[¹⁸F]FDG PET/CT predicts progression-free survival in patients with idiopathic pulmonary fibrosis

We asked all consecutive patients with a diagnosis of IPF, with a typical usual interstitial pattern (UIP) on high resolution CT of the chest (HRCT), to undergo [¹⁸F]FDG PET. According to ATS/ERS/JRS/ALAT guidelines, a typical UIP pattern was defined as the combination of reticular abnormalities, honeycombing with or without traction bronchiectasis, predominantly localized in sub-pleural areas with absence of features listed as inconsistent [19]. Patients receiving oral corticosteroids, immunosuppressants, or any antifibrotic therapy (including pirfenidone and nintedanib) within 3 months before PET completion were excluded. All patients received a diagnosis after a multidisciplinary discussion involving review of all clinical, functional, imaging and pathological data, according to French and international guidelines [19, 20]. Former smokers were defined by smoking cessation for at least 6 months.

For all patients, we assessed clinical data, bronchoalveolar lavage (BAL) cytology results, pulmonary function test results, and 6MWT distance performed within 1 month after PET. We applied the GAP index and staging system to each patient to obtain the GAP stage according to Ley et al. [6].

HRCT was obtained within 2 months after PET evaluation. The extent of fibrosis and ground glass was quantified for each lung lobe according to Kazerooni et al. [21]. The fibrosis score and ground-glass score for the left and right lung was calculated as the sum of the scores for each lobe. The total lung score was calculated as the geometric mean of the left and right lung scores. Duration of follow-up was defined as the time between PET completion and the date of the latest news, transplantation or death. During follow-up, we analyzed the changes in FVC and DLCO during the 12 months after PET completion. Acute exacerbation, defined according to Collard et al [22], was recorded. In further analysis, disease progression was defined as all-cause mortality, acute exacerbation, 10% or more decline in absolute FVC value at 12 months, or 15% or more decline in absolute DLCO value at 12 months, according to Collard et al. [4].

The control group included patients with gastrointestinal neuroendocrine tumor without thoracic localization who underwent [¹⁸F]FDG PET for pre-therapeutic evaluation of the tumor. None of these patients had evidence of current or past respiratory disease or had received thoracic radiotherapy or systemic corticosteroids. None had parenchymal abnormalities on chest CT coupled with PET.

[¹⁸F]FDG PET/CT involved a PET-CT scanner (GE Discovery 690, GE Health-care, Milwaukee, WI, USA). All patients fasted for at least 6 h before the examination. The serum glucose levels were < 9 mmol/L. PET-CT images were obtained 60 min after injection of 4 MBq/kg [¹⁸F]FDG. We performed iterative construction with attenuation correction, scatter correction, variant point spread function compensation and time of flight. The time of the image acquisition at 60 min was 3 min per bed. The regions of interest were manually identified. A visual analysis described the regional distribution of FDG uptake. Quantitative analysis was then performed. The SUV mean and SUV max were obtained for the right and left lung. Hereafter, the SUV mean is the geometric mean of SUV mean for the left and right lungs and the SUV max is the geometric mean of SUV max for the left and right lungs. The whole lung was delineated by an automatic lung extraction available in Planet software (DOSIsoft, Cachan, France) (Fig. 1, outlined yellow area). The bronchi were not excluded. The uptake from liver mediastinum (including heart but excepting adenopathy) was manually excluded when needed. Figure 1 shows the representative regions of interest used for analysis in a control patient (at the top) and in IPF patient (at the bottom). The area outlined in yellow represents the lung volume. The area outlined in red represents the metabolic lung volume. The MLV was determined considering a threshold of 1, considering the normal lung SUV mean to be 0.5. This was determined by measuring the mean SUV in lung of normal patients. We then applied a safety margin by using a threshold of 1. Therefore, we ensured that the MLV contained no or very few normal lung voxels. The relevant PET-related variables were extracted for the MLV and the whole lung. The variables were the maximum SUV (SUV max) and average SUV (SUV mean), from which we derived the TLG [ie, the product of the MLV and SUV mean [23]].

Statistical analysis involved use of GraphPad v6 and R v2.15. Because the data followed a normal distribution, we used parametric tests for analysis. Student t test was used to compare data for two groups and Pearson correlation test to determine correlation between two variables. Survival was estimated by the Kaplan-Meier survival curves and compared by log-rank test. Univariate and multivariate Cox proportional-hazards models were used to analyze the relationship between the variables of interest and survival and progression-free survival, estimating hazard ratios (HRs) and 95% confidence intervals (CIs). Because SUV, TLG and MLV were highly correlated (r > 0.9), they were not included together in the multivariate models. P < 0.05 was considered statistically significant.

In the control group, the lung SUV mean was 0.50 ± 0.11 and SUV max was 1.84 ± 0.82. SUV mean and SUV max did not differ between the left and right lungs. Because of the lack of hypermetabolic volume in controls, the TLG value was null.

All IPF patients showed increased [¹⁸F]FDG uptake as compared with controls, both in the quantitative and visual analysis. SUV mean and SUV max were higher for IPF patients than controls (1.0 ± 0.4, p < 0.00001, and 3.8 ± 2.5, p < 0.0001, respectively) (Fig. 2). SUV mean and SUV max did not differ between the right and left lungs (Table 2). The mean TLG was 1426 ± 1030 cm³ and mean MLV 950 ± 571 cm³. Increased pulmonary [¹⁸F]FDG uptake was predominant in sub-pleural areas and corresponded to honeycombing/reticulation areas.

We found no significant correlation between SUV mean, SUV max, MLV or TLG and epidemiological data, BAL cytology analysis or extent of fibrosis indicated by HRCT fibrosis and ground-glass scores. By contrast, we found a negative correlation between SUV mean, SUV max, MLV, TLG and lung function test results for FVC, DLCO, force expiratory volume in 1 s and total lung capacity (expressed in liters or as a percentage predicted value), which indicates increased uptake of [¹⁸F]FDG correlated with more severe disease (Fig. 3). We also found a negative correlation between partial pressure of oxygen (PaO2) and SUV mean and SUV max (Fig. 3). Furthermore, SUV max was negatively correlated with 6MWT distance, with no correlation between SUV mean or TLG and this variable (Fig. 4). High GAP stage was associated with high SUV mean, MLV and TLG (Fig. 5).

SUV mean, SUV max, MLV or TLG did not differ between former or active smokers and non-smokers and were not correlated with pack-years.

Follow-up results of pulmonary function tests were available for 22 patients: the mean FVC decline was 290 mL ± 251 mL (8.4% ± 8.0%, absolute value) and mean DLCO decline 10.1% ± 8.1% (absolute value). We found no correlation between the FVC and DLCO decline and SUV mean, SUV max, MLV or TLG. We then assessed [¹⁸F]FDG uptake by disease progression (Table 3), defined as all-cause mortality, acute exacerbation, 10% or more decline in FVC (absolute value) at 12 months, or 15% or more decline in DLCO (absolute value) at 12 months. In all, 12 patients (44%) were considered to have disease progression. TLG and MLV were significantly higher in patients with than without disease progression (Table 3). Similarly, TLG and MLV were higher for non-survivors than survivors during follow-up. SUV mean and SUV max did not differ between patients with and without disease progression and survivors and non-survivors (Table 3).

On univariate analysis, risk of death was increased for patients with FVC or DLCO decline and those with increased TLG or MLV; risk of death was increased with high SUV mean but not significantly (Table 4). On multivariate analysis, FVC (likelihood ratio test = 7.03, p = 0.02) and DLCO (likelihood ratio test = 6.58, p = 0.04) were the only independent factors associated with survival.

On univariate analysis, risk of death or disease progression was increased for patients with high SUV mean, TLG or MLV during the 12 months after PET completion (Table 5). In a multivariable model including SUV mean, TLG or MLV together with age, FVC and DLCO, risk of death or disease progression remained increased with high TLG and MLV (Table 5) even when the GAP index was included in the model (Table 5).

When the cohort of patients was divided by median TLG value (1218 cm³), progression-free survival was better for patients with TLG values below than above the threshold (367 days [317 - ∞] vs 211 [55 - ∞]; p = 0.02) (Fig. 6). We observed the same result when the cohort was divided by median MLV (862 cm³). Progression-free survival was better for patients with MLV values below than above the threshold (367 days [317 - ∞] vs 211 days [55 - ∞]; p = 0.02).