Introduction
Periprosthetic joint infection (PJI) is regarded as a destructive complication after total joint replacement (TJA) [
1], accounting for approximately 15% and 25% of the failure factors of total hip and knee arthroplasty, respectively [
2,
3]. With the extension of the average life span of human beings, there are an increasing number of TJA operations, and the resulting number of PJI patients has also been increasing year by year [
4]. The existence of PJI seriously reduces the quality of life of patients, exacerbates the difficulty of treatment for orthopedic surgeons, and aggravates the financial burden of the national health system [
5‐
7]. Moreover, it is extremely difficult to accurately diagnose PJI, due to the atypical symptoms of many PJI patients [
8]. Many scholars jointly established the American Academy of Orthopedic Surgeon (AAOS)’s guidelines, Musculoskeletal Infection Society (MSIS), and International Consensus Meeting (ICM) diagnostic criteria in 2010, 2011, and 2013 respectively, which have been widely used [
9‐
11]. However, currently, there is still no diagnostic standard or indicators that can achieve 100% diagnostic accuracy, and some standards including synovial fluid detection are invasive, expensive, and inconvenient to obtain [
12]. Given this situation, there is an urgent demand for joint surgeons to look for diagnostic markers with efficiency, cheapness, and convenience [
13]. D-dimer is a fibrin degradation product formed by fibrinolysis of fibrin clots, which reflects the state of blood coagulation and increases in systemic or local infections, thrombosis, and neoplastic diseases [
14‐
16]. Based on this principle, Shahi et al. and Li et al. speculated that D-dimer in PJI patients might serve as a neoteric diagnostic biomarker [
17,
18]. Nevertheless, their findings are contradictory. There is also no agreement on the conclusions of the other similar studies. Therefore, we conducted a systematic review and meta-analysis of these literatures to study the diagnostic value of D-dimer in PJI.
Methods
Our study is carried out strictly in accordance with the criterion of the preferred report items of systematic review and meta-analysis report. The research scheme is determined by all the authors, and the steps of literature retrieval, literature quality evaluation, data statistics, result merging, and report writing are completed in turn.
Retrieval strategy
Under the guidance of the Cochrane Review method, two authors (Haitao Zhang and Xingyang Zhu) search online databases such as PubMed, Embase, Web of Science, and Cochrane Library. The search subject words and Mesh words are as follows: “periprosthetic joint infection” or “prosthesis-related infections” stands for disease, “D-dimer” or “D-dimer fibrin” or “D-dimer fragments” or “fibrin fragment D1 dimer” or “fibrin fragment DD” or “fibrin fragment D-dimer” represents target index. The range of retrieval dates is from the establishment of the database to April 2020. When searching, we only include English literature. After database screening, we manually searched some of the references included in the literature to obtain the valuable literature for this study.
Study selection
The title, abstract, or full text of all search results are reviewed by two censors (Haitao Zhang and Xiaobo Sun) in detail. When there is still a disagreement between the two examiners after reading the full text, it will be left to Professor Yirong Zeng to make the final decision. The inclusion criteria of the literature are as follows: (1) using D-dimer as an index for the diagnosis of PJI, (2) have integrated data (including true positive, false negative, false positive, and true negative) to construct a 2 × 2 table, and (3) there is a definite gold standard such as Musculoskeletal Infection Society (MSIS) or International consensus on infection (ICM) to compare the diagnostic accuracy with D-dimer.
Data extraction and quality assessment
The following data extraction and literature quality evaluation are completed by Pengfei Xin and Ke Jie, two researchers who are familiar with the knowledge of statistics, back to back, and the extracted data are input into a table in Excel. The data include the following: the author of the study, the year, the country or region in which the article was published, the design type of the study, the number of cases, sex ratio, patient’s age and BIM index, the gold standard used in the study, the detection method, and the cutoff value of D-dimer. In addition, the diagnostic accuracy of D-dimer (AUC) and the true positive, false positive, true negative, and false negative data used to construct 2 × 2 table were also recorded in detail.
The Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) in the Revman (version5.3) software was used to evaluate the quality of all the literature included in the study. QUADAS-2 is an updated version of the original QUADAS, including four aspects of patient selection, index test, reference standard, and flow and timing, which has a more accurate bias level and applicability to the original research than the original QUADAS.
Statistical analysis
All data analysis and picture production are carried out by using the commands in the Stata14.0 software. The bivariate random effect model was selected to analyze the tp, fp, fn, and tn values of 2 × 2 table extracted in the study and to test the heterogeneity. The sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic score, and diagnostic odds ratio (DOR) were obtained after integration. Among them, the higher the value of DOR, the higher the diagnostic value.
Additionally, the summary receiver operating characteristic (SROC) was drawn by the Midas command, and the area under the curve (AUC) was calculated. AUC represents the diagnostic accuracy of D-dimer.
The heterogeneity is expressed as the inconsistency index (I2) statistic, the smaller the I2, the smaller the heterogeneity. When I2 is 75%, 50%, and 25% respectively, it corresponds to large, medium, and small literature heterogeneity, respectively. If the heterogeneity is large, meta-regression and subgroup analysis are performed to identify the source of heterogeneity. We believe that the variables that may affect the heterogeneity are the type of study design, the threshold used in the study, the number of cases, the index of diagnostic gold, the sample type, and the country or region in which the literature is published.
In order to definitely judge the publication bias, the funnel chart (Deeks’ funnel plot) was drawn. Besides, the change of the diagnostic value of D-dimer on the incidence of PJI can be clearly shown by drawing a Fagan plot diagram.
Discussion
Due to the considerable cost of treating PJI, it has caused a growing number of orthopedists to pay attention to this daunting disease [
26]. As a prerequisite for determining the therapeutic regimen, an accurate diagnosis of PJI is urgent and necessary for us. As a matter of fact, it is quite arduous to accurately distinguish between aseptic loose artificial joints and PJI joints. This is because a biofilm is often formed on the surface of the prosthesis in patients with PJI, and the culture of pathogens sometimes shows negative results. Furthermore, when PJI patients present with chronic deep infection, there is no significant difference in clinical characteristics compared with aseptic loosened joints [
27,
28]. It is gratifying that in recent years, an increasing number of biomarkers for the diagnosis of PJI have been found, including synovial quantitative alpha-defensin, serological white blood cell count (WBC), erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) interleukin-6 (IL-6), and procalcitonin [
29]. Among them, ESR and CRP have been recommended by the American Infectious Diseases Association (IDSA) to conduct routine screening among all suspected PJI patients [
30].
Recently, D-dimer has been proposed by a number of experts as a novel serum marker. D-dimer is a fibrin degradation product of disseminated intravascular coagulation and has been widely used in the diagnosis of venous thromboembolism (VTE) and infection in patients [
31,
32]. According to Ribera et al., the level of synovial D-dimer is higher than normal in foal infectious joint disease [
14]. The results of the study done by Bytniewski et al. show that the level of D-dimer in patients with early postoperative TJA changes faster than that of ESR and CRP, and can rise rapidly and return to normal level in a short time [
33]. Subsequently, the diagnostic value of D-dimer in patients with PJI began to be valued by joint surgeons. Shahi et al. put forward the threshold of D-dimer (850 ng/ml) for the first time and considered that serum D-dimer has high sensitivity (89%) and specificity (93%). It is a valuable biomarker for the diagnosis of PJI [
18]. Qin et al. used D-dimer in combination with ESR or CRP to diagnose PJI and found that combined use had higher diagnostic performance than a single test [
22]. Since then, several scholars have published controversial research results.
Although a recent meta-analysis on circulating D-dimer versus fibrinogen in the diagnosis of PJI has been published, we believe that one of their included literatures is inconsistent with other subjects and should not be included in the meta-analysis. Because its research content is about D-dimer predicts persistent infection before reimplantation in two-stage exchange arthroplasty for PJI [
34]. After excluding this study, they included only 5 studies that pooled the accuracy of D-dimer diagnosis. As a result, we suspect that their conclusion is unreliable. Compared with Zhang et al. [
35], our meta-analysis included more studies and patients (1592 patients in 9 studies) after rigorous screening and literature quality evaluation. The overall pooled AUC of D-dimer for diagnosis PJI was 0.85 (95% CI, 0.82~0.88), which is higher than the AUC of D-dimer calculated by Zhang et al. (AUC = 0.74). As a consequence, we deem that D-dimer has a good diagnostic value in PJI, which is inconsistent with the research conclusion of Zhang et al. [
35]. In addition, the pooled sensitivity and specificity were calculated to be 0.82 and 0.73 respectively. From another point of view, this is equivalent to a higher false positive rate (27%) and a lower false negative rate (18%) in D-dimer for PJI diagnosis. The LR and DOR usually indicate the effectiveness of diagnostic indicators in clinical practice [
36]. A guideline defines that LR + > 2, LR − < 0.5, or DOR > 4 is considered a viable predictor, and LR + > 5, LR − < 0.2, or DOR > 10 is considered a good predictor [
37]. Therefore, in terms of LR, D-dimer is a predictable index for the diagnosis of PJI, and D-dimer is a good predictive index when DOR is used as a reference parameter. Posttest probability is another parameter widely used by clinicians, including positive predictive value and negative predictive value. It reflects the probability of patients with PJI when the test results are negative or positive. The Fagan plot diagram shows that the ability of D-dimer to distinguish between PJI is general.
There is great heterogeneity in the studies of Zhang et al., but they did not provide a reasonable explanation [
35]. Another advantage of this study is that we performed reasonable meta-regression and novel subgroup analysis to find the source of heterogeneity in this study. We originally considered the threshold as the main cause of heterogeneity, but the results were unexpected. In the subgroup analysis, the AUC of the same threshold group was completely consistent with the AUC of overall studies, and DOR was almost the same. We found that in the study design, the AUC and DOR of the retrospective group fluctuated downward, while the AUC and DOR of the prospective group were almost twice as high as those of all studies. Furthermore, the results of the serum group were similar with those of the prospective group. In meta-regression, the heterogeneity source of sensitivity is obviously reflected in the sample type (serum or plasma), while the specific heterogeneity source is mainly embodied in the country of study (China and USA). Therefore, we believe that the study design, sample type, and the country of study are the main factors that affect the diagnostic accuracy of D-dimer.
Admittedly, there are certain limitations in our research. First, a handful of studies have not excluded patients with inflammatory diseases, which, as mentioned above, affect D-dimer levels. Second, there is still no gold standard for the detection of PJI, and the gold standard test used in the study is only approximate. Several positive patients still miss diagnosis because the gold standard cannot detect. Third, due to the incompleteness of the original data, it is impossible to calculate the optimal cutoff value of D-dimer and conduct more detailed subgroup analysis such as the location of the joint.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.