed. However, RPT of >10 mm on CXR has previously shown to give 96% specificity for prediction of restrictive abnormalities in patients with tuberculous pleural effusions, which means that all patients with RPT in the present study would most likely present with a restrictive pattern on pulmonary function testing.22 Finally, chest CT would probably be more accurate for detection and measurement of RPT but was not performed because it is associated with high levels of radiation exposure.
In conclusion, this study demonstrated the value of CRPser for prediction of RPT and the requirement for drainage in patients with non-purulent PPE. The study also showed that the combination of classical criteria with either CRPser or CRPpf could be a useful adjunct for decision-making in relation to treatment of patients with non-purulent PPE.