A systematic review and meta-analysis of 18F-FDG-PET interpretation methods in vascular graft and endograft infection

Citation for published version (APA): Reinders Folmer, E. I., von Meijenfeldt, G. C. I., Te Riet Ook Genaamd Scholten, R. S., van der Laan, M. J., Glaudemans, A. W. J. M., Slart, R. H. J. A., Zeebregts, C. J., & Saleem, B. R. (2020). A systematic review and meta-analysis of 18F-FDG-PET interpretation methods in vascular graft and endograft infection. Journal of Vascular Surgery. https://doi.org/10.1016/j.jvs.2020.05.065


12
The pattern of uptake as an interpretation method of 18 F-FDG PET(/CT) in diagnosing VGEI 13 showed the highest pooled sensitivity and specificity in this meta-analysis, compared to the FDG 14 uptake intensity and SUVmax methods. Therefore, the pattern of uptake method appears to have 15 the best discriminative ability in diagnosing VGEI. tomography as the current standard. However, the availability and varied use of several 1 interpretation methods, without consensus on which interpretation method is best, complicates 2 clinical use. The aim of this study was to evaluate the diagnostic performance of different 3 interpretation methods of 18 F-FDG PET(/CT) in diagnosing VGEI.

17
The uptake pattern interpretation method demonstrated the best positive and negative post-test 18 probability-82% and 10%, respectively.

19
Conclusion 20 This meta-analysis identified the FDG uptake pattern as the most accurate assessment method of  Incidence depends on the anatomical region and the technique used for implementation of the 19 graft. 2,3,4,5,6 Studies report an overall incidence between 0.1-6%, whereas the incidence in initial 20 endovascularly treated patients is much lower, at 0.1-1. 2%. 7,8,9 Grafts located in the groin appear 21 to be infected most often (6%). 10

23
Diagnosis of VGEI is complicated since symptoms vary and can be non-specific. 11,12 Positive 1 cultures, which can be obtained percutaneously or during surgery, are still considered the 2 reference standard. However, retrieving material for culture is not possible in all patients either 3 because it cannot be obtained percutaneously, the material is contaminated, or because a surgical 4 procedure is too invasive for the patient. Confirming the diagnosis can thus be difficult or 5 impossible.

7
Although surgical intervention is the preferred treatment, not all patients are in a medical 8 condition conducive to major surgery. Therefore, some patients are treated conservatively with 9 antibiotics. Since surgical treatment is invasive and carries risk, a correct diagnosis or exclusion 10 of graft infection is of great importance. 13

12
Over the last three decades, several imaging modalities have been used to non-invasively 13 diagnose VGEI, with a wide range of sensitivity and specificity. 14 The current nuclear hybrid 14 imaging techniques are promising for diagnosis of a vascular graft and/or endograft infection. anatomic localization of an existing infection and has the ability to tell if the graft is involved in 18 the infectious process. 14 Early and correct diagnosis is of major importance, since false negatives 19 may be fatal and false positives may result in overtreatment with potentially major consequences.

20
A recent meta-analysis of the diagnostic performance of different imaging modalities used in 21 patients suspected of VGEI described the accuracy of the existing imaging techniques. 18 F-FDG-

22
PET scans (with or without low-dose CT) yielded high sensitivity and specificity, whereas the 23 6 results were marginal for computerized tomography with or without angiography (CT(A)). 15 1 2 For FDG-PET(/CT) in the context of VGEI, no clear interpretation criteria exist; different 3 assessment methods are used to score 18 F-FDG PET(/CT) findings. These assessment methods 4 include visual scoring based on either (i) the uptake intensity of FDG, often assessed by a visual 5 grading scale (VGS); or (ii) the uptake pattern of the FDG, i.e., whether it is focal or diffuse.

6
When the uptake intensity is assessed, the amount of FDG uptake is quantified, often using a 7 five-point VGS. The pattern of FDG uptake can be evenly diffusely distributed (homogeneous) 8 or focally distributed (heterogeneous). VGEI often exhibits higher uptake intensity and a more 9 focal pattern of FDG uptake.

10
The other assessment methods involve semi-quantitative scoring by (iii) quantifying 11 the maximum standardized uptake value (SUVmax), corresponding with the highest FDG signal; 12 and (iv) calculating the tissue-to-background ratio (TBR) by dividing the SUVmax of the graft 13 by the SUVmean of the reference organ-e.g., the SUVmean of the liver, bladder, or caval vein 14 (blood pool).

16
The aim of this study is to identify the most optimal interpretation method of 18 F-FDG PET(/CT) 17 for diagnosis of VGEI.

21
The Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-22 P) 2015 statement and the Cochrane Handbook for Diagnostic Test Accuracy Reviews were 23 followed to conduct this meta-analysis. 16 The objective of this review was to assess the diagnostic value of the four different assessment 4 methods of 18 F-FDG PET(/CT) used in the diagnostic work-up of patients with suspected VGEI.  retrospective studies. Studies with fewer than five patients, case reports, abstracts, reviews, or 5 animal studies were excluded. Checks for duplicates and overlapping databases were performed 6 both electronically and manually. outcome data (true positives, false positives, false negatives, true negatives) were extracted. If 20 necessary, authors were contacted to obtain missing data. When data and/or vital information on 21 inclusion criteria was unavailable, that particular study was excluded from the analysis. Assessment of study quality 1 Two reviewers (ERF and RtRS) evaluated the methodological quality of the included 2 observational cohort studies. The Quality Assessment of Diagnostic Accuracy Studies tool 3 (QUADAS-2) was used to assess the applicability and risk of bias. 18 Several categories were 4 labeled as 'low risk, 'high risk,' or 'unclear risk,' such as patient selection, the index test, the 5 reference standard, and flow and timing. The available data was separated per method of assessment. All assessment methods with 9 sufficient available data were further analyzed during meta-analysis. Sensitivity and specificity 10 forest plots were drawn using RevMan version 5.3.3. 19 Pooled sensitivities and specificities were 11 calculated by using 2x2 contingency tables. The heterogeneity among the studies was evaluated 12 using Chi 2 and I 2 statistics and drawn in hierarchical summary receiver operating characteristics 13 (HSROC) curves. The I 2 statistic was interpreted as follows: 0% to 40% was considered not 14 important, 30% to 60% represented moderate heterogeneity, 50% to 90% represented substantial 15 heterogeneity, and 75% to 100% indicated considerable heterogeneity. 20  SUVmax, and (iv) TBR. Table 1 gives an overview of the study characteristics, patient 19 characteristics, index test, and reference standards of the included studies. Only two studies 20 investigated the TBR and therefore we could not perform a meta-analysis of this category. 23,24

21
As a reference standard, microbiological assessment was used for VGEI in all of the studies.

22
Clinical follow-up was used in more than half of the studies and ranged from 4 to 36 months. The number and anatomical location of the vascular grafts and endografts are shown in Table 1. 6 Six studies only included patients with central grafts; the other seven studies included peripheral 7 grafts as well. Aortic grafts were classified as thoracic, abdominal, or both. reference standard other than microbiological culture, such as follow-up, was considered to 3 confer a high risk of bias. Therefore, several studies scored as "high risk" of bias in the category 4 "flow and timing" and "reference standard."  The forest plots of the sensitivities and specificities of the different interpretation methods are 1 shown in Figure 5, with the confidence intervals (CI) given per study. The pooled diagnostic 2 odds ratios are shown in Figure 6. FDG uptake intensity exhibits lower specificity, which means a higher number of false 17 negatives. Therefore, the discriminative ability of the FDG uptake pattern and SUVmax appears 18 superior to the FDG uptake intensity method.

20
Pre-and post-test probabilities 21 To interpret the value of a positive or negative test result of one of the three interpretation 22 methods, the pre-and post-test probabilities were calculated (Figure 7). The pre-test 23 probabilities of all three interpretation methods are high, as the included studies comprised 1 patients who were already suspected of VGEI and not a random cohort of patients with a 2 vascular prosthesis in situ. For example, a patient suspected of VGEI had a risk of 52% of having 3 VGEI prior to the 18 F-FDG PET(/CT) using the uptake pattern interpretation method (pre-test 4 probability). After a positive scan, the risk of actually having VGEI is 82% (positive post-test 5 probability). If the test is negative, the risk of having VGEI anyway is 10% (negative post-test 6 probability). The uptake pattern interpretation method had the highest positive post-test probability (82%), 9 followed by the SUVmax method (77%) and FDG uptake intensity method (75%). The uptake 10 pattern method had the lowest negative post-test probability of 10%, which corresponds with 11 having the highest pooled specificity (0.81) of the three interpretation methods.   Although VGS is often used as an uptake intensity interpretation method in diagnosing VGEI, 7 the pooled outcome demonstrated the lowest accuracy. The pattern of uptake method showed the 8 highest accuracy of the three interpretation methods and hence appears to have the best 9 discriminative ability, resulting in fewer missed diagnoses and fewer over-treated patients. Since only two articles were available on the TBR, this data could not be pooled and is therefore 15 not included in the meta-analysis.  The pattern of uptake method group exhibited negligible heterogeneity among the included 2 studies. However, moderate heterogeneity was seen in the FDG uptake intensity and SUVmax 3 groups; therefore, the pooled diagnostic sensitivity and specificity should be interpreted with 4 caution.

6
Comparing the diagnostic performance of different interpretation methods of 18 F-FDG PET(/CT) 7 in suspected VGEI has its limitations. Being a rare complication, the number of patients per 8 included study is limited. Among the included studies, no randomized controlled trials could be 9 included. The ten included studies were all observational cohort studies, of which all but three 10 used prospective methodology. Diagnostic test accuracy reviews often show high heterogeneity 11 among the included studies, since patient and study characteristics differ. Fortunately, only 12 moderate heterogeneity was seen between two groups in this meta-analysis.  The influence of antibiotics on the number of false-negatives is a repeatedly discussed dilemma. Since this meta-analysis supports the hypothesis that heterogeneous, focal, and high 18 F-FDG 13 uptake is associated with infection, the uptake pattern of 18 F-FDG activity may help identify 14 VGEI with higher diagnostic precision. Another tool for quantifying distribution is textural and could therefore not be included in this meta-analysis.    ('blood vessel'/exp OR 'blood vessel':ab,ti OR vascular*:ab,ti OR aort*:ab,ti) AND ('blood vessel graft'/exp OR 'blood vessel prosthesis'/exp OR graft*:ab,ti OR prosthesis:ab,ti OR prosthet*:ab,ti) AND ('infection'/exp OR 'q fever':ab,ti OR 'q-fever':ab,ti) AND ('positron emission tomography'/exp OR 'positron emission tomograph*':ab,ti OR 'positron-emission tomograph*':ab,ti OR pet*:ab,ti OR 'fluorodeoxyglucose f 18'/exp OR 'Fluorodeoxyglucose F18':ab,ti OR 'fluor-18-deoxyglucose':ab,ti OR '18f-fdg':ab,ti OR fdg:ab,ti OR 18fdg:ab,ti)