Diagnostic Accuracy of C-Reactive Protein for Periprosthetic Joint Infection: A Meta-Analysis

Abstract

Background:

We conducted a study to assess the diagnostic accuracy of the C-reactive protein (CRP) assay for periprosthetic joint infection (PJI).

Methods:

For the purpose of the study, we conducted a search of PubMed, selecting only studies that described the diagnosis of PJI through measurement of the serum CRP concentration. The results were meta-analyzed by pooling estimates of sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR−), and diagnostic odds ratio (DOR) and constructing respective summary receiver-operating characteristic (SROC) curves. We also conducted subgroup analyses according to cutoff values of the serum CTP concentration.

Results:

A total of 25 studies met the selection criteria for inclusion in the analysis. The pooled estimates for sensitivity, specificity, and the area under the curve (AUC) for the CRP assay were 0.82 (95% CI 0.80–0.84), 0.77 (95% CI 0.76–0.78), and 0.877±0.016, respectively. In the 10 mg/L cutoff subgroup, the pooled estimates for sensitivity, specificity, and the AUC were 0.881 (95% CI 0.859–0.901), 0.728 (95% CI 0.706–0.749), and 0.8496±0.0248, respectively.

Conclusion:

We found that the serum CRP assay had good diagnostic accuracy for PJI but that its specificity was low. We recommend that when the CRP assay yields a normal result, other tools be used adjunctively with it to reduce the rate of false-negative diagnoses of PJI.

Total joint arthroplasty (TJA) is a safe and effective procedure that improves the quality of life and restores function to patients with arthritis. Such arthroplasty has been increasingly used in the past 50 y, especially in the growing population of elderly persons. However, periprosthetic joint infection (PJI) remains one of the most severe complications of TJA. In a retrospective cohort study of 9,245 patients undergoing primary hip or knee arthroplasty, Pulido et al. reported an incidence of PJI of 0.7%, and incidences of 1.1% for knee arthroplasty and 0.3% for hip arthroplasty, respectively [1]. In a 15-y prospective survey of 10,735 patients undergoing primary hip or knee replacement, Phillips et al. reported PJI in 0.57% of hip replacements and 0.86% of knee replacements [2]. Periprosthetic joint infection is often accompanied by pain, decreased range of motion (ROM), and functional disability, leading ultimately to implant failure and the need for revision arthroplasty. In a retrospective review, Clohisy et al. reported PJI as the reason for revision surgery in 7% of cases of hip arthroplasty [3]. The economic burden of PJI is expected to increase substantially over the coming decade. A study by Kurtz et al. found that the annual cost to hospitals in the United States of revision surgery for infected prosthetic joints increased from $320 million to $566 million from 2001–2009, and was projected to exceed $1.62 billion by 2020 [4].

Traditionally, microbiologic examination of tissue or fluid samples obtained from an affected prosthetic joint during revision surgery have been considered the gold standard for PJI. However, the diagnosis of PJI in the pre-operative period can be particularly difficult. Some researchers have used the results of culture of joint fluid aspirates for the pre-operative diagnosis of PJI. In the aspiration of joint fluid for microbiologic investigation, correct needle placement must be confirmed, especially in patients who are obese or those with acetabular protrusions or periarticular ossifications. However, Müller et al. reported that the culture of joint aspirates in patients with TJAs without removal of the prosthesis yielded false-negative results in 28% [5]. C-reactive protein (CRP) is an acute-phase serum protein whose concentrations increase in response to inflammation. A meta-analysis has already shown good accuracy for measurement of the serum concentration of CRP in the diagnosis of bacterial infections in hospitalized patients, with a sensitivity of 75% (95% CI 62%–84%) and specificity of 67% (95% CI 56%–77%) [6]. The serum CRP assay is also used widely in the diagnosis of PJI, but its diagnostic accuracy has varied across studies. Dupont et al. reported that the CRP assay had a sensitivity of 58.3% and specificity of 100% at a threshold CRP concentration of 25 mg/L; at a threshold of 18 mg/L the sensitivity and specificity of the assay were 66.7% and 88.9%, respectively [7]. Johnson reported a sensitivity of 95% and specificity of 20% for the CRP assay in PJI [8]. To evaluate adequately the diagnostic accuracy of the CRP assay for PJI, we systematically reviewed and conducted a meta-analysis of studies that investigated serum CRP concentrations as markers for PJI.

Methods

Search strategy

We attempted to identify studies of the diagnostic accuracy of the serum CRP assay for PJI that were published in the PubMed database before February 2013. The terms used in our search of PubMed were: C-Reactive Protein [title/abstract] OR CRP [title/abstract]) AND (periprosthetic [title/abstract] OR arthroplasty [title/abstract] OR replacement [title/abstract]) AND infection* [title/abstract] AND (“Sensitivity and Specificity” [MeSH Terms] OR specificity*[tw] OR accuracy [tw] OR predictive value*[tw] OR ROC[tw] OR likelihood ratio*[tw]). We supplemented our searches by reviewing manually the references of all relevant studies. No language restrictions were applied in these searches.

Inclusion and exclusion criteria

We included all studies that met the criteria of: (1) Assessing the diagnostic accuracy of the CRP assay for PJI; (2) comparing the diagnostic accuracy of the serum CRP assay with a reference standard (gold standard); and (3) providing true-negative (TN), false-positive (FP), true-positive (TP), and false-negative (FN) results, based on the data presented in the study, so that the sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR–), and diagnostic odds ratio (DOR) could be calculated. Two reviewers judged independently the eligibility of studies for inclusion in the meta-analysis while screening the citations of the studies in PubMed. Disagreements of the reviewers were resolved by a third reviewer.

Data extraction and quality assessment

In our search of the PubMed database, we used a standard form to extract data on the first author of a study, its year of publication, the country in which the study was done, the study design, the diagnostic gold standard used in the study, the study population and number of patients, the patients' genders and ages, the prevalence of PJI, and the cutoff value used for the concentration of CPR. We also recorded the number of TN, FP, TP, and FN results, allowing us to construct two-by-two tables. We used the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tool to assess the quality of our investigations of predictive validity. Each item of the QUADAS tool was answered with a reply of “Yes,” “No,” or “Unclear” [9]. Two reviewers assessed independently the quality of the studies brought up in the search. Disagreements of the reviewers were resolved by a third reviewer.

Statistical analysis

The overall pooled values for sensitivity, specificity, and diagnostic odds ratio (DOR) for PJI of the serum CRP concentration in the studies used in the meta-analysis, with 95% confidence intervals (CIs), were estimated with DerSimonian and Laird's random-effects model [10]. In addition, summary receiver-operating characteristic (SROC) analysis was done to examine the interaction between sensitivity and specificity [11] and to quantify the performance of the CRP assay using the AUC and Q* value [12]. The heterogeneity of the studies used in the meta-analysis was assessed with the Cochran Q test and the I² statistic. A value of p<0.05 with the Cochran Q test indicated significant heterogeneity; A value of I² >50% indicated substantial heterogeneity. We conducted threshold analyses using the Spearman rank correlation test for Logit(TPR) and Logit(1–TNP). Subgroup analyses were conducted on the basis of cutoff values of CRP. All analyses were done with Meta DiSc 1.4 software version 0.6 [13] (http://www.hrc.es/investigacion.metadisc.en.htm).

Results

Eligible studies

Our initial search yielded 72 citations, of which 31 were excluded after screening of the abstracts and titles. We retrieved 41 studies for full-text reading, of which 20 met our inclusion criteria. Another five studies were added after a manual review of relevant studies. Ultimately, 25 study papers [5,7–8,14 –35] were included for analysis. The flow diagram showing the process of selection of the studies is shown in the Supplementary material (www.liebertpub.com/sur). The 25 included studies were conducted in nine countries between 1989 and 2013, and included 4,088 patients. Twelve of the studies used an appropriate prospective cohort-study design and seven studies used a retrospective cohort or case–control design. All of the studies included in the meta-analysis used clinical criteria to define PJI. The populations of the studies consisted largely of patients undergoing total hip arthroplasty (THA) or TKA. The cutoff value of the serum CRP concentration varied from 0.06–500 mg/L, although a value of 10 mg/L was used in most studies. The details of all 25 studies included in the meta-analysis are shown in Table 1.

Table 1.

Characteristics of Studies Included in Meta-Analysis

First author/year	Country	Study design	Diagnosis gold standard	Population	No. of patients	Gender (male/female)	Age	Prevalence
Fink 2013	Germany	Prospective cohort	Microbiologic and histologic analyses of periprosthetic tissue	Revision of hip endoprostheses	100	NR	NR	45.0%
Parvizi 2012	USA	Prospective cohort	Three of the following four thresholds were abnormal: (1) ESR 30 mm/h or greater; (2) CRP 10 mg/L or greater; (3) synovial WBC 1,760×10⁹ cells/L or greater; or (4) synovial PMNs 73% or greater	Revision of knee arthroplasty	64	NR	NR	45.5%
Johnson 2011	USA	Prospective cohort	One of the following four criteria were required: (1) Two or more positive cultures of the same organism, (2) histologic evidence of acute inflammatory response observed on intra-operative frozen section, (3) gross purulence, or (4) a draining sinus tract that communicated with the joint space	Revision of total knee arthroplasty for suspected periprosthetic infection	113	61/52	61 (28–89)	92.9%
Buttaro 2010	Argentina	Prospective cohort	Positive histopathologic evidence of infection or growth of bacteria on culture	Revision of total hip arthroplasty	69	19/50	68 (38–91)	24.6%
Piper 2010	USA	NR	At least one of the following was present: (1) Visible purulence surrounding the implant; (2) acute inflammation on histopathologic examination of permanent tissue sections (as determined by the clinical pathologist); (3) a sinus tract communicating with the implant; or (4) positive periprosthetic tissue culture and positive sonicate fluid culture for the same microorganism	Knee, hip or shoulder arthroplasty, or spine implant removal	297 Knee arthroplasty	NR	NR	64.6%
					221 Hip arthroplasty	NR	NR	45.3%
					64 Shoulder arthroplasty	NR	NR	73.1%
					54 Spine implant removal	NR	NR	70.0%
					297 Knee arthroplasty	NR	NR	75.9%
					221 Hip arthroplasty	NR	NR	45.9%
					64 Shoulder arthroplasty	NR	NR	61.3%
					54 Spine implant removal	NR	NR	51.9%
Tohtz 2010	Germany	Prospective cohort	Met one of the criteria of: Identification of the same bacterial organism in at least two tissue samples, purulence in the joint, permanent sections showing evidence of infection, a hip-associated sinus tract was present	Component loosening after hip arthroplasty	64	26/37	66.7 (24–89)	29.7%
Ghanem 2009	Canada	Retrospective cohort	Met one of the following criteria: (1) An abscess or sinus communicating with the joint space, (2) positive pre-operative culture of joint aspirate on solid media, (3) two or more positive intra-operative cultures or one positive culture on solid media in conjunction with the presence of other indicators of infection including gross intracapsular purulence or an elevated cell count and differential count in aspirate fluid.	Revision of total hip arthroplasty	479	254/225	66 (23–93)	26.5%
Pfitzner 2008	Germany	Retrospective	Microbiologic and histopathologic findings	Total hip arthroplasty or total knee arthroplasty	50	16/34	67.4 (48–81)	50.0%
Müller 2008	Germany	Prospective	At least one of the following three criteria was met: An open wound or sinus communicating with the joint, a systemic infection with pain in the hip and purulent fluid within the joint, or a positive result on at least three tests in the presence of positive histologic evaluation or positive intra-operative cultures	Two-stage revision for assumed prosthetic joint in fection	50	23/27	69 (46–84)	74%
Dupont 2008	France	NR	NR	Total hip arthroplasty or total knee arthroplasty	48	NR	NR	25.0%
Della Valle 2007	USA	NR	Organism grew on solid media in at least two of three cultures or two of the following three criteria were met: At least one culture was positive, the final histopathology was consistent with infection, or gross purulence was seen at the time of revision surgery	Total knee arthroplasty	94	NR	NR	43.6%
Greidanus 2007	Canada	Prospective	NR	Revision of total knee arthroplasty	151 Knees	NR	NR	29.8%
Baré 2006	Canada	Retrospective	NR	Revision of total knee arthroplasty	295	NR	NR	26.8%
Itasaka 2001	Japan	NR	Meeting both of the criteria of: (1) A positive intra-operative sample culture and (2) five or more PMN in a highly magnified view (×400) on pathologic examination.	Revision of total hip arthroplasty	46 Patients with 48 THA	8/38	67.5 (49–78)	12.5%
Spangehl 1999,	USA	Prospective	NR	Revision of total hip arthroplastiy	178 Patients with 202 revision hip replacements	NR	NR	17.3%
Austin 2008	USA	Retrospective	Either pre-operative or intra-operative cultures were positive; the leukocyte count was more than 1,760×10⁹ cells/L and PMN were more than 73%, or there was either a draining sinus tract or abscess.	Revision of total knee arthroplasy	296	125/171	66	39.2%
Bottner 2007	Germany	Prospective	Positive culture and histologic evidence of deep infection	Revision oif total knee or hip replacement	78	41/37	64 (19–90)	26.9%
Chevillotte 2009	USA	Retrospective	Two or more samples of the intra-operative tissue or the intra-articular aspiration were positive for growth on bacterial culture	Periprosthetic hip fractures	204	93/111	70	11.6%
Kanner 2008	USA	Retrospective	Culture results from intra-operative samples	Hip and knee prosthetic revisions	244	NR	NR	9.8%
Simonsen 2007	Denmark	Retrospective	One or more of the following criteria were fulfilled: Positive cultures from biopsy specimens, purulence at the time of surgery defined as the presence of pus identified by the surgeon, or acute inflammation consistent with infection on histopathologic examination, defined as the presence of more than 10 neutrophil granulocytes per high power field in fresh-frozen sections	Painful hip prosthesis where loosening and/or infection were suspected.	76 Scintigraphies were performed in 66 patients.	Uninfected 19/30; Infected 16/11	Uninfected 73 (47–92); Infected 74 (63–95)	35.5%
Nilsdotter- Augustinsson 2007	Sweden	Prospective	(1) The orthopedic surgeon classified prosthesis loosening pre-operatively as being due to a deep infection of the device, based on clinical symptoms, laboratory and radiographic findings, pre-operative cultures of synovial fluid aspirates, or occurrence of fistulas; or (2) the surgeon deemed the patient infected because bacterial growth was observed in cultures from tissue samples taken at revision surgery, and the surgeon decided to treat the patient as being infected.	Revision surgery for radiographically confirmed loosening of hip prostheses	213	14/32 Patients in control group; 35/25 in revision group; 15/10 in infection group,	69 (43–89) Patients in control group; 74 (38–86) in revision group; 70 (30–83) in infection group	17.7%
Bernard 2004	Switzerland	NR	Minimum of one intra-operative culture was positive for a particularly virulent organism and associated purulence was present	Post-hip or -knee prosthesis who underwent a diagnostic work-up for suspected prosthetic joint infection	230	NR	69±18 (40–100)	91.0%
Sanzén 1989	Sweden	NR	Multiple positive bacteriologic cultures of biopsy specimens from the bone-cement interface obtained during revision surgery.	Revision of total hip arthrop!asty	108	14/9 Patients in infection group	74 (40–80) Patients in infection group; 68 (39–84) in group 1; 69 (42–79) in group 2	20.7%
Di Cesare 2005	USA	Prospective	At least10 PMN per high-power field on post-operative histologic examination of pathologic specimens and the presence of bacteria on culture of intra-operative specimens obtained before the administration of antibiotics		Total hip or knee replacement and re-operation	25/33	63 (25–87)	29.3%
Fink 2008	Germany	Prospective	At least one of the following conditions was fulfilled: Demonstration of the same micro-organism in at least two cultures; or demonstration of a micro-organism in at least one sample and at least five PMN per high-power field (x 400) in the associated histologic preparation.	Revision of total knee replacement	144 Patients with 145 knees	63/81	68 (30–87)	27.6%

CRP=C-reactive protein; ESR=erythrocyte sedimentation rate; NR=not reported; PMN=polymorphonuclear netrophils; THA=total hip arthroplasty; WBC=white blood cell count.

The quality of the included studies was modest. Four studies did not provide a sufficient spectrum of patients for testing and four other studies did not provide a reference standard for the diagnosis of PJI. Moreover, none of the studies included in the meta-analysis reported uninterpretable or indeterminate results, and none provided information about patient withdrawal or dropout. A graph and summary of the methodologic quality of the studies is shown in the appendix.

Main meta-analysis

The performance of the serum CRP assay for the diagnosis of PJI in the 25 studies included in the meta-analysis is presented in Table 2. The sensitivities of the CRP assay varied from 0.421–1.000 (a range of 0.579) and its specificities varied from 0.204–1.000 (a range of 0.796). The LR+varied from 1.163–42.692 (a range of 41.529) and the LR– varied from 0.033–0.686 (a range of 0.653). The overall pooled sensitivity of the CRP assay for PJI was 0.817 (95% CI 0.798–0.836; χ² ₍₃₄₎=238.11, p=0.000; I²=85.7%). The overall specificity of the CRP assay for PJI was 0.771(95% CI 0.757–0.784; χ²₍₃₄₎=362.54, p=0.000; I²=90.6%). The combined LR+and LR− were 3.659 (95% CI 2.917–4.591; χ²₍₃₄₎=347.56, p=0.000; I²=90.2%) and 0.262 (95% CI 0.205–0.334; χ²₍₃₄₎=167.52, p=0.000; I²=79.7%), respectively, yielding a DOR of 17.012 (95% CI 11.378–25.435; χ² ₍₃₄₎=145.14, p=0.000; I²=76.6%). There was a threshold effect between the studies (r=0.339; p=0.047). The overall weighted AUC of the SROC curve was 0.8736±0.0158, and the overall diagnostic accuracy (Q*) was 0.8040±0.0158. The forest plots for the pooled sensitivity, specificity, LR+, LR−, and DOR are shown in the Appendix, and the SROC curve is shown in Figure 1.

FIG. 1.

Plot of summary receiver-operating curve for serum C-reactive protein in the diagnosis of periprosthetic joint infection in studies included in the meta-analysis.

Table 2.

Diagnostic Performance of Serum C-Reactive Protein Assay for Periprosthetic Joint Infection in Studies Included in Meta-Analysis

First author/year	Cutoff (mg/L)	TP	FP	FN	TN	Sensitivity (95%CI)	Specificity (95%CI)	LR+(95%CI)	LR− (95%CI)
Fink 2013	10	29	14	16	41	0.644 (0.488−0.781)	0.745 (0.610−0.853)	2.532 (1.533−4.181)	0.477 (0.313−0.728)
Parvizi 2012	16.5	19	2	6	28	0.760 (0.549−0.906)	0.933 (0.779−0.992)	11.400 (2.935−44.280)	0.257 (0.127−0.520)
Johnson 2011	10	100	6	5	2	0.952 (0.892−0.984)	0.250 (0.032−0.651)	1.270 (0.849−1.899)	0.190 (0.044−0.832)
Buttaro 2010	10	12	5	5	47	0.706 (0.440−0.897)	0.904 (0.790−0.968)	7.341 (3.021−17.841)	0.325 (0.155−0.683)
Piper 2010	10	68	45	14	170	0.829 (0.730−0.903)	0.791 (0.730−0.843)	3.962 (3.001−5.231)	0.216 (0.133−0.350)
	10	25	41	9	146	0.735 (0.556−0.871)	0.781 (0.715−0.838)	3.354 (2.393−4.699)	0.339 (0.193−0.597)
	10	8	7	11	38	0.421 (0.203−0.665)	0.844 (0.705−0.935)	2.707 (1.144−6.403)	0.686 (0.458−1.026)
	10	8	6	6	34	0.571 (0.289−0.823)	0.850 (0.702−0.943)	3.810 (1.602−9.057)	0.504 (0.272−0.936)
	14.5	65	26	17	189	0.793 (0.689−0.874)	0.879 (0.828−0.919)	6.555 (4.496−9.556)	0.236 (0.154−0.361)
	10.3	25	40	9	147	0.735 (0.556−0.871)	0.786 (0.720−0.843)	3.438 (2.445−4.834)	0.337 (0.191−0.593)
	7	12	12	7	33	0.632 (0.384−0.837)	0.733 (0.581−0.854)	2.368 (1.308−4.289)	0.502 (0.272−0.929)
	4.6	11	13	3	27	0.786 (0.492−0.953)	0.675 (0.509−0.814)	2.418 (1.432−4.082)	0.317 (0.114−0.886)
Tohtz 2010	10	11	9	8	36	0.579 (0.335−0.797)	0.800 (0.654−0.904)	2.895 (1.439−5.823)	0.526 (0.305−0.910)
Ghanem 2009	10	116	82	11	270	0.913 (0.850−0.956)	0.767 (0.719−0.810)	3.921 (3.220−4.775)	0.113 (0.064−0.199)
	20.5	120	7	67	285	0.642 (0.568−0.710)	0.976 (0.951−0.990)	26.769 (12.776−56.086)	0.367 (0.303−0.445)
Pfitzner 2008	500	13	3	12	22	0.520 (0.313−0.722)	0.880 (0.688−0.975)	4.333 (1.405−13.366)	0.545 (0.354−0.841)
Müller 2008	NR	35	5	2	8	0.946 (0.818−0.993)	0.615 (0.316−0.861)	2.459 (1.231−4.913)	0.088 (0.021−0.362)
Dupont 2008	25	7	0	5	36	0.583 (0.277−0.848)	1.000 (0.903−1.000)	42.692 (2.617−696.54)	0.429 (0.227−0.810)
	18	8	4	4	32	0.667 (0.349−0.901)	0.889 (0.739−0.969)	6.000 (2.192−16.422)	0.375 (0.167−0.842)
Della Valle 2007	10	40	13	1	40	0.976 (0.871−0.999)	0.755 (0.617−0.862)	3.977 (2.474−6.394)	0.032 (0.005−0.225)
Greidanus 2007	13.5	41	15	4	91	0.911 (0.788−0.975)	0.858 (0.777−0.919)	6.439 (3.993−10.381)	0.104 (0.041−0.265)
Baré 2006	NR	47	80	32	136	0.595 (0.479−0.704)	0.630 (0.561−0.694)	1.606 (1.249−2.066)	0.643 (0.483−0.856)
Itasaka 2001	300	5	10	1	32	0.833 (0.359−0.996)	0.762 (0.605−0.879)	3.500 (1.830−6.695)	0.219 (0.036−1.320)
Spangehl 1999,	NR	29	40	6	127	0.829 (0.664−0.934)	0.760 (0.688−0.823)	3.459 (2.539−4.714)	0.225 (0.108−0.469)
Austin 2008	10	109	47	7	133	0.940 (0.880−0.975)	0.739 (0.668−0.801)	3.599 (2.803−4.621)	0.082 (0.040−0.168)
Bottner 2007	320	20	2	1	55	0.952 (0.762−0.999)	0.965 (0.879−0.996)	27.143 (6.934−106.26)	0.049 (0.007−0.334)
	150	20	5	1	52	0.952 (0.762−0.999)	0.912 (0.807−0.971)	10.857 (4.675−25.216)	0.052 (0.008−0.354)
Chevillotte 2009	10	17	81	4	102	0.810 (0.581−0.946)	0.557 (0.482−0.631)	1.829 (1.405−2.381)	0.342 (0.140−0.833)
Kanner 2008	NR	21	77	3	143	0.875 (0.676−0.973)	0.650 (0.583−0.713)	2.500 (1.976−3.163)	0.192 (0.066−0.557)
Simonsen 2007	10	25	39	2	10	0.926 (0.757−0.991)	0.204 (0.102−0.343)	1.163 (0.974−1.389)	0.363 (0.086−1.538)
Nilsdotter- Augustinsson 2007	10	18	30	4	72	0.818 (0.597−0.948)	0.706 (0.607−0.792)	2.782 (1.942−3.985)	0.258 (0.105−0.631)
Bernard 2004	10	200	4	7	17	0.966 (0.932−0.986)	0.810 (0.581−0.946)	5.072 (2.100−12.255)	0.042 (0.020−0.089)
Sanzén 1989	10	23	20	0	35	1.000 (0.852−1.000)	0.636 (0.496−0.762)	2.675 (1.886−3.794)	0.033 (0.002−0.514)
Di Cesare 2005	10	16	9	1	32	0.941 (0.713−0.999)	0.780 (0.624−0.894)	4.288 (2.378−7.729)	0.075 (0.011−0.508)
Fink 2008	13.5	29	20	11	85	0.725 (0.561−0.854)	0.810 (0.721−0.880)	3.806 (2.456−5.899)	0.340 (0.204−0.567)

CI=confidence interval; FN=false negative; FP=false positive; LR–=negative likelihood ratio; LR+= positive likelihood ratio; NR=not reported; Sen=sensitivity; Spe=specificity; TN=true negative; TP=true positive.

Subgroup analyses

A total of 14 studies provided a CPR cutoff value of 10 mg/L. In this 10 mg/L cutoff subgroup, the overall pooled sensitivity for PJI was 0.881 (95% CI 0.859–0.901; χ²₍₁₆₎=115.66, p=0.000; I²=86.2%) and the overall specificity was 0.728 (95% CI 0.706, 0.749; χ²₍₁₆₎=122.90, p=0.000; I²=87.0%). There was a threshold effect among the studies (r=0.505; p=0.039). The overall weighted AUC of the SROC curve was 0.8496±0.0248, and the overall Q* was 0.7809±0.0234. The forest plots for the pooled sensitivity, specificity, LR+, LR−, and DOR are shown in the Appendix and the SROC curve is shown in Figure 2.

FIG. 2.

Plot of summary receiver-operating curve for serum C-reactive protein in diagnosing periprosthetic joint infection in the 10 mg/L cutoff subgroup.

Discussion

In 2011, the workgroup of the Musculoskeletal Infection Society (MSIS) proposed a new definition for PJI, specifying that: (1) It must exhibit a sinus tract communicating with the prosthesis; or (2) a pathogen isolated by culture from two separate tissue or fluid samples from the affected joint; or (3) the existence of four of the following six criteria: (a) An elevated erythrocyte sedimentation rate (ESR) and CRP concentration; (2) an elevated synovial leukocyte count; (3) an elevated percentage of synovial neutrophils; (4) purulence in the affected joint; (5) isolation of a microorganism; or (6) greater than five neutrophils per high-power field in five high-power fields observed from histologic analysis of periprosthetic tissue at X400 magnification [36]. The American Academy of Orthopedic Surgeons (AAOS) recommended strongly in 2010 that CRP testing be done for patients being assessed for PJI [37]. The importance of the serum CRP assay has been well accepted for the diagnosis of PJI. Our investigation assessed the accuracy of the serum CRP assay when used alone for the diagnosis of PJI. Pooled data from the 25 studies included in our meta-analysis yielded an overall weighted AUC for the SROC curve of 0.8736±0.0158. An AUC of 0.5 indicates that a test has no discriminatory ability, whereas an AUC of 1.0 indicates perfect diagnostic capability [38]. Although the AUC should be in the region of 0.97 or greater to demonstrate excellent accuracy, an AUC of 0.93 to 0.96 is very good and an AUC of 0.75 to 0.92 is good, but an AUC of less than 0.75 indicates obvious deficiencies in the diagnostic accuracy of an assay [39]. The overall AUC and 95% CI for SROC curve were 0.8736 and 0.8426–0.9046, respectively. This result showed that the diagnostic accuracy of the serum CRP assay for PJI was good. The overall diagnostic accuracy (Q*) of the assay was 0.8040±0.0158, with a 95% CI of 0.7730–0.8450, which also indicated good diagnostic accuracy.

We also found that the overall pooled sensitivity of the serum CRP assay for PJI was 0.817 (95% CI 0.798–0.836) and that its overall pooled specificity was 0.771 (95% CI 0.757–0.784). The sensitivity of a diagnostic test, also called the true positive rate, measures the proportion of truly positive results that are correctly identified as such (in the present context, the percentage of patients with PJIs who are correctly identified as having PJIs), whereas the specificity of a test, also called the true negative rate, measures the proportion of truly negative results of the test that are correctly identified as such (the percentage of people without a PJI who are correctly identified as not having a PJI). In the meta-analysis done in our study, the overall pooled specificity of the CAP assay was low. This means there were many patients with true PJIs for whom the CRP assay yielded normal results. The reason for this may have been that some patients failed to mount a sufficient immune response to the pathogen responsible for their PJI. In a prospective cohort study, Johnson et al. reported that the serum CRP assay had a specificity of only 20% (95% CI 6%–50%) at a cutoff value of 10 mg/L [8]. Therefore, when the serum CRP concentration is normal, additional means for making a diagnosis of PJI should be used adjunctively to reduce the rate of false-negative results.

We also did subgroup analyses as part of the present study. Because a CRP cutoff value of 10 mg/L was used in most of the studies, we analyzed the subgroup for which this cutoff value was used. The overall weighted AUC of the SROC curve in this analysis was 0.8496±0.0248, and the overall Q* was 0.7809±0.0234; the overall pooled sensitivity was 0.881 (95% CI 0.859–0.901) and the overall specificity was 0.728 (95% CI 0.706–0.749). The diagnostic accuracy of the CRP assay for PJI was also good, but the specificity was low, as it had been in the original meta-analysis. We were unable to conduct subgroup analyses for THA or TKA subgroups because of insufficient data. The serum CRP assay may have different diagnostic accuracies for THA and TKA. Piper et al., using the 10 mg/L cutoff value for CRP, reported a sensitivity of 83% for the CRP assay in the TKA subgroup as compared with a sensitivity of 74% in the THA subgroup. The specificity with a 14.5 mg/L may be the best cutoff in the TKA subgroup; and 10.3 mg/L was the best cutoff in the THA subgroup. With an optimized CRP cutoff value, the sensitivity of the assay was 79% in the TKA subgroup vs. 88% in the THA subgroup, and the specificity was 79% in the TKA subgroup as compared with 78% in the THA subgroup [17]. Some prospective cohort studies remain necessary for comparing the diagnostic accuracy of the serum CRP assay for PJI in the subgroups of patients with TKA and THA.

Our meta-analysis has some limitations. First, there is currently no agreement about a gold standard for the diagnosis of PJI. Different reference standards were used in the studies included in the meta-analysis, and four studies did not report a reference standard. Second, most of the studies included in the meta-analysis enrolled patients because of suspected PJI. The incidence of PJI ranged from 11.6%–92.9%, which was higher than the 1% incidence of PJI in all patients who had total joint arthroplasty in the studies included in the meta-analysis. The different reference standards and higher incidence of PJI also lead to bias in assessing the diagnostic accuracy of the CRP assay. Third, there was an obvious threshold effect in the meta-analyses. These limitations need to be borne in mind when evaluating the conclusions that can be drawn from the results of the present study.

In conclusion, our meta-analysis indicates that the serum CRP assay for PJI has good diagnostic accuracy but a low specificity. We recommend that a normal result of the serum CTP assay be accompanied by adjunctive other diagnostic methods for reducing the rate of false-negative results.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

References

Pulido

, Ghanem

, Joshi

, et al. Periprosthetic joint infection: the incidence, timing, and predisposing factors. Clin Orthop Relat Res, 2008; 466:1710–1715.

Phillips

, Crane

, Noy

, et al. The incidence of deep prosthetic infections in a specialist orthopaedic hospital: A 15-year prospective survey. J Bone Joint Surg Br, 2006; 88:943–948.

Clohisy

, Calvert

, Tull

, et al. Reasons for revision hip surgery: A retrospective review. Clin Orthop Relat Res, 2004; 429:188–192.

Kurtz

, Lau

, Watson

, et al. Economic burden of periprosthetic joint infection in the United States. J Arthroplasty, 2012; 27:61–65.e1.

Müller

, Morawietz

, Hasart

, et al. Diagnosis of periprosthetic infection following THA: Evaluation of the diagnostic values of pre- and intraoperative parameters and the associated strategy to preoperatively select patients with a high probability of joint infection. J Orthop Surg, 2008; 3:31.

Simon

, Gauvin

, Amre

, et al. Serum procalcitonin and C-reactive protein levels as markers of bacterial infection: A systematic review and meta-analysis. Clin Infect Dis, 2004; 39:206–217.

Dupont

, Rodenbach

, Flachaire

. The value of C-reactive protein for postoperative monitoring of lower limb arthroplasty. Ann Readapt Med Phys, 2008; 51:348–357.

Johnson

, Zywiel

, Stroh

, et al. Serological markers can lead to false negative diagnoses of periprosthetic infections following TKA. Int Orthop, 2011; 35:1621–1626.

Whiting

, Rutjes

, Reitsma

, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol, 2003; 10:25–38.

10.

DerSimonian

, Laird

. Meta-analysis in clinical trials. Control Clin Trials, 1986; 7:177–188.

11.

Moses

, Shapiro

, Littenberg

. Combining independent studies of a diagnostic test into a summary ROC curve: Data-analytic approaches and some additional considerations. Stat Med, 1993:1293–1316.

12.

Walter

. Properties of the summary receiver operating characteristic (sROC) curve for diagnostic test data. Stat Med, 2002; 21:1237–1256.

13.

Javier

Zamora

, Victor

Abraira

, Alfonso

Muriel

, et al. Meta-Disc: A software for meta analysis of test accuracy data. BMC Med Res Methodol, 2006; 6:31–47.

14.

Fink

, Gebhard

, Fuerst

, et al. High diagnostic value of synovial biopsy in periprosthetic joint infection of the hip. Clin Orthop Relat Res, 2013; 471:956–964.

15.

Parvizi

, Jacovides

, Adeli

, et al. Coventry Award: synovial C-reactive protein: A prospective evaluation of a molecular marker for periprosthetic knee joint infection. Clin Orthop Relat Res, 2012; 470:54–60.

16.

Buttaro

, Tanoira

, Comba

, et al. Combining C-reactive protein and interleukin-6 may be useful to detect periprosthetic hip infection. Clin Orthop Relat Res, 2010; 468:3263–3267.

17.

Piper

, Fernandez-Sampedro

, Steckelberg

, et al. C-reactive protein, erythrocyte sedimentation rate and orthopedic implant infection. PLoS One, 2010; 5:e9358.

18.

Tohtz

, Müller

, Morawietz

, et al. Validity of frozen sections for analysis of periprosthetic loosening membranes. Clin Orthop Relat Res, 2010; 468:762–768.

19.

Ghanem

, Antoci

Jr , Pulido

, et al. The use of receiver operating characteristics analysis in determining erythrocyte sedimentation rate and C-reactive protein levels in diagnosing periprosthetic infection prior to revision THA. Int J Infect Dis, 2009; 13:e444–449.

20.

Pfitzner

, Krocker

, Perka

, et al. C-reactive protein. An independent risk factor for the development of infection after primary arthroplasty. Orthopade, 2008; 37:1116–1120.

21.

Della Valle

, Sporer

, Jacobs

, et al. Preoperative testing for sepsis before revision TKA. J Arthroplasty, 2007; 22:90–93.

22.

Greidanus

, Masri

, Garbuz

, et al. Use of erythrocyte sedimentation rate and C-reactive protein level to diagnose infection before revision TKA. A prospective evaluation. J Bone Joint Surg Am, 2007; 89:1409–1416.

23.

Baré

, MacDonald

, Bourne

. Preoperative evaluations in revision TKA. Clin Orthop Relat Res, 2006; 446:40–44.

24.

Itasaka

, Kawai

, Sato

, et al. Diagnosis of infection after THA. J Orthop Sci, 2001; 6:320–326.

25.

Spangehl

, Masri

, O'Connell

, et al. Prospective analysis of preoperative and intraoperative investigations for the diagnosis of infection at the sites of two hundred and two revision total hip arthroplasties. J Bone Joint Surg Am, 1999; 8:672–683.

26.

Austin

, Ghanem

, Joshi

, et al. A simple, cost-effective screening protocol to rule out periprosthetic infection. J Arthroplasty, 2008; 23:65–68.

27.

Bottner

, Wegner

, Winkelmann

, et al. Interleukin-6, procalcitonin and TNF-alpha: Markers of peri-prosthetic infection following total joint replacement. J Bone Joint Surg Br, 2007; 89:94–99.

28.

Chevillotte

, Ali

, Trousdale

, et al. Inflammatory laboratory markers in periprosthetic hip fractures. J Arthroplasty, 2009; 24:722–727.

29.

Kanner

, Saleh

, Frierson

Jr . Reassessment of the usefulness of frozen section analysis for hip and knee joint revisions. Am J Clin Pathology, 2008; 130:363–368.

30.

Simonsen

, Buhl

, Oersnes

, Duus

. White blood cell scintigraphy for differentiation of infection and aseptic loosening: A retrospective study of 76 painful hip prostheses. Acta Orthop, 2007; 78:640–647.

31.

Nilsdotter-Augustinsson

, Briheim

, Herder

, et al. Inflammatory response in 85 patients with loosened hip prostheses: A prospective study comparing inflammatory markers in patients with aseptic and septic prosthetic loosening. Acta Orthop, 2007; 78:629–639.

32.

Bernard

, Lübbeke

, Stern

, et al. Value of preoperative investigations in diagnosing prosthetic joint infection: retrospective cohort study and literature review. Scand J Infect Dis, 2004; 36:410–416.

33.

Sanzén

, Carlsson

. The diagnostic value of C-reactive protein in infected total hip arthroplasties. J Bone Joint Surg Br, 1989; 71:638–641.

34.

Di Cesare

, Chang

, Preston

, et al. Serum interleukin-6 as a marker of periprosthetic infection following total hip and knee arthroplasty. J Bone Joint Surg Am, 2005; 87:1921–1927.

35.

Fink

, Makowiak

, Fuerst

, et al. The value of synovial biopsy, joint aspiration and C-reactive protein in the diagnosis of late peri-prosthetic infection of total knee replacements. J Bone Joint Surg Br, 2008; 90:874–878.

36.

Workgroup Convened by the Musculoskeletal Infection Society. New definition for periprosthetic joint infection. J Arthroplasty, 2011; 26:1136–1138.

37.

Parvizi

, Della Valle

. AAOS Clinical Practice Guideline: Diagnosis and treatment of periprosthetic joint infections of the hip and knee. J Am Acad Orthop Surg, 2010; 18:771–772.

38.

Swets

. Measuring the accuracy of diagnostic systems. Science, 1988; 240:1285–1293.

39.

Jones

, Athanasiou

. Summary receiver operating characteristic curve analysis techniques in the evaluation of diagnostic tests. Ann Thorac Surg, 2005; 79:16–20.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

1.19 MB

0.00 MB