18F-fluorodeoxyglucose–positron emission tomography scanning is more useful in followup than in the initial assessment of patients with Erdheim-Chester disease

Abstract

Objective

Erdheim-Chester disease (ECD) is a rare form of non–Langerhans' cell histiocytosis. The aim of this study was to assess the value of whole-body scanning with ¹⁸F-fluorodeoxyglucose–positron emission tomography (FDG-PET) in a large cohort of ECD patients from a single center.

Methods

We retrospectively reviewed all PET scans performed on 31 patients with ECD who were referred to our department between 2005 and 2008. PET images were reviewed by 2 independent nuclear medicine specialist physicians and were compared with other imaging modalities performed within 15 days of each PET scan.

Results

Thirty-one patients (10 women and 21 men; median age 59.5 years) underwent a total of 65 PET scans. Twenty-three patients (74%) were untreated at the time of the initial PET scan, whereas 30 of the 34 followup PET scans (88%) were performed in patients who were undergoing immunomodulatory therapy. Comparison of the initial and followup PET scans with other imaging modalities revealed that the sensitivity of PET scanning varied greatly among the different organs studied (range 4.3–100%), while the specificity remained high (range 69.2–100%). Followup PET scans were particularly helpful in assessing central nervous system (CNS) involvement, since the PET scan was able to detect an early therapeutic response of CNS lesions, even before magnetic resonance imaging showed a decrease in their size. PET scanning was also very helpful in evaluating the cardiovascular system, which is a major prognostic factor in ECD, by assessing the heart and the entire vascular tree during a single session.

Conclusion

The results of our large, single-center, retrospective study suggest that the findings of a FDG-PET scan may be interesting in the initial assessment of patients with ECD, but its greater contribution is in followup of these patients.

Erdheim-Chester disease (ECD) is a rare form of non–Langerhans' cell histiocytosis first described by Jakob Erdheim and William Chester in 1930 (1). By November 2008, a total of 319 distinct cases of ECD have been described in the medical literature. While bone pain is the most common presenting symptom of the disease, half of all patients may experience extraskeletal manifestations, including central nervous system (CNS) and cardiovascular involvement, retroperitoneal and perirenal infiltration, interstitial lung disease, osteosclerosis of the paranasal sinuses, and exophthalmos (2).

The clinical course of ECD is largely dependent on the extent and distribution of the disease. Some patients may be asymptomatic or have only mild and limited symptoms, while other patients may have a more aggressive form of the disease, with a poor prognosis, particularly when CNS and cardiovascular systems are involved (3, 4). A reliable diagnosis of ECD can only be made if clinical, radiologic, and pathologic findings showing infiltration of involved tissues by CD68+ and CD1a– foamy histiocytes are considered together. While common radiologic investigations in ECD include radiography, ^99mTc bone scintigraphy, computed tomography (CT), and magnetic resonance imaging (MRI), none of these imaging modalities is able to provide a global assessment of the lesions during a single session.

Positron emission tomography (PET) using ¹⁸F-labeled fluorodeoxyglucose (FDG) is a noninvasive metabolic imaging modality based on the regional distribution of radiolabeled FDG, which accumulates in hypermetabolic tissues. Since its introduction to the field of clinical imaging, FDG-PET scanning has rapidly become a major tool for the management of oncologic diseases (5). Previous studies have suggested that PET scanning may play a role in the management of inflammatory diseases by detecting areas of increased glucose metabolism that are associated with inflammatory infiltrates (6). A few case reports have shown that FDG-PET scanning could be useful in assessing the extension of ECD lesions (7–12).

The aim of the present study was to determine the value of whole-body scanning with FDG-PET in the initial and followup assessments of the extension and activity of ECD lesions in a large cohort of patients at a single center.

PATIENTS AND METHODS

Patients.

We retrospectively reviewed 65 FDG-PET scans performed in 31 consecutive patients with a diagnosis of ECD. These patients were referred to the Internal Medicine Department of Hôpital Pitié-Salpêtrière between January 2005 and August 2008. PET scans were performed routinely, as part of the imaging strategy for ECD we use in our department. ECD was diagnosed based on the following criteria: 1) typical histologic findings of infiltration with foamy histiocytes nested among polymorphic granuloma, and fibrosis or xanthogranulomatosis with CD68+ and CD1a− immunohistochemical staining, and 2) typical skeletal findings, with radiographs showing bilateral and symmetric cortical osteosclerosis of the diaphyseal and metaphyseal regions of the long bones and/or symmetric and abnormally intense uptake in the distal ends of the long bones of the legs and, in some cases, the arms on ^99mTc bone scintigraphy. These criteria for ECD were used in our previous studies (3, 13, 14). Informed consent was obtained from all participants before the study began.

FDG-PET scanning and initial data processing.

Data from the FDG-PET scan were acquired using a Gemini Dual PET-CT scanner (Philips GXL Medical Systems, Cleveland, OH), which combines a helical 16-slice CT and a 3- dimensional PET scanner. Patients fasted for >6 hours before the study. Patients were scanned for 60 minutes after injection with ¹⁸F-labeled FDG (5 MBq/kg). A helical CT scan without injection of contrast material was performed first (scan field of 600 mm, increment of 5 mm, slice thickness of 5 mm, pitch of 1, 0.75 seconds per rotation, matrix of 512 × 512 at 120 kV, 150 mA). The whole-body PET scan was started at the end of the CT acquisition and consisted of 10–11 bed positions of 3 minutes each. Total acquisition time per patient varied from 35 to 40 minutes.

Qualitative and quantitative analyses of the PET images.

Images from the whole-body PET scans were displayed in coronal, transaxial, and sagittal slices, as well as in rotating 3-dimensional images. The images were reviewed by 2 independent nuclear medicine specialist physicians (ZM and FA), who were informed of the diagnosis of ECD but were blinded with regard to the other clinical, biologic, and radiologic data. In cases of disagreement between the 2 nuclear medicine specialist physicians, the final interpretation was determined by majority opinion after additional review of the images with a third nuclear medicine specialist physician (AK). A region of interest was placed over the most intense areas of FDG accumulation to minimize any partial volume effect.

A quantitative analysis of uptake was performed using the standardized uptake value (SUV), which is defined as the ratio of the measured FDG activity to the injected activity, which was then normalized to the body mass. FDG uptake was considered significant in large vessels only when the maximum SUV (SUV_max) was equal to or exceeded the mean SUV in the liver. The intensity of vascular uptake was expressed quantitatively using the vascular uptake index, which is defined as the ratio of the vascular SUV_max divided by the mean SUV in the liver, as described previously (15).

Analysis of other imaging modalities.

Images from MRI and CT scans performed within 15 days of each PET scan were reviewed for the presence or absence of involvement of each organ system that is normally involved in ECD. These findings were compared with those from the PET scan images.

Images obtained upon MRI of the CNS with dedicated sequences of the orbits as well as upon CT of the paranasal sinuses were reviewed by a trained neuroradiologist (AD). MRIs of the long bones were reviewed by a radiologist specialized in bone radiology (DZ). Bone scans were only performed as part of the initial assessment, and bone MRIs were performed as part of the followup. Thus, images from the initial PET scan were compared with those from the bone scan, and images from the followup PET scan were compared with those from the bone MRI for assessment of osseous involvement. Images obtained upon MRI of the heart, as well as images from the chest and abdominal CT scans were reviewed by a radiologist specialized in thoracic and abdominal radiology (DT). These 3 radiologists were blinded to the results of the PET scan analysis.

Statistical analysis.

Qualitative variables were expressed as numbers and percentages and are reported as the median and range. The median SUV_max was calculated considering only the foci with increased FDG uptake. The degree of agreement between the nuclear medicine specialist physicians' interpretations was assessed using the percentage of agreement (for the presence or absence of foci of increased FDG uptake) and using Spearman's correlation test (for interobserver SUV_max agreement).

Comparisons between PET scans and other imaging modalities (bone scans, CT scans, and MRIs) were assessed using sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), considering that these latter imaging modalities were the gold standards. Thus, we studied the diagnostic value of PET scanning as compared with that of bone scanning, CT scanning, and MRI, but not the actual diagnostic value of these latter modalities, which we assumed were 100% accurate. Because our study is retrospective and some data were not available, we calculated sensitivities, specificities, PPVs, and NPVs considering only cases in which both the PET scan data and the other imaging data were available. Such data were available in at least 61% of the patients, depending on the imaging modality being considered. Wilcoxon's signed rank test was used to assess differences between the initial and followup SUV_max.

Extension of vascular uptake was expressed as a score ranging from 1 to 9. This score was calculated considering only patients who had increased vascular uptake, by scoring the presence or absence of any significant uptake in the following vascular territories: carotid arteries, brachiocephalic trunk, subclavian arteries, ascending aorta, aortic arch, descending aorta, abdominal aorta, iliac arteries, and femoral arteries (1 point each if present). The number of visceral systems involved was expressed as a score ranging from 0 to 8. This score was calculated considering the presence or absence of any significant uptake in the osseous, CNS, vascular, cardiac, pulmonary, retroperitoneal, paranasal sinus, and orbital organ systems (1 point each if present). Statistical analyses were performed using the GraphPad Prism software package version 5.0 (GraphPad Software, San Diego, CA).

RESULTS

Characteristics of the study patients.

Thirty-one patients with ECD (10 women and 21 men; median age at the time of the PET scan 59.5 years [range 19–79 years]) underwent a total of 65 FDG-PET scans. Twenty-three of these 31 patients (74%) were untreated at the time of the initial PET scan. The remaining patients were being treated with corticosteroids, interferon-α, or imatinib mesylate (13, 16, 17). The characteristics of the ECD patients and their treatments are presented in Table 1.

Table 1. Characteristics of the patients with Erdheim-Chester disease and the treatments given*

Patient/sex	Age, years		PET scan number	Age at PET scan, years	Treatment
	At first symptom	At diagnosis
* PET = positron emission tomography; IFNα = interferon-α; MIU = million IU; PEG-IFN = PEGylated IFNα; IV = intravenous.
1/M	43	46	1A	57.7	None
2/M	53	54	2A	55.0	None
			2B	55.5	IFNα 9 MIU × 3/week
3/M	41	66	3A	67.5	None
4/M	52	53	4A	58.7	IFNα 3 MIU × 3/week
			4B	59.7	PEG-IFN 135 μg/week
5/M	63	65	5A	67.3	IFNα 3 MIU × 3/week
			5B	67.5	PEG-IFN 180 μg/week
			5C	68.4	None
6/F	54	59	6A	60.1	IFNα 9 MIU × 3/week
7/M	59	59	7A	62.5	None
8/M	57	57	8A	57.9	None
			8B	58.4	Prednisone 10 mg/day
			8C	58.8	Prednisone 10 mg/day
9/M	76	77	9A	78.7	IFNα 3 MIU × 3/week
10/M	50	51	10A	50.5	None
11/M	13	26	11A	30.3	None
			11B	30.8	IFNα 9 MIU × 3/week
			11C	31.3	IFNα 9 MIU × 3/week
			11D	31.8	IFNα 9 MIU × 3/week
12/M	52	52	12A	58.3	None
			12B	58.8	IFNα 3 MIU × 3/week
			12C	59.4	IFNα 3 MIU × 3/week
			12D	60.1	IFNα 3 MIU × 3/week
			12E	60.9	None
13/M	26	43	13A	42.9	None
			13B	44.2	IFNα 9 MIU × 3/week plus prednisone 15 mg/day
			13C	44.5	IFNα 9 MIU × 3/week plus prednisone 10 mg/day
			13D	45.0	IFNα 9 MIU × 3/week plus prednisone 10 mg/day
14/M	60	60	14A	63.5	Imatinib mesylate 200 mg/day
			14B	65.1	None
			14C	66.1	IFNα 3 MIU × 3/week plus IV pamidronate × 1/month
15/M	71	72	15A	72.1	None
			15B	72.4	IFNα 3 MIU × 3/week
16/F	54	61	16A	62.6	None
			16B	63.0	Imatinib mesylate 700 mg/day
			16C	63.3	Imatinib mesylate 800 mg/day
			16D	63.9	IFNα 9 MIU × 3/week
			16E	64.2	IFNα 9 MIU × 3/week
			16F	64.7	IFNα 9 MIU × 3/week
			16G	65.3	PEG-IFN 250 μg/week
17/F	41	42	17A	42.4	None
			17B	43.0	IFNα 9 MIU × 3/week
18/F	71	72	18A	77.4	None
19/M	52	59	19A	66.4	IFNα 2 MIU × 3/week
20/F	66	69	20A	69.5	None
21/M	64	65	21A	65.5	None
22/F	51	52	22A	52.5	None
			22B	54.9	PEG-IFN 180 μg/week
23/M	38	41	23A	47.8	None
24/M	43	43	24A	43.9	IFNα 3 MIU × 3/week plus prednisone 10 mg/day
25/M	51	57	25A	61.7	IFNα 3 MIU × 3/week
			25B	62.1	PEG-IFN 180 μg/week
26/F	41	41	26A	47.6	None
27/F	6	16	27A	18.6	None
			27B	19.2	IFNα 3 MIU × 3/week
			27C	20.0	None
28/F	17	21	28A	33.7	None
29/M	29	29	29A	29.9	None
			29B	30.5	IFNα 9 MIU × 3/week
30/F	63	63	30A	63.4	None
			30B	64.1	IFNα 3 MIU × 3/week
			30C	64.6	PEG-IFN 180 μg/week
31/F	70	73	31A	73.0	None
			31B	75.2	PEG-IFN 180 μg/week

Interobserver agreement.

The 2 independent nuclear medicine specialist physicians agreed on the presence or absence of foci of increased FDG uptake in 98.5% of cases. The interobserver agreement for the SUV_max was excellent (r = 0.90 by Spearman's correlation; P < 0.0001).

Analysis of initial PET scan images.

Detailed localization of uptake in the 31 initial PET scans and comparison with other imaging modalities are presented in Table 2. Typical examples of PET scans performed in ECD patients are presented in Figures 1 and 2.

Table 2. Findings of initial PET scans and comparison with other imaging modalities*

PET scan number	Long bones		Cardiac			Large vessels		CNS		Pleuro-pulmonary		Perirenal		Sinuses		Orbits
	Bone scan	PET	MRI	CT	PET	CT	PET	MRI	PET	CT	PET	CT	PET	CT	PET	MRI	PET
* The presence or absence of involvement identified with each imaging method is indicated. PET = positron emission tomography; CNS = central nervous system; MRI = magnetic resonance imaging; CT = computed tomography; NA = not available; PPV = positive predictive value; NPV = negative predictive value. † The cardiac MRI results were compared with the cardiac PET results.
1A	+	+	NA	+	+	+	+	NA	–	+	–	+	–	NA	–	NA	–
2A	+	+	+	NA	–	+	–	+	+	+	–	+	–	+	+	–	–
3A	+	–	NA	–	–	+	–	–	–	–	+	–	–	+	–	–	–
4A	+	–	NA	–	–	–	–	NA	–	–	–	+	–	NA	–	NA	–
5A	+	+	NA	NA	–	NA	–	NA	–	NA	–	NA	–	NA	+	NA	–
6A	+	–	+	NA	–	–	–	–	–	–	–	+	–	+	–	–	–
7A	+	+	+	NA	+	+	–	+	+	+	–	+	–	+	–	+	+
8A	+	+	+	NA	+	+	–	+	+	+	+	+	–	–	–	–	–
9A	+	+	+	+	+	NA	–	NA	–	NA	–	NA	–	NA	–	NA	–
10A	+	–	NA	–	–	+	+	–	–	–	–	NA	–	–	–	–	–
11A	+	–	–	–	–	–	–	NA	+	–	–	–	–	NA	+	NA	–
12A	+	–	NA	–	–	+	+	–	–	–	–	+	–	–	+	–	–
13A	+	–	NA	NA	–	+	+	+	–	–	–	+	–	+	–	+	–
14A	+	+	NA	NA	–	NA	–	–	–	NA	–	NA	–	+	–	–	–
15A	+	+	+	+	–	+	+	–	–	+	+	+	–	–	–	–	–
16A	+	–	+	NA	–	+	–	+	+	+	+	–	–	+	–	+	+
17A	+	+	–	NA	–	–	–	+	+	NA	–	NA	–	+	–	–	–
18A	+	+	–	NA	–	+	–	NA	–	NA	–	NA	–	NA	–	NA	–
19A	+	+	+	NA	–	+	+	–	–	+	–	+	–	+	–	+	–
20A	+	+	+	+	+	+	+	–	+	NA	–	+	+	+	–	+	+
21A	+	+	–	NA	–	–	+	–	–	–	–	–	–	–	–	–	–
22A	+	+	NA	NA	–	–	–	–	–	–	–	–	–	–	–	–	–
23A	+	–	–	NA	–	+	–	+	–	+	–	+	–	+	–	–	–
24A	+	+	+	+	–	+	–	NA	–	–	–	+	–	+	–	–	–
25A	+	+	NA	NA	+	+	+	+	–	NA	–	+	–	–	–	–	–
26A	–	–	NA	NA	–	NA	+	NA	–	NA	–	NA	–	–	–	–	–
27A	–	–	NA	NA	–	NA	–	NA	–	NA	–	NA	–	+	–	–	–
28A	+	–	NA	NA	–	NA	–	–	–	NA	–	NA	–	+	+	–	+
29A	+	–	NA	NA	–	NA	–	+	+	NA	–	NA	–	–	–	–	–
30A	+	+	–	–	–	+	+	–	–	–	–	–	–	–	–	–	–
31A	+	–	NA	+	–	+	+	–	–	NA	–	+	–	+	–	–	–
PET vs. other modality†
Sensitivity, %	58.6		9.1	33.3		55.6		66.7		37.5		6.7		13.3		60.0
Specificity, %	100		100	100		83.3		92.3		90.9		100		90.0		95.0
PPV, %	100		100	100		90.9		85.7		75.0		100		66.7		75.0
NPV, %	14.3		28.6	78.9		38.5		80.0		66.7		30.0		40.9		90.5

Figure 1. ¹⁸F-fluorodeoxyglucose–positron emission tomography (FDG-PET) scans showing involvement of various organ systems in patients with Erdheim-Chester disease. A, Typical bilateral and symmetric uptake in the diaphyseal and metaphyseal regions of the long bones is seen (arrows). B, Cardiovascular involvement, with uptake in the carotid arteries, the aortic arch, and the ascending aorta (arrows), is shown. C, Perirenal infiltration (encircled areas) on an image from a computed tomography (CT) scan (right) and absence of corresponding uptake on an image from a PET scan (left) are demonstrated. D, Neurologic involvement, with multiple cerebral uptake foci (arrows), is demonstrated. E, Uptake in the right maxillary sinus (arrows) is shown. F, Images from a CT scan, a PET scan, and fusion of the PET and CT images, demonstrating bilateral pulmonary uptake (arrows), are shown.

Figure 2. Typical distribution of lesions seen on whole-body ¹⁸F-fluorodeoxyglucose–positron emission tomography scans of 4 different patients with Erdheim-Chester disease.

Osseous involvement.

FDG uptake was present in the long bones in 17 of the 31 initial PET scans (55%) and always in a bilateral and symmetric manner. The median SUV_max of osseous uptake was 2.8 (range 1.5–20.0). Globally, osseous uptake was localized in the lower limbs in 16 of these 17 PET scans (94%) and in the upper limbs in 7 (41%). Other uptake sites included the calcaneum (2 patients), the dorsal vertebrae (2 patients), the tarsal bones (2 patients), and the ribs, sternum, and acetabular roof (1 patient each). Compared with the bone scans, the PET scans yielded a sensitivity of 58.6%, a specificity of 100%, a PPV of 100%, and an NPV of 14.3%.

Neurologic involvement.

The presence of foci of abnormal cerebral uptake was observed in 8 of the 31 initial PET scans (26%), with a median SUV_max of 12 (range 5.1–16). Five patients had a single CNS lesion (located in the cerebral trunk in 2 patients and in the sellar, left occipital, or right posterior cingulate regions in 1 patient each). Two patients had 2 distinct CNS lesions (left intraventricular and left occipital lesions in one patient and right frontal and left frontal lesions in the other patient). One patient (patient 16) had 6 distinct lesions (right parietal, left temporal, right occipital, left occipital, right temporo-occipital, and left temporo-occipital lesions). Compared with the MRIs of the CNS, the PET scans yielded a sensitivity of 66.7%, a specificity of 92.3%, a PPV of 85.7%, and an NPV of 80%.

Cardiovascular involvement.

The presence of uptake in the large vessels was observed in 12 of the 31 initial PET scans (38%), with a median vascular uptake index of 1.51 (range 1.1–6.37). Vascular uptake was localized in the ascending aorta (8 patients), aortic arch (8 patients), abdominal aorta (5 patients), descending aorta (3 patients), subclavian arteries (3 patients), carotid arteries (2 patients), brachiocephalic trunk (2 patients), iliac arteries (2 patients), and superior mesenteric artery (1 patient). One patient had increased uptake in the superior vena cava. Considering only the patients who had increased vascular uptake, the median number of vascular territories involved on the initial PET scans was 2 (range 1–7). Compared with the thoracic CT scans for the assessment of large vessel involvement, the PET scans yielded a sensitivity of 55.6%, a specificity of 83.3%, a PPV of 90.9%, and an NPV of 38.5%.

Abnormal FDG uptake was observed in the heart in 6 of the 31 initial PET scans (19.3%), with a median SUV_max of 4.0 (range 2.8–8.2). Compared with the MRIs, the PET scans yielded a sensitivity of 9.1%, a specificity of 100%, a PPV of 100%, and an NPV of 28.6% for the assessment of heart involvement in ECD. Compared with the CT scans, the PET scans yielded a sensitivity of 33.3%, a specificity of 100%, a PPV of 100%, and an NPV of 78.9%.

Pulmonary involvement.

Evidence of pleuropulmonary uptake was found in 4 of the 31 initial PET scans (12.9%), with a median SUV_max of 2.4 (range 1.8–3.5). FDG uptake was localized in both the lung parenchyma (2 patients) and the pleural region (3 patients). Compared with the high-resolution CT scans of the chest, the PET scans yielded a sensitivity of 37.5%, a specificity of 90.9%, a PPV of 75.0%, and an NPV of 66.7% for the assessment of respiratory involvement.

Retroperitoneal and perirenal involvement.

Only 1 of the 31 PET scans (3.2%) showed FDG uptake in the retroperitoneal and perirenal region (SUV_max 5.9). Compared with the CT scans of the abdomen, the PET scans yielded a sensitivity of 6.7%, a specificity of 100%, a PPV of 100%, and an NPV of 30.0% for the assessment of retroperitoneal and perirenal involvement.

Paranasal sinus involvement.

Increased FDG uptake was present in the paranasal sinuses in 5 of the 31 initial PET scans (16%). The median SUV_max was 5 (range 3.0–11.4). FDG uptake was localized in the right maxillary sinus in 4 patients, the left maxillary sinus in 2 patients, and the frontal and ethmoid sinuses in 1 patient each. Compared with the CT scans of the paranasal sinuses, the PET scans yielded a sensitivity of 13.3%, a specificity of 90.0%, a PPV of 66.7%, and an NPV of 40.9%.

Orbital involvement.

Orbital uptake was present in 4 of the 31 initial PET scans (12.9%), with a median SUV_max of 6.3 (range 3.8–8.4). Compared with the MRIs of the orbits, the PET scans yielded a sensitivity of 60.0%, a specificity of 95.0%, a PPV of 75.0%, and an NPV of 90.5%.

Analysis of followup PET scans.

Followup PET scans were performed in 17 of the 31 patients (55%). Eight patients underwent a total of 2 PET scans, 5 patients underwent 3, 2 patients underwent 4, 1 patient underwent 5, and 1 patient underwent 7 PET scans. The median duration between followup PET scans was 6.7 months (range 2.2–28.7 months). Thirty of the 34 followup PET scans (88%) were performed in patients undergoing immunomodulatory treatment, whereas 13 of the corresponding 17 initial PET scans (76%) were performed in untreated patients (Table 1).

A detailed description of the uptake findings in the followup PET scans is presented in Table 3. Briefly, osseous uptake was observed in 18 of these 34 followup PET scans (24%), abnormal CNS uptake in 13 (38%), large vessel uptake in 9 (26%), abnormal heart uptake in 2 (6%), pulmonary uptake in 5 (15%), retroperitoneal and perirenal uptake in 1 (3%), paranasal sinus uptake in 14 (41%), and orbital uptake in 1 (3%).

Table 3. Findings of followup PET scans and comparison with other imaging modalities*

PET scan number	Long bones		Cardiac			Large vessels		CNS		Pleuro-pulmonary		Perirenal		Sinuses		Orbits
	MRI	PET	MRI	CT	PET	CT	PET	MRI	PET	CT	PET	CT	PET	CT	PET	MRI	PET
* The presence or absence of involvement identified with each imaging method is indicated. PET = positron emission tomography; CNS = central nervous system; MRI = magnetic resonance imaging; CT = computed tomography; NA = not available; PPV = positive predictive value; NPV = negative predictive value. † The cardiac MRI results were compared with the cardiac PET results.
2B	NA	+	NA	NA	–	+	–	+	+	+	–	+	+	+	+	–	–
4B	NA	–	NA	NA	–	–	–	NA	–	NA	–	+	–	NA	–	NA	+
5B	NA	+	+	NA	–	+	–	NA	–	NA	–	+	–	NA	+	NA	–
5C	NA	+	+	NA	–	+	–	NA	–	NA	–	+	–	+	+	–	–
8B	NA	+	NA	+	+	–	–	+	+	+	+	+	–	–	–	–	–
8C	NA	+	NA	+	+	–	–	+	+	+	+	+	–	–	–	–	–
11B	NA	–	NA	NA	–	NA	–	+	+	NA	–	NA	–	+	+	–	–
11C	NA	–	NA	–	–	–	–	+	+	–	–	–	–	+	+	–	–
11D	NA	–	NA	NA	–	NA	–	+	+	NA	–	NA	–	+	+	–	–
12B	NA	–	NA	–	–	+	+	–	–	–	–	+	–	–	+	–	–
12C	NA	–	NA	–	–	+	+	–	–	–	–	+	–	–	+	–	–
12D	NA	–	NA	–	–	+	+	NA	–	–	–	+	–	NA	+	–	–
12E	NA	–	NA	–	–	+	+	–	–	–	–	+	–	–	+	–	–
13B	NA	–	NA	–	–	+	–	+	–	–	–	+	–	+	–	+	–
13C	NA	–	NA	NA	–	NA	–	+	–	–	–	NA	–	+	–	+	–
13D	NA	–	NA	NA	–	NA	–	NA	–	NA	–	NA	–	NA	–	NA	–
14B	+	+	NA	–	–	+	–	–	–	–	–	+	–	+	–	–	–
14C	+	+	NA	–	–	+	–	NA	–	–	–	+	–	NA	–	NA	–
15B	NA	+	NA	+	–	+	+	NA	–	+	–	+	–	NA	–	NA	–
16B	NA	–	+	NA	–	+	–	+	+	+	+	NA	–	+	+	+	–
16C	NA	+	NA	NA	–	+	–	+	+	+	+	NA	–	+	+	+	–
16D	NA	+	+	NA	–	+	–	+	+	+	+	NA	–	+	+	+	–
16E	+	+	NA	NA	–	+	–	+	+	+	–	NA	–	+	–	+	–
16F	NA	+	+	NA	–	+	–	+	+	+	–	–	–	+	–	+	–
16E	NA	+	+	NA	–	+	–	+	+	+	–	–	–	+	+	+	–
17B	NA	+	NA	NA	–	NA	–	+	+	NA	–	NA	–	+	–	–	–
22B	+	+	NA	NA	–	–	–	NA	–	–	–	NA	–	NA	–	NA	–
25B	NA	+	+	NA	–	NA	+	NA	–	NA	–	NA	–	NA	–	NA	–
27B	NA	–	–	–	–	–	–	–	–	–	–	–	–	+	–	–	–
27C	NA	–	NA	NA	–	NA	–	NA	–	NA	–	NA	–	+	–	–	–
29B	NA	–	NA	NA	–	NA	–	+	–	NA	–	NA	–	–	–	–	–
30B	+	+	NA	–	–	+	+	NA	–	–	–	–	–	NA	–	NA	–
30C	+	+	NA	–	–	+	+	NA	–	–	–	–	–	NA	–	NA	–
31B	NA	–	NA	+	–	+	+	NA	–	–	–	+	–	NA	–	NA	–
PET vs. other modality†
Sensitivity, %	100		9.1	33.3		48.3		78.3		57.1		4.3		41.7		10.0
Specificity, %	100		100	100		100		100		100		100		69.2		100
PPV, %	100		100	100		100		100		100		100		71.4		100
NPV, %	100		28.6	78.9		40.0		68.8		76.9		33.3		39.1		75.7

Changes in the SUV_max between the initial PET scan and the last followup PET scan performed for each patient are presented in Table 4. Considering the patients undergoing immunomodulatory treatment, the SUV_max decreased in most osseous, CNS, and pulmonary foci, whereas the SUV_max increased in most large vessel and paranasal sinus foci. The SUV_max increased in 1 of the 2 cardiac foci and decreased in the other. The median number of organ systems involved (overall range 0–8) was 1.5 (range 0–3) in the initial PET scans and 2 (range 0–4) in the last followup PET scans. This difference was not significant (P = 0.81 by Wilcoxon's test).

Table 4. Evolution of SUV_max on followup PET scans*

Organ system	Total no. of uptake foci on initial PET scan	SUVmax status on followup PET scan, no. (%) of foci†
		Increased SUVmax	Stable SUVmax	Decreased SUVmax
* Only patients with available followup positron emission tomography (PET) scans were analyzed. The intensity of uptake in large vessels was assessed according to the vascular uptake index, rather than the maximum standardized uptake value (SUVmax). CNS = central nervous system; NA = not available. † Comparing the initial PET scan with the last PET scan performed.
Osseous	37	8 (22)	5 (14)	24 (65)
CNS	15	3 (20)	0 (0)	12 (80)
Large vessels	15	8 (53)	2 (13)	5 (33)
Cardiac	2	1 (50)	0 (0)	1 (50)
Pulmonary	2	0 (0)	0 (0)	2 (100)
Retroperitoneal	NA	NA	NA	NA
Sinus	10	6 (60)	2 (20)	2 (20)
Orbital	NA	NA	NA	NA

The sensitivity, specificity, PPV, and NPV achieved in comparisons of the followup PET scans with other imaging modalities are presented in Table 3. Briefly, PET scans had excellent specificity and PPV as compared with most other imaging modalities. In contrast, the sensitivity and NPV varied greatly among the different sites of involvement we studied.

DISCUSSION

Only a few case reports have highlighted the potential utility of FDG-PET scanning in the evaluation of osseous (7, 9, 18), neurologic (7, 11), vascular (8), pericardial (8, 11), or orbital (8, 18) involvement in patients with ECD. In this retrospective study of 65 PET scans performed in 31 consecutive patients, we found that whole-body FDG-PET scanning was able to depict globally both the extent and the activity of the ECD lesions. Because most initial PET scans were performed in untreated patients and most followup PET scans were performed in patients undergoing immunomodulatory treatment, we were able to evaluate the usefulness of PET scanning in the assessment of the therapeutic response in ECD. Our study also provides a direct comparison between PET scanning and the more classic imaging modalities. To the best of our knowledge, this is the first large-scale study of the usefulness of whole-body PET scanning in the initial and followup assessment of patients with ECD.

Up to now, radiography, ^99mTc bone scintigraphy, MRI, and CT have been the main imaging methods used in the diagnosis and followup of patients with ECD. However, none of these imaging modalities is able to provide a global assessment of the extent and activity of the disease. While conventional imaging modalities have mostly been confined to the study of a single region, FDG-PET scanning is usually applied as a whole-body examination, and thus, it appears to be very advantageous in this regard.

All ECD patients referred to our department were evaluated on a regular basis according to a standardized imaging protocol that included ^99mTc bone scintigraphy, MRI of the CNS with dedicated sequences of the orbits, CT of the paranasal sinuses, and contrast-enhanced thoracoabdominal CT with high-resolution chest CT. Nineteen patients have also undergone dedicated cine-MRI of the heart. We decided to analyze only radiologic investigations performed within 15 days of each PET scan, thus allowing reliable comparisons between the PET scans and these radiologic examinations.

Comparison of both the initial and the followup PET scans with other imaging modalities revealed that the sensitivity of PET scanning varied greatly among the different organ systems studied (ranging from 4.3% to 78.3%), while the specificity was high (ranging from 69.2% to 100%).

Both radiography and ^99mTc bone scintigraphy may reveal osteosclerosis of the long bones, which is a typical finding in ECD (19). Assessment of osseous involvement of ECD is mostly performed as part of the diagnosis process. It is of minor importance in the followup of ECD patients, since pathologic fractures are very uncommon (2). Positivity of either radiography or ^99mTc bone scintigraphy are part of the diagnostic criteria for ECD that we used in the present study, along with the infiltration of involved tissues by CD68+ and CD1a– foamy histiocytes. PET scan assessments of osseous involvement revealed typical bilateral and symmetric uptake of FDG in the long bones (Figure 1). This uptake pattern is similar to that observed with ^99mTc bone scintigraphy. This is an important finding, since this typical pattern is very specific for ECD, and thus, it is highly evocative of this diagnosis (19). When compared with ^99mTc bone scintigraphy, the initial PET scan had moderate sensitivity (58.6%) but excellent specificity (100%). Because the interest of ^99mTc bone scintigraphy is limited to the assessment of osseous involvement in ECD, we believe that ^99mTc bone scintigraphy may reasonably be performed only in ECD patients in whom initial PET scans do not show typical osseous uptake. About half of all ECD patients may experience extraskeletal manifestations, including CNS, cardiovascular, retroperitoneal, respiratory, paranasal sinus, and orbital involvement of the disease (2).

We believe that PET scanning was very helpful in assessing CNS involvement in this disease, which is one of the major prognostic factors in ECD (4). Both the initial and the followup PET scans had good sensitivity (66.7% and 78.3%, respectively) as compared with MRI of the CNS. Specificity was high on both the initial and followup PET scans (92.3% and 100%, respectively). We believe that PET scanning was particularly useful in the followup of CNS lesions, since we were able to monitor not only the number of lesions, but also the SUV_max of the lesions, which reflects the underlying metabolic activity. While the proportion of patients undergoing immunomodulatory treatment increased between the initial and the last followup PET scans, the SUV_max decreased in 80% of CNS foci and returned to normal in most patients (Table 4). The corresponding lesions could still be observed on followup MRIs (Table 3), suggesting that PET scanning is able to detect an early therapeutic response in CNS lesions, even before the MRI shows a decrease in the size of these lesions. Interestingly, MRI of the CNS revealed a circumscribed T2 hyperintensity of the dentate nuclei in patients 23 and 25, while the intensity of FDG uptake was normal in the corresponding regions. Ribeiro et al (20) have previously reported such MRI findings in the neurodegenerative form of Langerhans' cell histiocytosis, along with a decrease in local FDG uptake. However, all the patients included in our study had biopsy-proven ECD, and none of them had clinically progressive cerebellar ataxia.

We have previously reported that cardiovascular involvement is a major prognostic factor in ECD (3). Thus, clear identification of ECD patients with cardiovascular involvement is a major concern for physicians. Both the initial and followup PET scans yielded a moderate sensitivity (55.6% and 48.3%, respectively) but a high specificity for the assessment of large vessel involvement, as compared with CT scans. One of the major interests of PET scanning was that we were able to assess the whole vascular tree during a single session. Due to improvements in MRI techniques, it is now possible to also acquire the entire vascular tree during a single MRI session, but this new imaging modality is not widely available because it requires very high magnetic field strengths (21).

The sensitivity of both the initial and followup PET scans for detecting cardiovascular involvement was low as compared with that of MRI or CT scans of the heart, probably because typical ECD lesions of the heart (i.e., pseudotumors of the right atrium) are usually of moderate size, and their FDG uptake may be masked by the physiologic uptake present in the myocardium.

Considering the pulmonary involvement of ECD, comparison of initial and followup PET scans with high-resolution chest CT scans yielded fair sensitivity (37.5%) and moderate sensitivity (57.1%), respectively. Specificity was excellent for both the initial and the followup PET scans (90.9 and 100%, respectively). All patients in whom significant pulmonary uptake was observed had extensive pleural and/or pulmonary infiltration of the disease. Conversely, all patients with false-negative findings on PET scans had limited, but typical, interlobular septal thickening on high-resolution chest CT scans (22), suggesting that PET scanning may best serve in the followup of respiratory involvement of the disease, rather than in its initial assessment.

The sensitivity of both the initial and followup PET scans for detecting retroperitoneal and perirenal involvement was poor (6.7% and 4.3%, respectively). This is quite surprising, since the nature of the retroperitoneal infiltrate is no different from what is encountered in the other sites involved, and most patients included in our study had the typical “hairy” perirenal infiltration of the disease (22).

There is preliminary evidence that PET scanning could be useful in evaluating patients with Langerhans' cell histiocytosis (23) and other inflammatory diseases, such as sarcoidosis (24) or primary large-vessel arteritides (15). In these diseases, FDG uptake is believed to be related to the high density of activated macrophages present in inflammatory lesions. While there is ongoing debate regarding the exact origin of the histiocytes in ECD (25, 26), these cells are of the monocyte/macrophage/dendritic cell lineages (27, 28). This may account for the observed uptake in other regions. While none of our patients with retroperitoneal infiltration have undergone any further followup PET scans, it seems very unlikely that retroperitoneal uptake was just delayed in these patients because most of them had longstanding disease. The specificity of both the initial and followup PET scans for assessing retroperitoneal and perirenal involvement was excellent (100%), suggesting that PET scanning may best serve in the followup, rather than in the initial assessment, of retroperitoneal involvement.

The sensitivity of the initial PET scanning for identifying paranasal sinus involvement was poor as compared with the corresponding CT scans (13.3%). CT scans mostly identified small, but typical, regions of osteosclerosis of the paranasal sinuses. These lesions were probably too small to be detected by PET scanning. The sensitivity of the PET scan increased to 41.7% during followup, probably because the metabolic activity of these lesions also increased, as indicated by the increased SUV_max in most paranasal sinus foci (Table 4).

The sensitivity of PET scanning for detecting orbital involvement was initially good (60%), but decreased to 10% when we analyzed followup data. The specificity was high for both the initial and followup PET scans (95% and 100%, respectively). Patient 16 had large bilateral lesions of both orbits that progressively diminished in size during followup as assessed with MRI, while FDG uptake disappeared even more rapidly (Table 3). Taken together, these results suggest that such “false-negative-like” findings of PET scans may de facto correspond to patients with marked and early therapeutic responses and, thus, that PET scanning may be very valuable in the therapeutic followup of orbital involvement.

Among the limitations of our study are its retrospective nature and the fact that data that were not available may have limited its robustness in terms of the statistics. However, this study is the largest single-center series of ECD patients ever published, and both the PET scan data and the other imaging data were available for comparison in more than 61% of cases. Another limitation is that we compared PET scans to bone scans, CT scans, and MRIs, assuming these imaging modalities are gold standards, whereas their actual sensitivity, specificity, PPV, and NPV in ECD patients are still unknown. This issue may not be seen as a major concern in our study because our objective was not to assess the exact diagnostic value of PET scanning in ECD, but rather, to directly compare PET scanning with the more traditional imaging modalities.

While radiologic investigations commonly used in ECD are unable to perform a global analysis of the lesions during a single session, we have shown that the whole-body PET scan was able to depict simultaneously many of the most relevant lesions encountered among ECD patients. The sensitivity varied greatly among the different sites of involvement we studied, but PET scanning showed excellent specificity when compared with most other imaging modalities. This suggests that PET scanning may be interesting in the initial assessment of selected patients, but may best serve during followup of the disease. Taken together, these data underline the key role of FDG-PET in the assessment of patients with ECD.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Amoura had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Arnaud, Malek, Archambaud, Kas, Piette, Amoura, Haroche.

Acquisition of data. Arnaud, Malek, Archambaud, Kas, Toledano, Drier, Zeitoun, Cluzel, Grenier, Chiras, Amoura, Haroche.

Analysis and interpretation of data. Arnaud, Malek, Archambaud, Kas, Toledano, Drier, Zeitoun, Cluzel, Grenier, Amoura, Haroche.