Quantitative 3D scintigraphy shows increased muscular uptake of pyrophosphate in idiopathic inflammatory myopathy

For the purpose of scintigraphy, 550 MBq of ^99mTc-PYP was administered intravenously. Imaging was performed on a Siemens Symbia T16 SPECT/CT scanner with low-energy high-resolution collimators with a 15% window centered on the 140 keV photo peak of ^99mTc. All acute patients had a whole-body scan 10 min p.i. as described in previous protocols [24]. All patients and controls underwent SPECT/CT of the thorax (arms along the body except in one case), and all but one acute patient had SPECT/CT of the pelvis and thighs 3 h p.i. SPECT parameters were 64 projections, 128 × 128 matrix, 40 s/frame at the upper body, and 20 s/frame at the lower body; the lower body scan was accomplished as a two-bed acquisition. CT was performed as a low-dose non-contrast-enhanced scan (130 kV, 20 mAs). SPECT data were reconstructed iteratively (four iterations, four subsets) with scatter and attenuation correction as well as resolution recovery and postfiltered with an 8-mm Gaussian filter. Three-dimensional models (maximum intensity projections) were interpreted visually and transaxial slices quantitatively. Since the investigators themselves collected data, they were not blinded at the time of scintigraphy, but data processing was performed later without looking at the clinical data. Unfortunately, the postponement of the processing resulted in a loss of data for some of the participants.

A qualitative assessment of tracer uptake including intensity, pattern, and symmetry was performed for both the upper and lower body in patients and controls. Images were analyzed on the GE Xeleris workstation (GE Healthcare Denmark, Brøndby, Denmark). First, two readers (KT and JS) evaluated scans by consensus (first observation). Prompted by previous scoring systems for planar ^99mTc-PYP images with focus on tracer uptake of cardiac tissue relative to that of the ribs [21], we graded peripheral muscular uptake compared to uptake in adjacent bones from 1 to 4 (1 = uptake predominantly in bone, 2 = bones >> soft tissue, 3 = bones ≥ soft tissue, and 4 = bones < soft tissue). Tracer uptake pattern was considered patchy in case of distinct spots with high uptake in muscles. Scans were described as asymmetrical if hot spots were seen on one side only and symmetrical if the muscular tracer uptake was diffuse or if hot spots were equally found on both sides. Later, a third reader (KB-O) evaluated a subset of the scans at a separate time point with the purpose of testing reproducibility (second observation). This reader was completely blinded to the clinical data.

Tracer uptake was measured in bilateral thigh muscles after semi-automated delineation based on CT scans as illustrated in Fig. 1. Siemens Inveon Research Workplace software (Siemens Healthcare, Ballerup, Denmark) was used to manually define upper and lower demarcation lines 5 cm below the trochanter major and 8 cm above the knee joint, respectively. Within the volume in between these demarcation lines, automated thresholding segmentation based on Hounsfield units (HU) was applied. Voxels with values within the range 200–2000 HU were defined as bone. Considering the spatial resolution of SPECT, 15 mm was then added to the bone volume in all directions in order to avoid spillover from bone uptake into the muscle volume. Voxel values within the range 0–90 HU were defined as muscle. The rather wide range of muscle HU values was chosen in light of a high noise level of the low-dose CT and varying fatty infiltrations of muscles of patients and healthy subjects. Feasibility of automated segmentation of muscles of the right-sided upper limb, shoulder girdle, and neck was also tested for in a small subset of patients and controls. Demarcation lines were placed right below the chin and at the elbow level. Again, voxels with values within the range 200–2000 HU were considered as bone and after expansion by 15 mm in all directions omitted from the volume of interest (VOI). Contrary to results for the lower limbs, however, voxel values within the range 0–90 HU were insufficient in defining muscles, probably due to higher noise levels. Including voxels with values within 0–150 HU yielded a fair delineation of the muscles judged visually.

The mean number of counts per voxel (cps mL⁻¹) in each VOI was registered, and the mean voxel activity (Bq mL⁻¹) was calculated from the knowledge of the camera sensitivity (cps MBq⁻¹). Camera sensitivity was measured using a known amount of ^99mTc [25]. The activity was decay corrected according to the time of injection and normalized to injected activity (MBq) per body weight (kg) of the participant. The result was a mean standardized uptake value (SUV_mean) measured in grams per milliliter. Based on a linear fit of the slice-by-slice SUV_mean down along each subject’s thigh muscles, a gradient (g mL⁻¹ mm⁻¹) was assessed.

Results are presented as frequencies (percentages) for categorical variables and descriptive statistics like mean (range) or mean ± standard error of the mean (SEM) for continuous variables, supplemented by 95% confidence intervals (95% CI) when appropriate. Differences in frequency distributions were compared by Fisher’s exact test or the chi-square test. Intergroup differences in continuous variables were tested by the unpaired Student’s t test or the Wilcoxon rank-sum test. The correlation between ordinal variables was assessed by means of Spearman’s rho. Results from interobserver analyses were reported as proportions of agreement and reliability in terms of weighted kappa (Cicchetti-Allison weights [26]) with bootstrapped 95% CI according to the guidelines proposed by Kottner et al. [27]. Univariate linear regression was used to explore the relationship between the SUV_mean and clinical parameters of disease activity according to the International Myositis Assessment and Clinical Study Group (IMACS) [28]. Goodness of fit of the model was assessed with the R-squared statistics. The significance level was set at 5%. Statistical analyses were performed using Stata/IC 11.2 (©StataCorp LP, College Station, TX, USA).

Early whole-body scans showed diffuse activity in soft tissue and organs corresponding to the perfusion of these and were judged not to contribute to the diagnosis of IIM.

Figure 2 shows examples of scintigraphy after 3 h, demonstrating tracer uptake in the soft tissue of a patient (right side of Fig. 2a, b), whereas in a control person, mainly the skeleton is visualized (left side of Fig. 2a, b). Figure 2c illustrates the patchy muscular uptake of a patient.

Results from the visual interpretation of the tracer uptake are shown in Table 2. According to the first observation, the median upper limb intensity score was 2 in patients and 1 in controls. According to the second observation, it was 3 and 1, respectively. From the first observation, the median lower limb intensity score was 2 in both groups with different distributions; in patients, the majority scored 2, whereas in the control group, frequencies of scores 1 and 2 were nearly equal, and a minimal number scored higher than 2. From the second observation, the median lower limb score was 2 and 1, respectively. Intergroup comparisons of the intensity scores of upper and lower extremities showed that high scores were significantly more prevalent in patients than in controls (p < 0.0001) in both observations. The correlation between upper and lower limb scores was 0.64 in patients and 0.63 in controls in the first observation and 0.57 versus 0.71 in the second observation, respectively (p for all < 0.005). The first observers found more patients than controls to have a patchy distribution of scintigraphic activity (24 vs. 6% for the upper limbs, p = 0.01; 22 vs. 6% for the lower limbs, p = 0.03), while the second observer almost never concluded patchiness. There was no significant difference in the symmetry/asymmetry of the tracer uptake between patients and controls according to any of the observers. Qualitative parameters did not differ between PM and DM patients or between sexes (data not shown).

The proportion of agreement between observers was 0.54 [0.40;0.67] for the upper limb score and 0.50 [0.37;0.63] for the lower limb score. Kappa was 0.53 [0.37;0.69] (p < 0.0001) for the upper limb score and 0.47 [0.30;0.64] (p < 0.0001) for the lower limb score. Considering dichotomized scores (1 or 2 vs. 3 or 4), the proportion of agreement was 0.84 [0.71;0.92] for both upper and lower limb scores; kappa was 0.65 [0.43;0.88] (p < 0.0001) for the upper limb score and 0.59 [0.34;0.83] (p < 0.0001) for the lower limb score.

Semi-automated quantification was done in 83 patients and 46 controls. Results are shown in Fig. 3 and in Table 3. The SUV_mean of patients was 60% higher than that of the controls with no overlap between 95% confidence intervals. Ranges were, however, wide and overlapping (Fig. 3, upper row). In patients, the decrease of activity down along the thigh was 2.5 times than that of the control subjects (Fig. 3, lower row). Muscular ^99mTc-PYP uptake and activity gradient did not differ between PM and DM patients (data not shown). There were no differences in ^99mTc-PYP uptake according to sex, whereas women had a steeper gradient than men (− 0.00025 SUV mm⁻¹ vs. − 0.00001 SUV mm⁻¹, p = 0.0002). The muscular ^99mTc-PYP activity of patients correlated with clinical parameters of disease activity; the manual muscle test including eight muscle groups (MMT8), the Health Assessment Questionnaire (HAQ), physician global activity using a 10-cm visual analogue scale (VAS), and patient global activity, but not with the serum level of creatine kinase (Table 4).

Semi-automated quantification of the right-sided muscles of the upper body was done in nine patients and eight controls and revealed results similar to those of the lower body. Ratios between lower and upper body SUVs were 1.2 in patients and 1.1 in controls (p = 0.16), and in spite of the small sample, results from the upper body also differed significantly between patients and controls (SUV_mean 0.29 [0.22;0.36] vs. 0.21 [0.19;0.23], p = 0.02).

Seeing that the group of patients was heterogeneous, we did an exploratory comparison of the muscular uptake of acute patients with that of chronic patients as well—well aware of the inequality in size of these subsets (quantitative data available in 13 vs. 70 patients). The SUV_mean did not differ between patient groups (0.36 [0.32;0.40] in the acute vs. 0.34 [0.31;0.36] in the chronic, p = 0.17). The gradient was significantly steeper in acute than in chronic patients (− 0.00044 [− 0.00061;− 0.00027] vs. − 0.00021 [− 0.00027;− 0.00014], p = 0.01).

Previous papers on scintigraphic appearance of IIM were case reports or presentations of non-controlled, smaller studies. We collected a large material, and furthermore, we compared the patients with a control group. The patients were at different disease stages. We obtained data systematically from standardized VOIs instead of just reading the maximum number of counts in affected areas. Gradients were calculated from simple linear fits, and we did not elaborate on goodness of fits.

Interobserver analyses revealed some, but far from perfect, reproducibility of qualitative scores. According to the reliability measures proposed by Landis and Koch [42], three of our kappa values could be interpreted as moderate reliability and the kappa for dichotomized upper limb scores as substantial reliability. Agreement parameters [43] in terms of proportion of agreement were also not striking but somewhat higher when considering dichotomized results. This shows that a subjective evaluation cannot stand alone but supports the potential diagnostic value of an interpretation referring to a low (1 and 2) or a high (3 and 4) ^99mTc-PYP uptake in muscles compared to bone. Interobserver differences in scores could, in part, be caused by varying the use of the intensity scale during scan reading since adjustment of the intensity level could have an impact on the visualization of bone. The same applies to the judgment of patchiness on which the different observers in our groups did not agree. Again, this points to room for improvement of a qualitative evaluation.