Diagnostic Accuracy of FDG PET/CT in Suspected LVAD Infections: A Case Series, Systematic Review, and Meta-Analysis

Abstract

OBJECTIVES

The purpose of this study was to describe our experience with fluorine-18 fluorodeoxyglucose (FDG) positron emission tomography computed tomography (PET/CT) in diagnosing left ventricular assist device (LVAD) infections and perform a meta-analysis of published studies to determine overall diagnostic accuracy.

BACKGROUND

Device-related infections are a common complication of LVADs and are linked to worse outcomes. Diagnosis of LVAD infections remains challenging. FDG PET/CT has demonstrated good diagnostic accuracy in several other infectious conditions.

METHODS

This was a single-center, retrospective case series of FDG PET/CT scans in suspected LVAD infection between September 2015 and February 2018. A systematic review of PubMed from database inception through March 2018 was also conducted to identify additional studies.

RESULTS

Nineteen FDG PET/CT scans were identified for the retrospective case series. The systematic review identified an additional 3 publications, for a total of 4 studies involving 119 scans assessing diagnostic performance. Axial (n = 36) and centrifugal (n = 83) flow LVADs were represented. Pooled sensitivity was 92% (95% confidence interval [CI]: 82% to 97%) and specificity was 83% (95% CI: 24% to 99%) for FDG PET/CT in diagnosing LVAD infections. Summary receiver-operating characteristic curve analysis demonstrated an AUC of 0.94 (95% CI: 0.91 to 0.95).

CONCLUSIONS

FDG PET/CT for suspected LVAD infections demonstrates good diagnostic accuracy, with overall high sensitivity but variable specificity.

Durable left ventricular assist devices (LVADs) are an important therapeutic option in the management of end-stage heart failure, with approximately 3,500 new device implants per year in the United States (1,2). Among the most common LVAD complications are device-related infections, occurring at a rate of 13.6 per 100 patient-months during the first 3 months post-implant and 4.6 per 100 patient-months thereafter. The presence of an LVAD infection carries increased morbidity and mortality (1).

Unfortunately, although some definitions and diagnostic approaches for LVAD infections have been published, use and implementation remain challenging (3,4). LVAD-specific infections can be classified into pump and cannula, pump pocket, and superficial and deep driveline infections. Diagnostic confidence is enhanced by assessing multiple major and minor criteria that integrate clinical, microbiologic, and histologic criteria (Supplemental Table 1). Standard work-up of suspected LVAD infections include a complete blood count, chest radiography, and 3 sets of blood cultures over a 24-h period. If there is suspicion for pump or driveline infections, gram stain and culture of the site is recommended. Additionally, other potential sources of infection should be considered. In some cases, erythrocyte sedimentation rate and C-reactive protein may have value. Transthoracic echocardiography and transesophageal echocardiography (TEE), ultrasonography, computed tomography (CT), and tagged white blood cell scans may also be useful depending on clinical circumstances.

Fluorine-18 fluorodeoxyglucose (FDG) positron emission tomography computed tomography (PET/CT) is a sensitive, specific, and accurate test for the diagnosis of infection, including infections of cardiac implantable electronic devices and prosthetic valves (5–9). Although case reports and series have described FDG PET/CT findings in patients with LVADs, these studies are few in number (10–21). Furthermore, no systematic reviews exist specifically assessing FDG PET/CT use in the LVAD population. As such, no practice guidelines for use in suspected LVAD infections exist.

We sought to describe our experience with FDG PET/CT for the diagnosis of possible LVAD infections and systematically review the published reports regarding the diagnostic accuracy of FDG PET/CT scanning for suspected device infections in the LVAD population.

METHODS

CASE SERIES.

Selection.

This single-center case series is a retrospective study of consecutive FDG PET/CT scans performed on LVAD patients between September 2015 and February 2018 at the University of Michigan. Studies were identified through screening all cardiac FDG PET/CT scans protocoled to assess for inflammation in this time frame. Inclusion criterion was LVAD device at the time of scan. Exclusion criterion was scan not performed to assess for possible infection. Demographic data as well as pertinent history, laboratory values, imaging studies and clinical outcomes were obtained from medical record review. Possible, probable, or proven LVAD-specific infections (Supplemental Table 1) were grouped together as having a clinician-determined device-specific infection and served as a reference standard for diagnostic testing. These determinations were based on provider assessment during the initial care encounter and confirmed on 30-day follow-up.

This study was approved by the Institutional Review Board at the University of Michigan and waiver of informed consent was granted for this retrospective case series.

FDG PET/CT protocol.

A standardized protocol at the University of Michigan was used for inflammatory cardiac PET studies. Patients were instructed to maintain a high-fat, low-carbohydrate, protein-permitted diet for all meals beginning with breakfast on the day before the FDG PET/CT scan. On the morning of the study, after an overnight fast, patients underwent rubidium-82 PET/CT resting perfusion imaging to help delineate myocardial borders. Immediately after perfusion imaging, patients were given a high fat drink. After a minimum of 3 h, patients were injected with 8 to 10 mCi of FDG. Three boluses of 10 units/kg of unfractionated heparin were injected at 15-min intervals beginning 10 min before FDG injection. In some cases, the perfusion scan was performed the day before FDG PET for scheduling reasons. Approximately 1 h after FDG injection, patients underwent PET/CT imaging (Siemens mCT-64, Siemens Medical Imaging, Knoxville, Tennessee) including cardiac gating. When necessary, additional bed positions were acquired to cover the entire LVAD system and any other cardiac devices. In cases where clinical suspicion warranted, whole-body imaging was also performed. Attenuation corrected and non-attenuation corrected images were obtained. Physicians interpreting the studies were not blinded to clinical data.

SYSTEMATIC REVIEW.

Selection.

This systematic review and meta-analysis was conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) Statement guidelines (23). A PubMed search was performed for all publications that included the use of FDG PET/CT in diagnosing suspected LVAD infections, from database inception through March 2018 (Supplemental Appendix). After duplicate search results were removed, all abstracts were independently assessed by 2 reviewers (M.C.T., V.N.P.) for the desired diagnostic test and patient population. Full-text reviews of screened studies were assessed for inclusion and exclusion criteria. Single-patient case reports and studies that did not report test characteristics (i.e., sensitivity and specificity) were excluded. All but the most recent publication for a given center’s database were excluded. The reference lists of the selected articles were also screened for suitable publications.

Data extraction.

Data extraction was performed independently by 2 reviewers (M.C.T., V.N.P.), with any differences reconciled by mutual agreement. Information on study size, number of PET scans, type of LVAD, age, time from implantation, use of attenuation correction, and treatment for LVAD infection was collected. The principle measure obtained was results of PET scans defined as true positive, true negative, false positive, or false negative using a reference standard of clinician-determined final diagnosis.

Quality of evidence.

Risk of bias was assessed by 2 reviewers (M.C.T., V.N.P.) using the QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies 2) tool (24), with any differences reconciled by mutual agreement.

DATA ANALYSIS AND META-ANALYSIS.

Demographic and clinical data were assessed by standard descriptive statistics for comparison groups. Continuous variables were described as median (interquartile range) and between-group differences were tested for statistical significance with Mann-Whitney U tests. Categorical variables were described as frequencies, and between-group differences were tested with Fisher exact tests. All statistical tests were 2-sided with a 0.05 significance level. Clinician-determined possible, probable, or proven LVAD-specific infection were grouped together for analysis. For the case series, sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio were determined. For the meta-analysis, extracted data were used to determine pooled sensitivity, specificity, and summary receiver-operating characteristic (SROC) curves with area under the curve (AUC). A generalized linear mixed model approach for diagnostic test accuracy using random effects (25) was used with meta-analysis commands (26). Study heterogeneity was assessed by Higgins I² (27). Publication bias was assessed by Deeks’ funnel plot (28). Analyses were performed with Stata version 13.0 (StataCorp, College Station, Texas).

RESULTS

CASE SERIES.

Overall, 20 FDG PET/CT scans assessing for inflammation in LVAD patients were performed between September 2015 and February 2018. One scan was excluded because it was ordered to assess sarcoidosis activity and not infection. Therefore, 19 scans performed on 18 patients were included for analysis. One patient had 2 scans for separate care episodes. LVAD devices studied include axial (n = 7) and centrifugal (n = 12) continuous-flow pumps. Baseline data for the groups with (n = 11) and without (n = 8) LVAD infections are shown in Table 1. A statistically significant difference was noted with respect to age (p = 0.013). No differences were noted between the remaining demographic and history characteristics. Importantly, no between group differences were noted in traditional infectious markers such as white blood cell count, erythrocyte sedimentation rate, and C-reactive protein. There were also no differences seen in markers of adequate study preparation including insulin level, C-peptide, serum glucose, triglycerides, and free fatty acid.

TABLE 1.

Baseline Case Series Patient Data

	Total (N = 19)	No LVAD Infection (n = 8)	LVAD Infection (n = 11)	p Value
Age, yrs	61.0 (50, 65)	63.0 (61.5, 68.0)	56.0 (45.0, 59.0)	0.013*
Female	4(21.1)	3 (37.5)	1 (9.1)	0.26
Ethnicity				0.34
African American	5 (26.3)	1 (12.5)	4 (36.4)
Caucasian	14 (73.7)	7 (87.5)	7 (63.6)
Etiology of heart failure				0.32
Cardiac sarcoidosis	2 (10.5)	1 (12.5)	1 (9.1)
Ischemic	9 (47.4)	3 (37.5)	6 (54.5)
Mixed ischemic/nonischemic	2 (10.5)	0 (0.0)	2 (18.2)
Nonischemic	5 (26.3)	4 (50.0)	1 (9.1)
Familial/viral	1 (5.3)	0 (0.0)	1 (9.1)
Type of LVAD				0.63
Axial flow pump	7 (36.8)	2 (25.0)	5 (45.5)
Centrifugal flow pump	12 (63.2)	6 (75.0)	6 (54.6)
LVAD indication				0.67
Bridge to candidacy	1 (5.3)	0 (0.0)	1 (9.1)
Bridge to decision	3 (15.8)	2 (25.0)	1 (9.1)
Bridge to transplant	7 (36.8)	2 (25.0)	5 (45.5)
Destination therapy	8 (42.1)	4 (50.0)	4 (36.4)
LVAD implant to FDG PET/CT time, days	59.0 (189,1,234)	584.5 (479.5,1,357.5)	309.0 (105,1,234)	0.32
Prosthetic valve	11 (57.9)	4 (50.0)	7 (63.6)	0.66
Cardiac electronic implantable device	18 (94.7)	8 (100)	10 (90.9)	1.00
Diabetes	10 (52.6)	3 (37.5)	7 (63.6)	0.37
Insulin use	6 (31.6)	2 (25.0)	4 (33.3)	1.00
White blood cell count, K/μl	13.10 (7.70,19.60)	12.50 (8.70,17.35)	13.10 (6.6, 23.7)	0.51
Erythrocyte sedimentation rate, MM	28.5 (19.0, 44.0)	28.0 (23, 29)	44.0 (19, 78)	0.35
C-reactive protein, mg/dl	3.65 (1.50, 8.40)	1.50 (1.5, 7.9)	3.70 (3.6, 8.4)	0.60
Lactate dehydrogenase, IU/l	340.0 (259.0, 381.0)	363.0 (259.0, 376.5)	319.0 (259, 441)	0.90
Insulin level at time of scan, μIU/ml	21.40 (13.20, 32.90)	13.80 (11.9, 28.65)	22.70 (16.9, 34.3)	0.19
C-peptide at time of scan, ng/ml	5.5 (4.4, 6.9)	5.1 (4.0, 7.2)	5.5 (4.6, 6.9)	0.74
Glucose at time of scan, mg/dl	111.0 (86.0,143.0)	106.5 (88.5,126)	118.0 (86,147)	0.59
Triglycerides at time of scan, mg/dl	140 (114, 229)	146 (132.5, 278.5)	136 (113, 229)	0.51
Free fatty acid at time of scan, mmol/l	1.090 (0.98,1.300)	1.120 (1.015,1.36)	1.090 (0.94,1.30)	0.87

Values are n (%) or median (interquartile range [25th, 75th]). *Significance with alpha level of <0.05 between no infection and infection group. FDG PET/CT = fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography; LVAD = left ventricular assist device.

Patient-level clinical data for the nineteen scans were collected (Supplemental Table 2). Compared with clinician-determined presence or absence an LVAD infection, FDG PET/CT scanning produced 11 true positives, 2 true negatives, 6 false positives, and 0 false negatives. Figures 1 to 3 highlight examples of these scans. Of the 11 true positive scans, 6 cases were treated with antibiotics alone, 1 required LVAD exchange, 2 underwent eventual heart transplantation, 2 died, and 1 had a concurrent uterine abscess. Of the 2 true negatives, 1 had cholecystitis (Figure 2) and 1 was treated empirically with antibiotics without a clear source of infection. Of the 6 false positives, 1 received antibiotics for an empyema, 1 had a cardiac implantable electronic device extracted for infection and eventual unrelated heart transplantation, 1 had an aortic endograft infection, 1 had concomitant active cardiac sarcoidosis requiring steroids, 1 had no clear source of infection but nonspecific uptake along mitral and tricuspid prostheses and treated empirically with antibiotics, and 1 was newly diagnosed with chronic myeloid leukemia (Figure 3). At our center, FDG PET/CT for diagnosing LVAD infections had 100% (95% confidence interval [CI]: 72% to 100%) sensitivity and 25% (95% CI: 3.1% to 65%) specificity, with a positive likelihood ratio of 1.33 (95% CI: 0.89 to 2.00) and a negative likelihood ratio of 0.00.

FIGURE 1. True Positive Scan.

A 33-year-old humans with a centrifugal flow left ventricular assist device who presented with fatigue and fevers, and subsequently was found to have Streptococcus anginosus bacteremia. Note the increased radiolabeled uptake along the device pump (red arrows) and tricuspid valve ring (yellow arrows), indicating high metabolic activity and possible infection. The patient was treated for device infection with intravenous antibiotics for 6 weeks and then suppressive therapy.

FIGURE 2. True Negative Scan.

A 71-year-old humans with an axial flow left ventricular assist device who presented with chest pain, worsening leukocytosis, increasing lactate dehydrogenase, and hypotension. Blood cultures were negative. On non–attenuation-corrected images, there was no significant radiolabeled uptake around the device pump, cannula, or driveline (red arrows). There was intense uptake along the gallbladder (yellow arrows), indicating high metabolic activity and possible infection. There was no clinical device infection and the patient was treated for purulent cholecystitis.

FIGURE 3. False Positive Scan.

A 65-year-old humans with a centrifugal flow left ventricular assist device who presented with worsening leukocytosis. Blood cultures were negative. Note the increased uptake along the device pump (red arrows) that occur in an interrupted pattern. These regions correspond to typical locations of pledgeted surgical suture material. There were no other clinical signs of infection. The patient was deemed not to have a device-specific infection, but eventually diagnosed with new chronic myeloid leukemia.

TEE scans were obtained in 10 of the cases (6 true positive, 1 true negative, and 3 false positive FDG PET/CT scans). None demonstrated definitive intracardiac vegetations, with only 1 scan demonstrating a likely Lambl’s excrescence. CT scans were obtained in 12 of the cases (7 true positive, 2 true negative, and 3 false positive FDG PET/CT scans). In patients treated as an LVAD infection, CT scans were suggestive of LVAD infection in 2 of 7 cases. In patients not treated as an LVAD infection, none of the 5 CT scans showed evidence of LVAD infection, though they did point to other sources of infection in 2 cases.

SYSTEMATIC REVIEW.

Selection.

There were 57 publications identified through the PubMed search, with no duplicates (Figure 4). All abstracts were screened, and 45 studies were excluded because they did not involve both FDG PET/CT testing and the LVAD patient population. Full-text review of the remaining 12 articles was completed, with an additional 9 articles excluded. Of these 9 publications, 4 were single-patient reports (10,11,13,17), 2 were older publications (14,16) from an overlapping dataset of a more recent publication (18), and 3 did not include information about diagnostic test characteristics (12,20,21). Notably, large studies by Kanapinn et al. (20) (assessing utility of quantitative changes in serial FDG PET/CT scans in 30 nonconsecutive LVAD patients) and Kim et al. (21) (assessing clinical outcomes related to location of FDG uptake in 35 consecutive LVAD patients) were excluded because the studies were not designed to assess the accuracy of FDG PET/CT for LVAD infections.

FIGURE 4. PRISMA Flow Diagram.

The results of a PubMed database search for publications assessing diagnostic value of fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography for left ventricular assist device infections. PRISMA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

Quality of evidence.

The QUADAS-2 analysis (Supplemental Table 3) showed overall low risk for bias for patient selection, unclear risk for bias for index test and flow or timing, and high risk of bias for reference standard. The high risk for bias with the reference standard domain is due to lack of a true gold standard test for diagnosing LVAD infections. There is overall low concern for applicability in all domains.

Study characteristics and outcomes.

Included publications are summarized in Table 2, totaling 119 FDG PET/CT scans. Axial flow (n = 36) and centrifugal flow (n = 83) LVAD scans were represented in this group. Median time from LVAD implant to FDG PET/CT scan in the studies varied from 134 to 559 days. Overall, there were 6 false negatives, 12 false positive, 32 true negative, and 69 true positive scans. All but 1 study (15) commented on use of both nonattenuation and attenuation corrected images. Treatments and outcomes of for patients with LVAD infections included prolonged antibiotics (n = 11), surgical exploration or device exchange (n = 18), heart transplant (n = 29), and death (n = 14).

TABLE 2.

Pooled Data of Studies for FDG PET/CT for Diagnosis of LVAD Infections

First Author, Year (Ref. #)	N	Scans	LVAD Type	Time From Device Implant	Age (yrs)	Dietary Preparation Before Scan	Scan Result	Images Analyzed	Definitive Therapy for LVAD Infection
Akin et al., 2017 (19)	9	10	HeartMate II (n = 10)	Median 134 (range 24–645) days	Mean 54 ± 15	1-day LC diet, 6-h fast	True negative (n = 2) True positive (n = 8)	AC and NAC	Antibiotics (n = 1) Heart transplant (n = 4) Surgery (n = 2)
Bernhardt et al., 2017 (15)	21	29	HeartWare HVAD (n = 29)	Median 287 (IQR: 487) days	Mean 53.7 ± 14.3	4-h fast	False negative (n = 2) True negative (n = 13) True positive (n = 14)	AC	Antibiotics (n = 4) Heart transplant (n = 4) Surgery (n = 9)
Dell’Aquila et al., 2017 (18)	47	61	HeartMate II (n = 13) HeartWare HVAD (n = 40) Incor (n = 6) Ventracor (n = 2)	Median 13.02 (IQR: 5.21) months	Median 64.13 (IQR: 18.77)	6-h fast	False negative (n = 4) False positive (n = 6) True negative (n = 15) True positive (n = 36)	AC and NAC	Death (n =12) Heart transplant (n = 19) Surgery (n = 6)
Current series, 2018	18	19	Axial flow pump (n = 7) Centrifugal flow pump (n =12)	Median 559 (IQR: 1,045) days	Median 61 (IQR: 15)	1-day HF LC diet, overnight fast, HF drink 3+ h prior	False positive (n = 6) True negative (n = 2) True positive (n = 11)	AC and NAC	Antibiotics (n = 6) Death (n = 2) Heart transplant (n = 2) Surgery (n = 1)
Total	95	119	Axial flow pump (n = 36) Centrifugal flow pump (n = 83)	—	—	—	False negative (n = 6) False positive (n = 12) True negative (n = 32) True positive (n = 69)		Antibiotics (n = 11) Death (n =14) Heart transplant (n = 29) Surgery (n =18)

AC = attenuation corrected; HF = high fat; IQR = interquartile range; LC = low carbohydrate; N/A = not available; NAC = non–attenuation corrected; other abbreviations as in Table 1.

META-ANALYSIS.

Figure 5 shows the forest plots of individual study sensitivity and specificity. Pooled sensitivity was 92% (95% CI: 82% to 97%) and specificity was 83% (95% CI: 24% to 99%) for FDG PET/CT in diagnosing LVAD infections. SROC analysis demonstrates an AUC of 0.94 (95% CI: 0.91 to 0.95) (Central Illustration).

FIGURE 5. Forest Plot.

The pooled sensitivity and specificity of individual studies assessing diagnostic accuracy of fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography for left ventricular assist device infections with measures of heterogeneity. CI = confidence interval.

CENTRAL ILLUSTRATION. Summary Receiver-Operating Characteristic Curve.

The summary characteristics and plot of studies included in the meta-analysis. AUC = AUC; SENS = pooled sensitivity; SPEC = pooled specificity; SROC = summary receiver-operating characteristic.

Heterogeneity between studies as assessed by the Higgins I² statistic was 0.00% (95% CI: 0.00% to 100.0’%) for sensitivity and 80.71% (95% CI: 62.08% to 99.33%) for specificity (Figure 5). Publication bias was assessed with a Deeks’ funnel plot (p = 0.82) (Supplemental Figure 1).

DISCUSSION

DIAGNOSTIC VALUE OF FDG PET/CT IN LVAD INFECTIONS.

This is the first systematic review and meta-analysis assessing the diagnostic accuracy of FDG PET/CT in LVAD infections. Overall, FDG PET/CT performed well despite a limited number of studies. There was a high diagnostic accuracy, with a SROC curve AUC of 0.94, indicating a potential expanded role for this imaging technology in assessing suspected LVAD infections. Of interest, there was a notably high sensitivity based on this case series (100%) and the meta-analysis (92%), suggesting that this modality may be particularly useful in ruling-out infections when there is a low to intermediate pretest probability. Specificity was lower for the case series (25%), though it was higher in the meta-analysis (83%) with wide CIs. This suggests less usefulness for ruling in infections. Overall, this meta-analysis assessed LVAD patients with a wide range of baseline characteristics, enhancing generalizability.

FACTORS INFLUENCING DIAGNOSTIC VALUE OF FDG PET/CT IN LVAD INFECTIONS.

Inflammatory cells have increased glucose metabolism and concentrate FDG. Therefore, increased FDG uptake in PET/CT scanning is seen in noninfectious inflammatory conditions as well, such as postoperative states, myocarditis, and sarcoidosis (29,30). Patient factors such as time from LVAD implant and etiology of heart failure need to be taken into consideration when interpreting the results of these scans. Normal myocytes also utilize glucose as a metabolic substrate. In the absence of adequate patient metabolic preparation before FDG PET/CT scanning, normal cardiac myocytes can concentrate FDG, which can alter test specificity. Careful metabolic preparation needs to be taken to drive myocytes toward fatty acid, rather than glucose, metabolism to suppress any background FDG uptake in the myocardium. Levels of insulin, C-peptide, serum glucose, triglycerides, and free fatty acid were measured in this case series to assess for adequate metabolic preparation. No significant differences were noted between groups, suggesting no difference in metabolic preparation between groups studied in this case series.

Of note, this case series includes 6 false positive scans, with 4 having concurrent bacteremia or other possible sources of infection, 1 having concurrent increased cardiac sarcoidosis activity, and 1 having a new diagnosis of chronic myeloid leukemia. It remains unclear if these concurrent infectious and inflammatory conditions play a role in the increased FDG uptake near the components of the LVAD but could contribute to false positive scans.

Antibiotic use before the imaging may also play a role in attenuating the degree of FDG uptake in areas of infection. This has the potential for increasing the rates of false negatives and reducing test sensitivity. While it is unclear the exact impact of antibiotic use and duration on the results of testing, rates of false negatives remained low, at 5.0% (6 of 119 scans) in this meta-analysis, and pooled sensitivity remained high.

Attenuation correction is used to enhance image quality and allow for better quantification of FDG uptake in PET/CT scans, but also can produce artifacts around metallic material such as an LVAD due to overcorrection for attenuation and subsequent over-estimation of FDG PET/CT activity (31). This is a source of false positive studies, though FDG uptake around metallic objects should be confirmed on nonattenuation corrected images as they are less prone to manifest the same artifact. Protocols for interpretation of FDG PET/CT studies in the LVAD population should include review of the nonattenuation corrected images for this reason.

Study heterogeneity as determined by the Higgins I² statistic for diagnostic testing is not an absolute measure, but increasing values indicate increasing likelihood of heterogeneity (32,33). For the studies included in the meta-analysis, heterogeneity was low with respect to sensitivity (I² = 0.00%), but given the wide CI, the significance is unclear. With respect to specificity (I² = 80.71%), heterogeneity was high. Although this may reflect random chance, methodology errors, or study design differences, there is likely a component of a threshold effect. Sensitivity and specificity are linked variables that change inversely based on the threshold for relative FDG uptake. Currently no universally accepted standard has been set, which may limit the current utility of this testing. The limited number of studies also leads to an underpowered analysis.

INTEGRATING FDG PET/CT WITH OTHER CLINICAL DATA IN DIAGNOSING LVAD INFECTIONS.

Suspected LVAD infections can be difficult to diagnose and often require integrating clinical and testing data (3,4). As noted in this case series, traditional blood biomarkers of infection (Table 1) and imaging with TEE and CT (Supplemental Table 2) were poor predictors for LVAD infection when viewed as standalone tests, further highlighting the need for a combined approach. FDG PET/CT can significantly add to current practice given the high diagnostic accuracy (AUC: 0.94) and pooled sensitivity (92%) found in this meta-analysis. Making an accurate diagnosis in this setting is important given the increased mortality with LVAD infections and the implications of subsequent therapy (prolonged antimicrobial therapy, LVAD exchange, heart transplant).

A proposed algorithm for incorporating FDG PET/CT testing is outlined here. For all patients with suspected LVAD infection, usual work-up includes a pertinent history and physical, blood biomarkers for infection, blood cultures, wound cultures if purulence is present at the driveline exit site, chest radiograph, and additional imaging if there is suspicion for a pump or intracardiac infection (transthoracic echocardiography with possible TEE, ultrasonography, or CT of the chest or abdomen). Based on this initial clinical information, cases of proven LVAD infection can proceed to appropriate treatment without further testing and cases of unlikely device infection should prompt a search for an alternate diagnosis. In the remaining cases where there is diagnostic uncertainty and risks of surgical exploration are high, then FDG PET/CT may be obtained. Positive results would indicate need for treatment of a LVAD infection and negative scans would indicate an alternate diagnosis. In some situations, incidental FDG uptake can be seen in non-LVAD components, leading to appropriate therapy for a different diagnosis.

STUDY LIMITATIONS.

For the case series, a limitation is the small sample size, which is not necessarily powered to detect difference in the baseline characteristics described. Therefore, the significance of certain markers in assessing for LVAD infections may be underestimated. For the meta-analysis, sensitivity analyses including tests of heterogeneity and publication bias are also underpowered with the number of studies included.

With the systematic review and meta-analysis, 1 limitation is the variation in dietary or metabolic preparation and imaging protocols used by different centers, which can alter the accuracy of testing. This likely contributed to the study heterogeneity. For our cases series, the dietary preparation used aligns with current data for effective myocardial suppression (22).

In general, selection bias is also important to consider. It is unlikely that all patients with suspected LVAD infections receive FDG PET/CT scans, leading to potential errors in assessing diagnostic accuracy. Similarly, no patients were evaluated who did not have suspected infection.

The lack of a true reference standard test for diagnosing LVAD infections also poses a major limitation in cases where no surgeries are performed or when surgical specimens at time of device explant are not feasible. Therefore, the diagnosis of infection relies on several clinical, laboratory, and imaging factors, which may lead to errors in determining the true diagnostic accuracy of any single test or imaging modality. Furthermore, for the cases included in the systematic review, interpretation of FDG PET/CT was not necessarily blinded from other clinical data and may have influenced the outcome of the scan.

Finally, the retrospective nature of this study also leads to potential confounding variables that may have affected the results. Further prospective studies are needed to address these issues.

CONCLUSIONS

In this case series and systematic review with meta-analysis of all pertinent studies on the use of FDG PET/CT in diagnosing LVAD infections, FDG PET/CT demonstrated good diagnostic accuracy, with overall high sensitivity but variable specificity.

PERSPECTIVES.

COMPETENCY IN MEDICAL KNOWLEDGE:

Assessing for suspected infection with FDG PET/CT scanning has significant clinical implications for the management of LVAD patients, given that making the diagnosis is challenging. Integration of FDG PET/CT into the current diagnostic algorithm provides an accurate and sensitive test of infection.

TRANSLATIONAL OUTLOOK:

Current practices for assessing suspected LVAD infections vary and advanced imaging such as FDG PET/CT scan should be considered. Diagnostic accuracy and sensitivity of this test is high, but future studies should focus on improving the specificity of the test to have even greater clinical importance.

ABBREVIATIONS AND ACRONYMS

AUC

area under the curve

CI

confidence interval

CT

computed tomography

FDG

fluorine-18 fluorodeoxyglucose

LVAD

left ventricular assist device

PET

positron emission tomography

SROC

summary receiver-operating characteristic

TEE

transesophageal echocardiography