Role and interpretation of fluorodeoxyglucose-positron emission tomography/computed tomography in HIV-infected patients with fever of unknown origin: a prospective study


Correspondence: Charlotte Martin, CHU Saint-Pierre, Infectious Diseases Dept, 322 Rue Haute, 1000 Brussels, Belgium. Tel: 00 32 2535 41 30; fax: 00 32 2539 36 14; e-mail:



The aim of the study was to evaluate prospectively the usefulness of fluorodeoxyglucose-positron emission tomography/computed tomography (FDG-PET/CT) in investigation of fever of unknown origin (FUO) in HIV-positive patients and to determine whether HIV viraemia impacts on FDG-PET/CT performance.


The FDG-PET/CT results of 20 HIV-infected patients with FUO were analysed and compared with the FDG-PET/CT results of 10 HIV-infected viraemic patients without FUO. The performance of FDG-PET/CT for identifying the aetiology of FUO was assessed. Final diagnosis for FUO was based on histopathology, microbiological assays, or clinical and imaging follow-up.


FDG-PET/CT contributed to the diagnosis or exclusion of a focal aetiology of the febrile state in 80% of patients with FUO. The presence of increased FDG uptake in the central lymph node has 100% specificity for focal aetiology of fever, even in viraemic patients. The absence of hypermetabolic central lymph nodes in FUO patients has 100% negative predictive value for focal disease. Lymph node biopsy in central hypermetabolic areas allowed, in 100% of cases, identification of underlying disease in patients with FUO. Biopsy of peripheral lymph nodes should be performed in lymph nodes with maximum standardized uptake value (SUVmax) ≥ 6–8 (sensitivity 62.5%; specificity 75%) and avoided in lymph nodes with SUVmax = 0–4 (specificity 0%). High HIV viraemia does not prevent correct interpretation of FDG-PET/CT.


As in HIV-negative patients, we confirm the usefulness of FDG-PET/CT in investigation of FUO in HIV-positive patients even if they are viraemic.


Fever of unknown origin (FUO) is a challenging clinical entity frequently encountered in HIV-positive patients. As in nonimmunosuppressed patients, the aetiology of FUO can be infectious, inflammatory, allergic or malignant. In HIV-infected patients, infectious aetiologies account for the majority of FUO [1-3] but may have atypical clinical presentation, which may delay the diagnosis. This has been well described for tuberculosis, one of the most frequent causes of FUO in HIV-infected patients, for whom conventional radiology is less sensitive and who present with extrapulmonary involvement in 60 to 70% of cases [4].

Inflammatory cells such as macrophages, lymphocytes, neutrophils and even fibroblasts have been shown to be avid for fluorodeoxyglucose (FDG) following stimulation by multiple cytokines, which increases the number of cell surface glucose transporters and hence increases the level of glycolysis. This explains the recognized role of FDG-positron emission tomography/computed tomography (PET/CT) in non-oncological indications for several years [5, 13-15].

The advantages of the technique are a whole-body evaluation with available results within 4 hours, potentially allowing accurate localization of the etiology of the problem.

FDG-PET(/CT) has been well validated in the work-up of FUO in HIV-negative patients [6-10], but only two small studies have been performed in HIV-positive patients with FUO [11, 12], one of which was conducted in the pre-highly active antiretroviral therapy (HAART) era [12].

Several studies on FDG-PET/CT in HIV-infected patients have shown increased FDG uptake in lymph nodes that could reflect sites of HIV replication [16-22]. Some of these studies suggest that asymptomatic HIV-infected patients with fully suppressed HIV viral loads (VLs) have no or little FDG uptake in lymph nodes, whereas patients with detectable VLs show FDG uptake in several lymph node groups, predominantly in cervical and axillary areas, followed by the inguinal area, according to the stage of the disease [18-20]. A correlation between FDG-PET signal and HIV viraemia was found in two studies [17, 19] but another study did not find such a correlation [21]. Metabolism of reactive lymph nodes can thus be variable [18, 21] and, when increased, can decrease the specificity of FDG-PET/CT in viraemic HIV-infected patients [22, 23]. This must be taken into account when interpreting FDG-PET/CT imaging in HIV-infected patients [12, 16].

To assess whether HIV VL could be a confounding factor in interpreting FDG-PET/CT of patients with FUO, we compared HIV-infected patients who had FUO with viraemic HIV-infected patients who did not have FUO.


We performed a prospective study of 20 HIV-positive patients presenting with FUO (group 1) and compared them with a control group of 10 HIV-positive asymptomatic but viraemic patients (group 2). Both groups underwent FDG-PET/CT.

The Ethics Committee of CHU Saint-Pierre approved this study. All patients signed an informed consent form. They were all ≥ 18 years old and were enrolled between July 2005 and October 2008.

Exclusion criteria were pregnancy and predominant neurological symptoms.

Group 1

Twenty HIV-positive patients presenting with FUO were enrolled for FDG-PET/CT. FUO was classically defined as a fever of more than 38.3°C for 3 weeks or more, with an inconclusive first-line work-up including blood analysis, routine microbiological examinations, fundoscopy, chest X-ray and abdominal ultrasound. FDG-PET/CT was performed immediately after the first-line work-up. Final diagnosis was established either by microbiological and/or histopathological examination or by favourable clinical evolution after targeted treatment. Drug-induced fever was defined as occurring after 7 to 10 days of drug administration, persisting as long as the drug was continued, disappearing soon after stopping the drug, and rapidly reappearing if the drug was restarted [29]. HIV-related fever was defined as fever in an HIV-positive patient for whom no etiology of fever was found after complete work-up with fever disappearing when HIV VL became suppressed on antiretroviral therapy. A comprehensive minimal 6 months of clinical follow-up was carried out for all patients.

Group 2

Ten HIV-positive asymptomatic viraemic patients were enrolled for FDG-PET/CT. Patients with a history or evidence of acute or chronic illness were excluded. They all had HIV VL > 10 000 HIV-1 RNA copies/ml. A comprehensive 6 months of clinical follow-up was carried out in order to exclude any occurrence of infectious, inflammatory or malignant disease.

Characteristics of the two groups are presented in Table 1.

Table 1. Characteristics of groups 1 and 2
 Group 1: FUO patients (n = 20)Group 2: viraemic patients (n = 10)
  1. FUO, fever of unknown origin; HAART, highly active antiretroviral therapy; VL, viral load.
Gender15 men, 5 women7 men, 3 women
Age (years) [mean (range)]41.7 (29–73)45 (19–73)
CD4 count (cells/μL) [median (range)]60 (1–566)268 (209–335)
VL (copies/ml) [median (range)]100 000 (0–100 000)90 000 (17 800–> 100 000)
On HAART (n/total)9/200/10

FDG-PET/CT imaging

A PET-CT Discovery LS (GE Medical Systems, Milwaukee, WI) was used for data acquisition. Prior to FDG injection, the patient fasted for at least 6 hours. The blood glucose level was systematically measured before injection of FDG and the patient was not injected if glucose was > 200 mg/dl. The mean blood glucose levels in groups 1 and 2 were 93.7 and 86.4 mg/dl, respectively, with no significant difference between these values.

Sixty minutes after intravenous injection of 220–315 MBq FDG, acquisition was performed with the patient lying in a supine position, from the mid-thigh to the base of the skull. No oral or iodine-based contrast medium was administered. The CT parameters were: 120 kV, 120 mA, pitch 1.5:1, and speed 15 mm/rot. The PET element operated in 2D mode, for 4 minutes per bed position. Attenuation correction was based on the CT data. PET and CT images were analysed by an experienced nuclear medicine physician. This physician knew that the patient was included in the FUO study (blind to the group) but had no access to the medical files of the patients.

FDG accumulation within the regional lymph nodes was studied. For each scan, four peripheral lymph node groups (cervical, axillary, iliac and inguinal) and four central lymph node groups (mediastinal/hilar, internal mammary, para-aortic and mesenteric) were studied. Nodes were designated as abnormal if FDG activity was increased relative to that of adjacent normal soft tissue. The SUVmax corresponds to the maximum standardized uptake value (SUV) in a voxel within a circular region of interest drawn manually for each lymph node group with FDG uptake. In addition to lymph node groups, any focal or diffuse FDG uptake above background whose location was not compatible with normal anatomy/physiology was also noted as pathological.


The significance of differences in SUVmax in lymph nodes between the two groups was estimated using the nonparametric Mann–Whitney test; p < 0.05 was considered to be statistically significant.

We used the highest value of SUVmax in an anatomic region of the patient (either peripheral or central) when calculating mean SUVmax.


Group 1

In the FUO group (group 1), 15 men and five women with a mean age of 41.7 years (range 29–73 years) were enrolled. Their CD4 count ranged from 1 to 566 cells/μL (median 60 cells/μL) and nine of the 20 patients were on HAART. Fifteen patients (of whom four were on HAART) had a HIV VL > 10 000 copies/mL (median > 100 000 copies/mL; range 0 to > 100 000 copies/mL) at the time of FDG-PET/CT. None of the FDG-PET/CT results were normal in the FUO group. Seventeen patients (85%) showed abnormal FDG uptake in peripheral lymph nodes with a mean SUVmax of 8.9 (range 2.3–22), distributed as follows: cervical area (n = 16), axillary (n = 11), iliac (n = 8) and inguinal (n = 6). Sixteen patients (80%) showed abnormal FDG uptake in central lymph nodes with a mean SUVmax of 11.6 (range 3.1–18). Fifteen patients (75%) showed abnormal FDG uptake in 20 extranodal sites, with a mean SUVmax on 12 of 20 available values of 7.6 (range 2.9–19.4).

Final diagnosis was obtained by lymph node biopsy in nine of the 20 patients, performed on lymph nodes with SUVmax > 7.2 in all cases except for one case of a lymph node with plasmablastic lymphoma (SUVmax = 4.9). Central lymph nodes were biopsied in six patients and peripheral lymph nodes in three patients. All lymph node biopsies revealed a diagnosis. In the remaining 11 patients, final diagnosis was obtained by pulmonary microbiology (n = 5), pleural cytology (n = 1), bone marrow examination (n = 1), dental examination (n = 1) or therapeutic challenge (n = 3).

Sixteen patients (80%) were diagnosed with focal lesions: tuberculosis (n = 8), lymphoma (n = 3), nontuberculous mycobacteriosis (n = 3), pneumococcal infection (n = 1) or dental infection (n = 1). The remaining four patients had inflammatory systemic (nonfocal) diseases: one had visceral disseminated leishmaniasis and three had drug-induced or HIV-related fever.

Every anatomical site whose analysis led to diagnosis by microbiology or anatomopathology was abnormal on FDG-PET/CT and showed increased FDG uptake (sensitivity 100%), especially lymph nodes (mean SUVmax = 13; range = 4.9–17.8). Extranodal hypermetabolic sites were consistent with final diagnosis in 13 of 15 patients (86%). Final diagnosis and mode of diagnosis for group 1 are presented in Table 2 and the FDG-PET/CT characteristics of group 1 in Table 3.

Table 2. Group 1 [fever of unknown origin (FUO)]: final diagnosis and mode of diagnosis for FUO
Final diagnosis for FUOnMode of diagnosis
  1. LNB, lymph node biopsy.
Tuberculosis8Pulmonary microbiology (n = 4); LNB (n = 4)
Atypical mycobacteriosis3Abdominal LNB
Drug-induced fever2Clinical diagnosis
HIV-induced fever1Clinical diagnosis
Lymphoma3LNB (n = 2); pleural cytology (n = 1)
Visceral leishmaniosis1Bone marrow puncture
Dental infection1Dental examination
Pneumococcal sepsis1Microbiology
Table 3. Fluorodeoxyglucose-positron emission tomography/computed tomography characteristics of patients of group 1 (n = 20)
Tuberculosis (n = 8) nSUVmax rangeNontuberculous mycobacteriosis (n = 3) nSUVmax range
Peripheral LNC73.2–20.1Peripheral LNC114.7
Central LNM75.1–17.1Central LNM26–14.5
Non-Hodgkin lymphoma (n = 2) nSUVmax rangeDrug- and HIV-related fever (n = 3) nSUVmax range
Peripheral LNC26.2–7.2Peripheral LNC23.5–6
Central LNM23.3–4.9Central LNM13.1
Hodgkin lymphoma (n = 1) nSUVmaxVisceral leishmaniosis (n = 1) nSUVmax
Peripheral LNC17.3Peripheral LNC12.9
 A110.3Dental infection (n = 1) nSUVmax
 I122Peripheral LNC14
IN0 A19.8
Central LNM110.2Pneumococcal infection (n = 1) nSUVmax
  1. VL, viral load; LN, lymph node; SUVmax, standard uptake value maximum; C, cervical; A, axillary; I, iliac; IN, inguinal; M, mediastinal; Mesent, mesenteric; IntMamm, internal mammary; P, para-aortic.
 IntMamm113.5Peripheral LNC16
Mesent0 I13.3
    Central LNM14.3

FDG-PET/CT was helpful in guiding other diagnostic procedures than lymph node biopsy: for instance, pleural puncture was performed because of FDG-PET/CT and revealed a diagnosis of primary effusion lymphoma, bone marrow aspiration was performed because of PET/CT and revealed a diagnosis of leishmaniasis, and dental examination was performed because of FDG-PET/CT as dental abscess was asymptomatic. Moreover, FDG-PET/CT allowed exclusion of focal diseases when it showed no focal uptake. For instance, in a case of HIV-related fever, HAART was started after FDG-PET/CT showed no pathological uptake requiring further investigation.

The most frequent diagnosis for FUO was tuberculosis, which was diagnosed in eight patients, four of whom were diagnosed by lymph node biopsy and four by pulmonary microbiology. Biopsied lymph nodes (thus demonstrated to contain Mycobacterium tuberculosis) had high FDG uptake (range 8.8–17.1) and every patient with tuberculosis had hypermetabolic central lymph nodes on the FDG-PET/CT.

Three cases of nontuberculous mycobacteriosis were diagnosed, all by lymph node biopsy (two abdominal and one cervical) showing high SUVmax (mean = 15.4; range = 13.8–17.8). FDG-PET/CT showed a common pattern in these three patients, with high FDG uptake in central lymph nodes (in the abdominal area in all three patients). In addition, one of these patients showed high FDG uptake in peripheral lymph nodes (cervical): as these lymph nodes were more accessible, one of them was biopsied and revealed a diagnosis.

Two human herpes virus 8 (HHV8)-related lymphoproliferative disorders were diagnosed: one plasmablastic lymphoma by lymph node biopsy and one primary effusion lymphoma by pleural cytology. They showed a similar PET/CT pattern, with peripheral and central lymph nodes having increased FDG uptake with moderate SUVmax (mean 5.6; range 3.3–8.6) and moderately hypermetabolic splenomegaly.

Among patients with drug-induced or HIV-induced fever, FDG-PET/CT showed no focal increased uptake in three of three patients. All had moderate FDG uptake in peripheral lymph nodes (SUVmax range 2.3–6.9). Of note, two patients had high HIV VL (> 100 000 copies/ml). One patient also showed moderately increased FDG uptake in the mediastinal lymph node and one patient showed moderately increased FDG uptake (5.1) in the right lung which corresponded to a known recent bacterial pneumonia. Three patients showed diffuse (oeso-)gastric FDG uptake (two oesophageal candidiasis and one chemical gastritis), with no evidence of a relationship with the final diagnosis for FUO.

Group 2

In the viraemic group without FUO (group 2), 10 patients (seven men and three women; median age 47 years) were enrolled. Their HIV VLs ranged from 17800 to > 100 000 copies/ml (median 90 000 copies/mL) and their CD4 count ranged from 209 to 335 cells/μL (median 268 cells/μL).

Characteristics of groups 1 and 2 are presented in Table 1. The FDG-PET/CT characteristics of group 2 are presented in Table 4. One patient (HIV VL 93 000 copies/mL) showed normal FDG-PET/CT. In nine of the 10 patients (90%), hypermetabolic peripheral lymph nodes were observed (cervical, n = 9; axillary, n = 8; inguinal, n = 7; iliac, n = 5). SUVmax in these peripheral lymph nodes ranged from 2.2 to 11.2 (mean 6.5). One patient presented with moderate FDG uptake in a central (mediastinal) lymph node (SUVmax = 3.8). No extranodal pathological FDG uptake was observed in group 2.

Table 4. Fluorodeoxyglucose-positron emission tomography/computed tomography characteristics of patients of group 2 (n = 10)
  n/10Mean SUVmax (range)
  1. The median HIV viral load of group 2 was 90 000 copies/ml.
  2. Comparison of SUVmax in peripheral lymph nodes of groups 1 and 2 showed higher SUVmax in the fever of unknown origin (FUO) group (group 1), with the difference being statistically significant (P = 0.014).
  3. LN, lymph node; SUVmax, maximum standardized uptake value.
Peripheral LNCervical95.5 (3.6–8.8)
Axillary85.1 (2.3–11.2)
Iliac54.7 (3.3–7.2)
Inguinal73.6 (2.2–5.1)
Central LNMediastinal13.8

Comparison of mean SUVmax in peripheral lymph nodes between groups 1 and 2 showed higher mean SUVmax in the FUO group (group 1), with the difference being statistically significant (P = 0.014). Central lymph nodes in group 1 showed a higher SUVmax than those in group 2, but the difference was not evaluable (only one central lymph node in group 2). There was no correlation between mean SUVmax and HIV VL in these patients.


In this prospective study, we evaluated the efficacy of FDG-PET/CT in investigating FUO in 20 HIV-positive patients. As HIV-positive patients with FUO are often viraemic, which can make interpretation of FDG-PET/CT difficult due to lymph node activation phenomenon [12, 16, 22, 23], we assessed the role of HIV viraemia in FDG-PET/CT interpretation with a control group of 10 HIV-infected viraemic patients without FUO.

Only two small studies have previously been performed to address the usefulness of FDG-PET/CT in the management of FUO in HIV-positive patients. O'Doherty et al. [12] performed a study during the pre-HAART era in 1997 on 80 HIV-positive patients with FUO and/or weight loss and/or confusion. Twenty-three patients had only a brain evaluation carried out. This heterogeneous population of patients who underwent a FDG-PET scan did not undergo CT at that time. The PET scan had a sensitivity of 92% and a specificity of 94% for identification of focal aetiology of the febrile state. The second study, which we performed, examined retrospectively 10 HIV-positive patients with FUO and found that FDG-PET/CT was helpful in diagnosis for nine of the 10 patients [11].

In accordance with the study of O'Doherty et al. [12] (FDG-PET) and our previous work [11], this study shows that FDG-PET/CT has excellent sensitivity for focal lesions in patients with FUO: 100% of our patients with FUO and focal disease had focal increased FDG uptake consistent with the final diagnosis. Increased FDG uptake in lymph nodes had a good positive predictive value for focal disease, especially in the central lymph node, confirming prospectively our previous observations [11]: the presence of increased FDG uptake in central lymph nodes has 100% specificity for focal disease, even in viraemic patients. These central lymph nodes are, by definition, nonpalpable and could thus be identified thanks to FDG-PET/CT. FDG-PET/CT was useful in showing the most hypermetabolic lymph nodes, but also the most easily reachable for biopsy. All biopsies performed in central areas identified the disease underlying the fever. Moreover, as in HIV-negative patients, the absence of hypermetabolic central lymph nodes has an excellent negative predictive value for focal lesions explaining FUO [27, 28].

FDG-PET/CT may thus help to select an appropriate site for biopsy if it shows hypermetabolic central lymph nodes. For instance, in patients with extrapulmonary tuberculosis without pulmonary involvement, FDG-PET/CT could be a useful tool to guide biopsy towards the most appropriate site. Of note, in our FUO group, 50% of patients with tuberculosis (four of eight) needed lymph node biopsy for diagnosis. The median HIV VL in FUO patients was high (> 100 000 copies/ml): both populations of our study were viraemic. We can thus hypothesize that patients with FUO also present reactive lymph nodes, and present a ‘mixed’ population of lymph nodes. In this study, conclusive biopsies in peripheral areas were performed in lymph nodes with SUVmax > 7 (n = 3). To avoid useless invasive procedures in viraemic patients, it seems logical not to perform a biopsy in peripheral lymph nodes with low SUVmax. Many technical factors can influence the measurement of SUVmax, and we observed an important overlap of SUVmax values between our two groups. However, a receiver operating characteristic (ROC) curve was built in an attempt to define a cut-off in peripheral SUVmax values for biopsy. Peripheral lymph node SUVmax values of 0–4 show 0% specificity (useless biopsies only). A peripheral SUVmax value of 6–8 seems to be the best cut-off point to consider a biopsy in this area, with a sensitivity of 62.5% and a specificity of 75%. If the chosen cut-off is 4–6, sensitivity increases to 68.75% but specificity drops to 25% (Fig. 1). The area under the curve (AUC) of this ROC curve is 0.672, while the AUC of the ROC curve of the central lymph node SUVmax values is 0.961.

Figure 1.

Receiver operating characteristic (ROC) curve for peripheral lymph node maximum standardized uptake values (SUVmax) in patients with fever of unknown origin (FUO).

In the FUO group, some specific patterns of FDG-PET/CT were observed according to pathological aetiology.

Tuberculous lymphadenitis showed high FDG avidity (SUVmax range 3.1–20.1) as previously described [11], and all the patients had at least central hypermetabolic lymph nodes. Nontuberculous mycobacteriosis showed very high SUVmax (range 5.1–17.8), mostly in the abdominal area (Fig. 2). HHV8-lymphoproliferative disorders showed central and peripheral lymph nodes and splenomegaly with moderate FDG uptake (range 3.3–8.6), as previously described [24-26]. In these low-grade lymphomas, hypermetabolic splenomegaly and central lymph nodes, usually not observed in reactive lymphadenopathy linked to HIV viraemia, can help with the diagnosis.

Figure 2.

Maximum intensity projection (MIP) images. Group 1 (left): infection with nontuberculous mycobacteria. Fluorodeoxyglucose-positron emission tomography/computed tomography shows increased fluorodeoxyglucose uptake in central (abdominal) and peripheral (cervical) lymph nodes. The fused image shows a hypermetabolic cervical lymph node. Group 2 [viraemic patients without fever of unknown origin (FUO)] (right): moderate fluorodeoxyglucose uptake is observed in cervical [maximum standardized uptake value (SUVmax) 5.1], axillary (SUVmax 2.8) and mediastinal (SUVmax 3.8) lymph nodes.

We compared the FDG-PET/CT patterns of the patients with FUO with those obtained in 10 HIV-positive viraemic and asymptomatic patients, to assess the impact of HIV viraemia on FDG-PET/CT interpretation. Both groups were highly viraemic. Mean SUVmax were higher in the FUO group compared with asymptomatic patients, and the difference was statistically significant in peripheral areas (p = 0.014). The FDG-PET/CT of asymptomatic patients showed a different pattern from that of patients with FUO: 86% of the asymptomatic patients showed hypermetabolic peripheral lymph nodes in the cervical and axillary areas and more than half of the asymptomatic patients showed hypermetabolic peripheral lymph nodes in the iliac and inguinal areas. One central moderately hypermetabolic lymph node was observed in only one patient. These observations are in agreement with previous studies performed in viraemic patients [18-20]. Maybe because of the small number of patients, we could not detect any correlation between HIV viraemia and SUVmax in our viraemic patients, as described in the work of Lucignagni et al. [21]. Nor could we find any correlation between HAART administration and SUVmax. As suspected, it appears that high VL (vascular compartment) is not the only factor responsible for the presence of hypermetabolic lymph nodes (extraplasmatic space) in asymptomatic patients, possibly reflecting compartmentalization of viral replication. Indeed, we observed one HAART-naïve patient with a high VL and normal FDG-PET/CT.

Our study has several limitations: the number of patients was small, subgroups of patients with and without HAART were not compared to assess the impact of antiretroviral treatment on FDG-PET/CT pattern, and the median VLs of groups 1 and 2 were similar.

However, our FUO definition was strict using a set of obligated investigations rather than a time-related criterion. The study was conducted in the post-HAART era, and thus reflects current practice. Moreover, it is the first study to compare patients who had FUO with a control group of viraemic patients to try to assess the potential role of VL in interpreting the results of FDG-PET/CT.

In conclusion, we confirmed prospectively that FDG-PET/CT is a useful tool for the investigation of FUO in HIV-infected patients, as shown retrospectively in our previous work [11] and as in HIV-negative patients [6-10]. FDG-PET/CT contributed to diagnosing or excluding a focal aetiology of the febrile state in 80% of patients. High HIV viraemia does not prevent correct interpretation of FDG-PET/CT. Reactive lymph nodes were observed essentially in peripheral areas with a mean SUVmax of 6.5 in asymptomatic viraemic patients, while in patients with FUO and underlying focal pathologies, we observed central hypermetabolic lymph nodes with a mean SUVmax of 11.6. The absence of hypermetabolic central lymph nodes in patients with FUO has a 100% negative predictive value for focal pathologies. Lymph node biopsy in central hypermetabolic areas allowed identification of underlying disease in patients with FUO in 100% of cases. In the peripheral area, biopsy should be performed in lymph nodes with SUVmax ≥ 6–8.

Some limitations of FDG-PET/CT must be highlighted, such as the low sensitivity in diagnosis of HHV-8 lymphoproliferative disorders (low-grade lymphomas) and the magnitude of overlap in peripheral SUVmax values. Despite the small number of patients, several important findings of this preliminary prospective study may be highlighted.

However, a randomized prospective study in a larger group of HIV-infected patients should be performed. Four groups of patients could be compared (viraemic patients with FUO, nonviraemic patients with FUO, viraemic asymptomatic patients and nonviraemic asymptomatic patients) to better characterize the ‘physiological’ aspects of FDG-PET/CT in viraemic/nonviraemic patients and the role of HAART, and to fully validate the usefulness of FDG-PET/CT in FUO in HIV-infected patients.

The cost-effectiveness of including FDG-PET/CT in FUO work-up in HIV-positive patients has still to be evaluated.