Is procalcitonin superior to CRP and ESR in the diagnosis of diabetic foot osteomyelitis? A systematic review and meta-analysis

Abstract

Background

In this study, we aimed to assess the diagnostic accuracy of inflammatory biomarkers in clinical practice to distinguish diabetic foot osteomyelitis (DFO) from soft tissue infection in diabetes.

Methods

Electronic databases were searched for relevant studies on the diagnosis of DFO using CRP, ESR, or PCT until Sep 2023. We pooled sensitivity, specificity, PLR, NLR, DOR, and ROC-AUC to assess the diagnostic value of biomarkers for DFO. We evaluated the included study with the QUADAS tool. The sensitivity analysis and publication bias were assessed.

Results

A total of 14 studies with 1698 patients were included in this meta-analysis. The combined sensitivity was 0.78 (95% CI: 0.66–0.87), specificity was 0.72 (95% CI: 0.64–0.78), and AUC was 0.79 for CRP diagnosis. The combined sensitivity was 0.77 (95% CI: 0.67–0.85), specificity was 0.74 (95% CI: 0.65–0.81), and AUC was 0.82 for ESR diagnosis. The combined sensitivity was 0.88 (95% CI: 0.63–0.97), specificity was 0.81 (95% CI: 0.63–0.92), and AUC was 0.91 for PCT.

Conclusion

This study evaluated the role of ESR, CRP, and PCT in the diagnosis of DFO, and found that PCT had the best diagnostic test accuracy in distinguishing DFO from soft tissue infection in diabetes.

Clinical trial

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-026-13071-6.

Diabetic foot osteomyelitis (DFO) is a serious complication of diabetic foot disease and constitutes the main cause of lower limb amputation in patients with diabetic foot ulcers [1]. It affects 50%-60% of patients with severe diabetic foot infection and approximately 20% of those with moderated infection [2]. Although bone histopathology and microbial culture remain the current gold standard for diagnosing DFO [3], their invasive nature underscores the need for more accessible diagnostic approaches. According to the Guidelines on the diagnosis and treatment of diabetes-related foot infection (IWGDF/IDSA 2023), a combination of the probe-to-bone test, serum inflammatory markers, and plain X-rays is recommended as a preliminary study for diagnosing foot osteomyelitis [4]. The typical radiographic signs of osteomyelitis include osteolysis, bone destruction, and periosteal reaction, which usually only manifest until a later stage [5]. Biomarkers of infection could assist in earlier and noninvasive diagnosis, especially when clinical symptoms are misleading [6].

Conventional biomarkers such as erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) are elevated in a wide range of inflammatory processes. Procalcitonin (PCT), a 116-amino acid precursor of the hormone calcitonin, has recently emerged as a promising biomarker for diagnosing infections [7]. Although several studies have investigated the role of PCT in diagnosing DFO [8], evidence regarding its superiority over established biomarkers remains limited. This study aimed to evaluate the diagnostic accuracy of commonly used inflammatory biomarkers in distinguishing DFO from soft tissue infection in clinical practice, and to directly compare the performance of PCT with that of traditional markers such as ESR and CRP.

Methods

Search strategies and study selection

This meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (S1) [9]. PROSPERO registration number was CRD42023487628. Two reviewers independently performed a systematic literature search of several electronic databases, including MEDLINE, BIOSIS, Embase, the Cochrane Central Register of Controlled Trials (CENTRAL), and Web of Science, to identify relevant studies published up to September 2023. No language restrictions were applied. Additional studies were sought from the reference lists of the selected articles and other recent reviews. The following keywords including ‘diabetic foot’, ‘osteomyelitis’, ‘infection’, ‘biomarker’, ‘CRP’, ‘C-reactive protein’, ‘ESR’, ‘erythrocyte sedimentation rate’, ‘PCT’, ‘procalcitonin’, and ‘diagnosis’ were used for the search terms. The concrete search policy was illustrated in S2 file. The search strategy was designed and refined, and any differences between two reviewers were dealt by a third reviewer.

Inclusion and exclusion criteria

Literature inclusion criteria: (1) prospective or retrospective studies that investigated inflammatory serum biomarkers for the diagnosis of DFO; (2) inflammatory serum biomarkers including CRP, ESR, and/or PCT were reported; (3) valid data available in the study. The exclusion criteria were (1) abstracts, reviews or letters; (2) animal models or basic studies; and (3) studies reported incomplete data or missing data which was not possible to extract usable data for the meta-analysis. Two reviewers independently determined whether the retrieved literature could be included in the study. If there was disagreement, a third reviewer wound make an independent judgment.

Data extraction

The extracted data included first author’s name, year of publication, country, study design, number of patients, sex, age, biomarker cut-off value, and outcome measures (sensitivity and specificity). Two reviewers extracted all data independently, and all disagreements were resolved by a third reviewer.

Study quality assessment

We used the Quality Assessment of Diagnostic Accuracy Studies (QUADAS)-2 tool provided by the Cochrane Collaboration system to assess the quality of the selected studies [10]. This tool evaluates the four biases in terms of patient selection, index test (pre-specified values), reference standard, and flow and timing. For each domain, the risk of bias was judged as ‘low’, ‘high’, or ‘unclear’.

Data synthesis and statistical analysis

This meta-analysis was performed using Stata 14.1 and Meta-Disc 1.4. We evaluated the threshold effect with the use of Meta-Disc software. Assessment criterion was the Spearman correlation coefficient, and a p-value > 0.05 indicated that no threshold impact. Pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR) were calculated for CRP, ESR, and PCT. Heterogeneity among studies was assessed using the Chi-square test and I² statistic. The random effects model was used when heterogeneity exceeded 50%. We also obtained the summary receiver operating characteristic (ROC) curve, and calculated the 95% confidence interval of the area under the curve (AUC 95%). If there was significant heterogeneity, a subgroup and meta-regression analysis was conducted to explore potential sources. We performed sensitivity analyses by analyzing one study at a time. Deek’s funnel plot test was used to evaluate the risk of publication bias. Clinical applicability of the studies was evaluated by a Fagan diagram. A P value < 0 0.05 indicated statistically significance.

Results

Fourteen studies were included for analysis according to the eligibility criteria [11–24]. The selection flowchart is shown in Fig. 1. Details of the included studies were presented in Table 1. There was a total of 1698 patients reviewed, of whom 812 patients had DFO and 886 patients had infected DFU without osteomyelitis. These studies involved cases from the USA, Australia, Greece, Turkey, Iran, South Korea, and China. The sample sizes of the individual studies ranged from 24 to 353. The average age varied between 54 and 67.4 years, and 70.08% of them were male (1190 cases). The included literature consisted of eight prospective cohort studies and six retrospective studies. Table 2 provided the details of cut-off value, sensitivity, and specificity for ESR, CRP, and PCT in our study.

Fig. 1.

Flow diagram for study selection

Table 1.

The characteristics of the included studies

	Study design	Country	DFO	Infected DFU	Reference standards	Mean age (years)	Gender (M/F)
Ertugrul 2009	Prospective	Turky	24	22	Bone histopathology, microbiological investigation on bone tissue, and MRI	64	30/16
Fleischer 2009	Prospective	USA	34	20	Bone histopathology	61.5	44/10
Mutluoglu 2011	Prospective	Turkey	13	11	MRI	61.9	18/6
Michail 2013	Prospective	Greece	27	34	Clinical findings, X-ray, nuclear scintigraphy, MRI, and bone probe test.	63.1	45/16
Park 2017	Prospective	South Korea	19	104	NR	67.4	105/18
Hayes 2018	Prospective	Australia	16	11	Clinical findings, X-ray, MRI, and bone probe test.	66.4	21/5
Hadavand 2019	Retrospective	Iran	110	90	MRI	61.26	143/57
Lavery 2019	Retrospective	USA	177	176	Clinical findings, X-ray, MRI, and bone probe test.	54	262/91
Xu 2020	Retrospective	China	111	86	Bone histopathology, culture of bone.	63.38	126/71
Moallemi 2020	Retrospective	Iran	71	71	Clinical findings, X-ray, MRI, and bone probe test.	61.2	94/48
Vangaveti 2021	Prospective	Australia	19	18	Clinical findings, X-ray, MRI, and bone probe test.	63.46	27/10
Soleimani 2021	Prospective	Iran	45	45	MRI	59.7	63/27
Balin 2022	Retrospective	Turkey	96	151	Clinical findings, X-ray, MRI, and bone probe test.	61.5	162/85
Eren 2022	Retrospective	Turkey	50	47	MRI	58.1	50/47

NR: Not report

Table 2.

Main findings of the included studies

	ESR (mm/hr)			CRP (mg/L)			PCT (ng/mL)
	Cut-off	Sen%	Spe%	Cut-off	Sen%	Spe%	Cut-off	Sen%	Spe%
Ertugrul 2009	65	88	73	NR	NR	NR	NR	NR	NR
Fleischer 2009	60	68	70	32	85	65	NR	NR	NR
Mutluoglu 2011	47	72.1	84.6	NR	NR	NR	NR	NR	NR
Michail 2013	67	84	75	14	85	83	0.3	81	71
Park 2017	NR	NR	NR	NR	NR	NR	0.59	94.7	88.5
Hayes 2018	NR	NR	NR	68.5	70.6	80	NR	NR	NR
Hadavand 2019	56.5	95.8	50	44	90.3	57	0.35	86.1	45.3
Lavery 2019	60	73	56	79	49	80	NR	NR	NR
Xu 2020	50	47.7	88.6	NR	NR	NR	NR	NR	NR
Moallemi 2020	49	74.6	57.7	35	76	54.9	NR	NR	NR
Vangaveti 2021	NR	NR	NR	NR	NR	NR	0.064	79	70
Soleimani 2021	19.5	86.7	73.3	53.5	84.4	91.1	0.085	100	97.8
Balin 2022	74.5	62.5	73.1	87	50	73.1	0.19	40.9	84
Eren 2022	68.5	84.0	85.1	41.3	92.0	80.9	NR	NR	NR

NR: Not report

Details of the quality of included studies through the QUADAS-2 tool were shown in Fig. 2. The risk of bias of the patient selection was deemed to be high risk in four studies. For the Reference Standard domain, eleven studies had a low risk of bias and three trials had an unclear risk of bias. In the Index Test domain and Flow and Timing domain, fourteen studies were at a low risk of bias.

Fig. 2.

Quality evaluation of the included studies

The Spearman correlation analysis revealed no significant threshold effect for any biomarker, with correlation coefficients of 0.191 (p = 0.574) for ESR, 0.033 (p = 0.932) for CRP, and − 0.429 (p = 0.397) for PCT. The absence of a distinct ‘shoulder-arm’ pattern in the receiver operating characteristic (ROC) space further corroborated this finding.

Nine studies evaluated the role of CRP in diagnosis DFO compared with infected DFU. The cut-off value ranged from 14 mm/hr to 87 mm/hr. The meta-analysis showed that the combined sensitivity was 0.78 (95% CI: 0.66–0.87, P < 0.01, I² = 92.90%) and the specificity was 0.72 (95% CI: 0.64–0.78, P < 0.01, I² = 80.19%) (Fig. 3A); the PLR was 2.8 (95% CI: 2.1–3.5, P < 0.01, I² = 67.18%), NLR was 0.30 (95% CI: 0.19–0.48, P < 0.01, I² = 92.26%), DOR was 9 (95% CI: 5–17, P < 0.01, I²=100%), and AUC was 0.79 (95% CI: 0.76–0.83) (Fig. 4A).

Fig. 3.

The combined sensitivity and specificity of CRP (A), ESR (B), and PCT (C) for diagnosis of DFO

Fig. 4.

The combined AUC of CRP (A), ESR (B), and PCT (C) for diagnosis of DFO

Eleven trials assessed the role of ESR in diagnosis DFO compared with infected DFU. The cut-off value ranged from 19.5 mm/hr to 74.5 mm/hr. The meta-analysis showed that the combined sensitivity was 0.77 (95% CI: 0.67–0.85, P < 0.01, I² = 88.71%) and the specificity was 0.74 (95% CI: 0.65–0.81, P < 0.01, I² = 84.81%) (Fig. 3B); the PLR was 3.0 (95% CI: 2.2-4.0, P < 0.01, I² = 70.41%), NLR was 0.31 (95% CI: 0.22–0.45, P < 0.01, I² = 85.88%), DOR was 10 (95% CI: 6–17, P < 0.01, I²=100%), and AUC was 0.82 (95% CI: 0.78–0.85) (Fig. 4B).

Six studies investigated the role of PCT in diagnosis DFO compared with infected DFU. The cut-off value ranged from 0.064 ng/mL to 0.59 ng/mL. The meta-analysis indicated that the combined sensitivity was 0.88 (95% CI: 0.63–0.97, P < 0.01 I² = 96.21%) and the specificity was 0.81 (95% CI: 0.63–0.92, P < 0.01, I² = 95.40%) (Fig. 3C); the PLR was 4.7 (95% CI: 2.0-11.4, P < 0.01, I² = 95.03%), NLR was 0.14 (95% CI: 0.03–0.59, P < 0.01, I² = 97.36%), DOR was 33 (95% CI: 4-272, P < 0.01, I²=100%), and AUC was 0.91 (95% CI: 0.88–0.93) (Fig. 4C).

We tried to identify potential sources of heterogeneity by subgroup and meta-regression analysis, and covariates included study design (prospective and retrospective studies), ethnicity (Asians vs. Non-Asians), and cut-off values (CRP: ≥50 mg/L vs. <50 mg/L; ESR: <60 mm/hr vs. ≥60 mm/hr; PCT: <0.1 ng/mL vs. ≥0.1 ng/mL). As shown in Table 3, study design, ethnicity and cut-off value were the main sources of heterogeneity for CRP. Study design was the main sources of heterogeneity in the ESR diagnosis of DFO, while ethnicity may be the primary sources of heterogeneity for PCT. We conducted sensitivity analyses to explore whether a study significantly influenced the results or highly contributed to the heterogeneity. After excluding one study, there were no significant changes in the overall pooled results, showing that the results of this study were stable and reliable (Fig. 5).

Table 3.

subgroup and meta-regression analysis for biomarkers in the diagnosis of DFO

	Parameter	Subgroup	No.of studies	Sensitivity	P of SEN	Specificity	P of SPE
ESR	Study design	Prospective	5	0.79 [0.67–0.92]	0.12	0.80 [0.69–0.90]	0.02
		Retrospective	6	0.76 [0.64–0.87]		0.70 [0.60–0.81]
	Ethnicity	Asia	4	0.80 [0.67–0.92]	0.12	0.75 [0.63–0.88]	0.12
		Non-Asia	7	0.75 [0.64–0.87]		0.73 [0.62–0.84]
	Cut-off value	< 60 mm/hr	5	0.78 [0.66–0.90]	0.15	0.77 [0.65–0.88]	0.07
		≥ 60 mm/hr	6	0.76 [0.64–0.88]		0.72 [0.60–0.83]
CRP	Study design	Prospective	4	0.83 [0.71–0.96]	0.10	0.76 [0.66–0.87]	0.01
		Retrospective	5	0.74 [0.61–0.88]		0.69 [0.61–0.77]
	Ethnicity	Asia	3	0.86 [0.74–0.97]	0.05	0.60 [0.53–0.67]	0.55
		Non-Asia	6	0.74 [0.60–0.87]		0.78 [0.73–0.83]
	Cut-off value	< 50 mg/L	5	0.87 [0.80–0.94]	0.00	0.67 [0.58–0.76]	0.27
		≥ 50 mg/L	4	0.63 [0.49–0.78]		0.77 [0.69–0.85]
PCT	Study design	Prospective	4	0.93 [0.84–1.00]	0.13	0.86 [0.74–0.99]	0.16
		Retrospective	2	0.67 [0.30–1.00]		0.68 [0.39–0.96]
	Ethnicity	Asia	3	0.96 [0.90–1.00]	0.03	0.85 [0.68–1.00]	0.59
		Non-Asia	3	0.67 [0.38–0.96]		0.78 [0.56–1.00]
	Cut-off value	< 0.1 ng/mL	2	0.96 [0.86–1.00]	0.49	0.91 [0.77–1.00]	0.22 |
		≥ 0.1 ng/mL	4	0.81 [0.58–1.00]		0.75 [0.57–0.94]

Fig. 5.

Sensitivity analysis of the meta-analysis, CRP (A), ESR (B), and PCT (C)

We determined publication bias through the Deeks’ regression test of asymmetry, and a low risk of publication bias was presented in relation to CRP (p = 0.09), ESR (p = 0.19) and PCT (p = 0.34) (Fig. 6).

Fig. 6.

Publication bias analysis of the meta-analysis, CRP (A), ESR (B), and PCT (C)

Fagan nomograms were constructed assuming a pre-test probability of 50%. The pre-test and post-test probabilities of CRP were 73% and 23%, respectively, indicating that the possibility of DFO increased from 50% to 73% after CRP diagnosis and that the possibility of no DFO decreased from 50% to 23%. For ESR, the corresponding post-test probabilities were 75% (positive test) and 24% (negative test). PCT demonstrated the highest clinical utility, with a post-test probability of 83% after a positive test and 13% after a negative test. (Fig. 7).

Fig. 7.

Fagan nomogram of CRP (A), ESR (B), and PCT (C) for diagnosis of DFO

Discussion

DFO is a severe complication of diabetes. Given the high rates of disability and mortality associated with DFO, prompt diagnosis is critical. This meta-analysis assessed the diagnostic utility of inflammatory biomarkers for DFO. The results showed that elevated PCT levels had a high diagnostic value for DFO (AUC = 0.91), with a sensitivity of 88%, specificity of 81%, and diagnostic OR of 33.

Diagnosis of DFO may be difficult, partly owing to the absence of universally accepted definitions or a reference standard, and partly because of the limited consistency of commonly used diagnostic tests [25]. The 2023 IWGDF/IDSA guidelines suggested combining probe-to-bone test, serum inflammatory markers, and plain X-rays as a preliminary study to diagnose osteomyelitis of the foot [4]. The probe-to-bone test is the most useful, but it was affected by the skill and experience of clinician’s, as well as ulcer’s location. X-rays are not sensitive to acute osteomyelitis, and often needed to be repeated within 2–3 weeks. Inflammatory biomarkers such as ESR, CRP, and PCT may play a specific role in the diagnosis of DFO, especially when it was difficult to diagnose clinically in the absence of probe-to-bone test and X-ray [26]. Serum tests for these common biomarkers are widely available, easy to obtain, and most are relatively inexpensive.

Two previous meta-analyses have evaluated the diagnostic value of inflammatory biomarkers in diagnosing DFU and osteomyelitis. Majeed et al. [27] concluded that ESR had a diagnostic accuracy of 0.84 in terms of diagnosing DFO. Due to a lack of data, it is not possible to determine the suitability of CRP and PCT for diagnosing DFO. Sharma et al. [28] found that PCT exhibited the highest AUC value of 0.844 when compared with other biomarkers. However, these earlier analyses included heterogeneous patient populations (combining soft tissue and bone infections) and were based on a limited number of studies. We therefore performed this update meta-analysis specifically to assess the ability of these biomarkers in differentiating DFO from infected DFU.

A recent meta-analysis evaluated the diagnostic characteristics of biomarker for DFO [29]. Pooled AUC for ESR, CRP and PCT were 0.83, 0.77, and 0.71, respectively. While their findings for ESR and CRP are consistent with our results (AUCs of 0.82 and 0.79, respectively), a significant discrepancy exists for PCT (0.91 in our analysis versus 0.71). This divergence can be attributed to differences in study inclusion; our analysis incorporated an additional study (Park 2017) and excluded one trial (Crisologo 2020 [30]) where biomarker measurements were taken after six weeks of antibiotic treatment, which could confound the results. Consequently, we found PCT to have the highest AUC value of 0.91, followed by ESR (AUC of 0.82), and CRP (AUC of 0.79). PCT is proved to be a good biomarker for diagnosing DFO and can assist in clinical diagnosis. Further studies are needed to confirm these findings and guide clinical practice.

In healthy individuals, PCT levels are typically low (< 0.05-0.1ng/mL), and they become elevated in the presence of inflammation, such as diabetes, infection, and autoimmune disease [31]. It has been used as a biomarker for the diagnosis of sepsis, severe sepsis and septic shock [32]. Recent studies suggest that PCT may be more effective than conventional biomarkers like CRP and ESR in the diagnosis of infected DFU [33]. Although PCT is emerging as a popular infection biomarker, its role in the diagnosis of DFO remains controversial [34]. Mutluoğlu et al. found that PCT failed to identify patients with bone infection [12], whereas Asten et al. recommended that it was useful to detecting osteomyelitis in patients with infected DFU [8]. Conversely, Glaudemans et al. [35] found that PCT had little diagnostic value and specificity for DFO. Therefore, we conducted this study to assess the diagnostic accuracy of PCT in differentiating DFO and soft tissue infection. The high DOR of 33 obtained in our analysis indicates a robust diagnostic performance for PCT.

There are several limitations to our review. First, the heterogeneity was high among the included studies. According to the subgroup and meta-regression analyses, study design, ethnicity and cut-off value may be the sources of heterogeneity. Furthermore, further unrecorded difference among these studies may also contribute. High-quality and well-designed research is required to minimize such variability. Second, both prospective and retrospective studies were included in our meta-analysis, which could cause selection bias. Larger-scale, randomized controlled trials are needed in the future. Third, the cut-off values of PCT for diagnosing DFO ranged from 0.064 ng/mL to 0.59 ng/mL. This wide range underscores the current lack of a standardized threshold for clinical use. Future studies should specifically aim to identify and validate an optimal, standardized cutoff for PCT in DFO diagnosis.

In conclusion, this study assessed the diagnostic roles of ESR, CRP, and PCT in DFO, and found that PCT had the highest diagnostic test accuracy for distinguishing DFO from soft tissue infection in diabetes. However, PCT should not be used as the sole definitive test for DFO diagnosis. Its results must be interpreted in conjunction with information from careful medical history, X-rays, and when feasible, the probe-to-bone test. More high-quality prospective controlled studies are needed in the future.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, JZ Dai; Investigation, SS Mei; Methodology, SS Mei; Writing – original draft, JZ Dai and SS Mei; Writing – review & editing, JZ Dai, and H Chen. All authors have read and approved the manuscript, and ensure that this is the case.

Funding

Not applicable.

Data availability

The data used to support the findings of this study are included within the article, which are available from the corresponding author upon request.

Declarations

Ethics approval and consent to participate

Declarations.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.