Assessment of 18F-FDG uptake in idiopathic pulmonary fibrosis: influence of lung density changes

Idiopathic pulmonary fibrosis (IPF) is a chronic, irreversible, and progressive fibrosing interstitial lung disease leading to a terminal respiratory insufficiency (Raghu et al., 2011). Although the aetiology of IPF remains unknown, it appears that recurrent alveolar epithelial cell injury play a central role with an aberrant healing process characterized by accelerated apoptosis of type II alveolar epithelial cells, abnormal recruitment of fibroblasts (which in turn differentiate to myofibroblasts) and exaggerated production of extracellular matrix components (Spagnolo et al., 2014).

Prognosis of untreated IPF is poor with a median survival time of 2 to 3 years from the time of diagnosis (Raghu et al., 2011). Apart from an ineluctable evolution towards respiratory failure, IPF is associated with an increased incidence of lung cancer (Archontogeorgis et al., 2012). Its clinical course is unpredictable with the majority of patients experiencing a slow progression over many years, while others show an accelerated decline with or without acute exacerbations (Raghu et al., 2011). Identification and validation of predictive markers of the clinical evolution are therefore crucial.

Some prognostic factors are already known such as a decline in forced vital capacity (FVC) or carbon monoxyde diffusion (DLCO) (Raghu et al., 2011). The GAP risk assessment system is a clinical tool estimating the prognosis of patients with IPF. The GAP index is a composite based on gender (G), age (A), and physiological variables (P) (FVC, DLCO). This score predicts the overall risk of mortality of a patient but it does not inform about disease activity at a specific time point and about rate of evolution (Ley et al., 2012; Salisbury et al., 2016).

Value of 2-deoxy-2-(18F)fluoro-D-glucose positron emission tomography with computed tomography ([18F]-FDG PET/CT) has been recently studied in IPF. Several studies showed that IPF patients have increased lung uptake of [18F]-FDG in areas of reticular/honeycombing and ground-glass opacity on high-resolution CT (HRCT) (Groves et al., 2009) and in areas of normal pulmonary parenchyma (Win et al., 2014). Moreover, [18F]-FDG lung uptake seems to be correlated with the severity of the disease, and to reflect not only the inflammatory process but also fibrogenesis (Groves et al., 2009; Meissner et al., 2006). This is supported by a study in a murine model of pulmonary fibrosis showing [18F]-FDG uptake during the initial inflammatory phase of the bleomycin-induced lung injury but also in the latter fibrotic phase (Bondue et al., 2015).

Two recent studies evaluated the prognostic value of the standardized uptake value (SUV) in IPF patients (Lee et al., 2014; Justet et al., 2017). The mean lung SUV (SUVmean) was negatively correlated with FVC and DLCO (Lee et al., 2014; Justet et al., 2017), and the decline in FVC was associated with an increment in the maximum lung SUV (SUVmax) (Lee et al., 2014). This suggests that [18F]-FDG uptake is related to the functional impairment and clinical severity in IPF.

However, the prognostic value of the [18F]-FDG PET/CT in IPF could be explained – at least partially – by a relationship with lung density, depending on the way the pulmonary compartment is considered for the quantification of [18F]-FDG uptake, i.e. including or excluding the air volume from the lung volume of interest. Indeed, IPF lungs are smaller, with a thickened interstitium resulting in a higher tissue/air ratio compared to normal lung. Even if tissue uptake remains constant, a higher tissue/air ratio results in higher SUV values, when the whole lung volume is used to quantify [18F]-FDG uptake in the pulmonary tissue. Pulmonary function impairment evolves in parallel with lung density, so any relationship between pulmonary function and [18F]-FDG uptake in an undifferentiated lung volume may be driven by density changes, regardless of the metabolic demands by the pulmonary cells involved in inflammatory or fibrotic processes. So, if [18F]-FDG PET/CT is used to assess the intrinsic activity of the disease (and not only its severity), the influence of lung density should be taken into consideration in the [18F]-FDG uptake evaluation (Lambrou et al., 2011; Holman et al., 2015).

The aim of this study was to correlate the severity and the evolution of IPF to lung uptake of [18F]-FDG, evaluated by SUVmean and SUVmax, corrected (SUVmean-corr; SUVmax-corr) and uncorrected (SUVmean-uncorr; SUVmax-uncorr) for lung density. We hypothesize that [18F]-FDG lung uptake corrected for lung density can be a prognostic factor of the pulmonary function decline.

Between 2013 and 2016, 31 IPF patients were included in the study. Their clinical characteristics at the time of the inclusion are summarized in Table 1. Ten patients died during the study period, and three patients underwent lung transplantation. A total of 29 patients reached a follow-up of 1 year after the PET/CT and were classified as having a progressive (n = 15) or a non-progressive (n = 14) disease. Two patients were lost to follow-up. A [18F]-FDG PET/CT was performed at baseline (Fig. 1). The SUVmean and the SUVmax, corrected or not for the tissue fraction, were determined.

Characteristics	Patients (n = 31)	Progressive (n = 15)	Non-progressive (n = 14)	p value
Age (years, mean ± SD)	69.8 ± 8.6	70.0 ± 9.0	68.6 ± 8.6	p = NS
Sex ratio (M/F) (n)	23/8	12/3	9/5	p = NS
Smoking status (n)
Never smoker	4	2	2
Current smoker	0	0	0
Former smoker	25	12	11
Passive smoker	2	1	1
Treatments (n)
Pirfenidone	15	7	7
Nintedanib	15	8	6
No treatments	1	0	1
GAP score (n)
Stage 1	10	2	8
Stage 2	16	8	6
Stage 3	5	5	0
C-reactive protein (mg/L) (median, range)	3.2 (1.9–7)	2.7 (1.7–5.6)	3.75 (1.72–8.2)	p = NS
Pulmonary function test (mean ± SD):
FVC (% predicted)	73.9 ± 19.0	64.4 ± 13.5	83.8 ± 19.9	p = 0.005 *
DLCO (% predicted)	44.3 ± 11.7	41.5 ± 14.4	47.1 ± 8.3	p = NS
6MWT (n = 28) (mean ± SD)
SaO2 post-exercise (%)	89.1 ± 4.7	87.3 ± 4.0	90.9 ± 4.2	p = 0.035 *
Distance traveled (m)	464.4 ± 111.7	464.8 ± 118.4	460.4 ± 114.8	p = NS
Bronchoalveolar lavage cytology (n = 24) (median, range)
Cellularity (/mm3)	293 (140–450)	405 (130–567.5)	261.5 (132.5–420)	p = NS
Macrophages (%)	79 (65–89)	85.5 (60.2–92)	79 (65–85)	p = NS
Lymphocytes (%)	4 (2.5–7)	3.5 (1.75–7)	5.5 (2.5–10)	p = NS
Neutrophils (%)	8 (3–15)	6 (3.75–10)	8 (2–16)	p = NS
Eosinophils (%)	1 (0–3)	1 (0–9.5)	2 (0–3)	p = NS
PET/CT parameters (mean ± SD)
SUVmean uncorr	0.95 ± 0.19	1.00 ± 0.20	0.89 ± 0.17	p = NS
SUVmax uncorr	3.24 ± 0.73	3.34 ± 0.92	3.12 ± 0.52	p = NS
SUVmean corr	2.70 ± 0.35	2.72 ± 0.43	2.65 ± 0.29	p = NS
SUVmax corr	29.89 ± 6.43	30.20 ± 6.74	30.48 ± 6.26	p = NS
CT mean (HU) (mean ± SD)	−580.04 ± 57.34	− 563.09 ± 51.78	− 598.49 ± 61.51	p = NS

p is the value of unpaired Student t tests or Mann-Whitney test for simple comparisons between two groups (progressive vs. non-progressive disease)

NS not significant, SD standard deviation, FVC Forced vital capacity, DLCO carbon monoxide diffusion, 6MWT 6-min walking test, SaO₂ oxygen saturation, SUV standardized uptake value

The PET/CT analysis at baseline showed a SUVmean-uncorr of 0.95 ± 0.19 (mean ± SD), which was strongly correlated with the lung density assessed by the chest CT (CTmean: − 580.04 ± 57.34 HU, mean ± SD) (r = 0.755; p <  0.001) (Fig. 2). Conversely, when correcting the SUVmean for the tissue fraction (SUVmean-corr), this correlation with the lung density completely disappeared (r = − 0.091; p = 0.627).

Other strong correlations were observed between the lung volumes (FVC, TLC) measured at baseline and the SUVmean-uncorr (FVC: p = 0.004; TLC: p = 0.001) and the lung density (CTmean) (FVC: p = 0.001; TLC: p = 0.001) (Fig. 3a and b). Once again, these correlations disappeared when correcting for the tissue fraction (SUVmean-corr) (Fig. 3c).

Of note, no significant correlation was found between the SUV values and the cytological analysis of the BAL (Additional file 1: Table S1).

Using the SUVmean-corr as a PET marker independent of the influence of the lung density, we demonstrated correlations between the SUVmean-corr and the DLCO on one side (r = − 0.399; p = 0.026) and the oxygen saturation during the 6MWT (post-exercise value and changes pre- and post-exercise) on the other side (r = − 0.376 and r = 0.402; p = 0.049 and p = 0.034, respectively). Interestingly, these correlations were not present for the SUVmean-uncorr or SUVmax corrected or not for lung density (Tab. 2 and Fig. 4a-c).

Pearson correlation tests			SUVmean-uncorr	SUVmean-corr	SUVmax-uncorr	SUVmax-corr	CTmean (HU)
PFT	VC (%)	r	−0.447	− 0.065	− 0.408	−0.082	− 0.518
		p	0.013*	0.733	0.025*	0.666	0.003*
	FVC (%)	r	−0.505	−0.08	−0.413	− 0.072	−0.57
		p	0.004*	0.667	0.021*	0.7	0.001*
	TLC (%)	r	−0.568	−0.165	−0.527	− 0.003	−0.595
		p	0.001*	0.382	0.003*	0.987	0.001*
	DLCO (ml/ min/mmHg)	r	−0.247	−0.398	−0.056	− 0.268	0.013
		p	0.18	0.026*	0.764	0.145	0.945
6MWT	SaO2 post-exercise	r	−0.344	−0.376	−0.064	− 0.062	−0.18
		p	0.073	0.049*	0.746	0.755	0.36
	Δ SaO2 (between the pre- and post-exercise)	r	0.366	0.401	0.056	0.172	0.174
		p	0.055	0.034*	0.777	0.38	0.375
	Δ SaO2 post-exercise (between the test at the time of PET and the test at 1 year)	r	0.38	0.527	0.017	0.261	0.074
		p	0.12	0.024*	0.947	0.296	0.769
	Distance traveled	r	0.037	−0.217	0.354	0.032	0.199
		p	0.85	0.266	0.064	0.87	0.311
	SUV mean-uncorr	r					0.755
		p					< 0.001*
	SUV mean-corr	r					−0.091
		p					0.627

Data presented were obtained using Pearson correlation tests with significant correlations highlighted in bold (corresponding to p values < 0.05). The r corresponds to the correlation coefficient. PFT = pulmonary function tests

CV vital capacity, FVC Forced vital capacity, TLC Total lung capacity, DLCO carbon monoxide diffusion, 6MWT 6-min walking test, SaO₂ oxygen saturation, SUV standardized uptake value

The patients were also classified according to the GAP score. There was no significant difference in SUVmax-corr between patients in various GAP stages (Additional file 2: Figure S1; p = 0.271), whereas for what concerns SUVmean-corr a significant difference was noticed between patients in stage 3 and in stage 2 (p = 0.027) and between patients in stage 3 and in stage 1 (p = 0.037) (Fig. 4d). The average SUVmean-corr for stage 1 was 2.62 ± 0.23, for stage 2 was 2.62 ± 0.33 and for stage 3 was 3.08 ± 0.43.

Next, hypothesizing that SUVmean-corr could reflect the intrinsic activity of the disease, we assessed whether it has some predictive value on the clinical outcomes after 1 year of follow-up. A significant correlation was only found between SUVmean-corr and the increase in oxygen desaturation after the 6MWT, assessed by the difference in oxygen saturation at the end of the 6MWT between baseline values and those measured at 1 year (Additional file 3: Figure S2). No significant correlation was found between SUVmean-corr values and changes in lung volumes, walk distance and DLCO at 1 year (Additional file 1: Table S1). When patients with a progressive or a non-progressive disease were compared, no significant difference was found for SUVmean-uncorr (p = 0.124), SUVmean-corr (p = 0.626), SUVmax-uncorr (p = 0.434) and SUVmax-corr (p = 0.909) (Fig. 5a and Table 1). Finally, we divided patients according to their vital status at the end of the study (alive vs. death or lung transplantation). Once again, no difference was found for SUVmean-uncorr (p = 0.794), SUVmean-corr (p = 0.979), SUVmax-uncorr (p = 0.8) and SUVmax-corr (p = 0.961) (Fig. 5b).

The aim of this study was to evaluate the prognostic value of [18F]-FDG PET/CT in IPF patients. Contrary to previous studies evaluating the usefulness of [18F]-FDG PET/CT in IPF (Meissner et al., 2006; Lee et al., 2014; Justet et al., 2017; Umeda et al., 2015; Win et al., 2018; Nobashi et al., 2016), SUV and derived parameters were corrected for lung density in the present study. We hypothesized that correction for lung density is important since lung density increases with the development of fibrotic tissue, leading to reduced lung volumes and to higher tissue/air ratios compared to normal lung. Even if tissue uptake remains constant, a higher tissue/air ratio will result in higher SUV values. Therefore, if [18F]-FDG PET/CT is used to assess the intrinsic activity of the disease (and not only its severity), the influence of lung density should be taken into consideration in the [18F]-FDG uptake evaluation (Lambrou et al., 2011; Holman et al., 2015).

As expected, mean pulmonary density, as well as uncorrected SUVmean and SUVmax, were negatively correlated with lung volumes (VC, FVC, TLC) measured at the time of PET/CT. This is easily explained by the fact that with the development of fibrotic tissue, density increases while pulmonary compliance decreases, resulting in a reduction in lung volumes. These results are consistent with a pilot study showing that SUVmean uncorrected for density is negatively correlated with FVC (Lee et al., 2014) and other studies in which CTmean values were correlated with pulmonary physiological variables (FVC, TLC) (Ohkubo et al., 2016; Best et al., 2003). SUV corrected for the tissue fraction are not correlated with pulmonary volumes, probably because the volumes are strongly related to density, leaving no residual variability when this primary effect is corrected for. Normalization for lung density thus eliminates correlations between SUVmean and respiratory function variables and reduces the influence of changes in volumes on the metabolic measurements. The normalized measurements are then more directly related to cellular and tissue changes in the lung parenchyma itself. Accordingly, SUVmean-corr was correlated with the DLCO, and the significance of this correlation was supported by the correlation with the desaturation post-exercise during the 6MWT. In other words, increase glucose metabolism estimated by SUVmean-corr indicates a reduction in gas diffusion in the lungs and an increase in oxygen desaturation of the blood after exercise. The study therefore suggests a link between the pulmonary tissue glucose uptake and the status of the alveolar-capillary barrier. In IPF, there is a proliferation of fibroblasts (especially in fibroblastic foci), covered by activated epithelial cells (Wolters et al., 2014). Activated fibroblasts have an increased metabolic activity with increased glucose consumption and glycolysis (Xie et al., 2015). Studies also revealed an increased expression of the glucose transporter 1 (GLUT-1) in fibroblasts from IPF patients as well as in epithelial cells and inflammatory cells (El-Chemaly et al., 2013; Cho et al., 2017; Andrianifahanana et al., 2016). Interestingly, GLUT-1 is the main transporter of glucose and [18F]-FDG (El-Chemaly et al., 2013). The increase in SUVmean-corr could therefore be attributed to activated fibroblasts, epithelial cells, or possibly some inflammatory cells, either by an increase in metabolic activity per cell, or by an increase in their number per dense tissue volume, or by a combination of both phenomena. The impact of fibroblastic activity on [18F]-FDG uptake in IPF is supported by a preclinical study on a mouse model of pulmonary fibrosis. In this study, Bondue et al. demonstrated [18F]-FDG lung uptake at the late fibrotic stage of the model, whereas labeled leukocyte recruitment is greatly reduced at this stage (Bondue et al., 2015). The limited influence of the inflammatory cells on the [18F]-FDG uptake in IPF is confirmed in the present study by the lack of correlation between the SUV values and the BAL cytology, in line with similar results obtained by Justet and colleagues (Justet et al., 2017).

Altogether PET/CT evaluation in IPF could provide some additional information about disease activity and prognosis. This is supported by a recent publication showing that adding uncorrected PET parameters to the GAP stage (PET-modified GAP staging) increases the predictive power of the GAP classification (Win et al., 2018). In our study, we demonstrate that the SUVmean-corr is also higher in GAP stage 3. This could indicate that a higher SUVmean-corr is associated with a higher mortality risk. Nevertheless, only 5 patients were classified in stage 3. Moreover, there is a relatively important overlap of the values between GAP stages without a clear threshold above which patients could be clearly differentiated among these stages. Therefore, further studies are needed to determine the extent to which PET data can improve the predictive value of the GAP classification.

Our study presents several limitations: it is a single-center study based on a relatively small number of patients. Also, the follow-up period was limited to 1 year. Finally, most patients received an antifibrotic treatment (pirfenidone or nintedanib) during the follow-up period, which may change their clinical evolution and influence correlations with baseline PET/CT values. These limitations could perhaps explain the lack of correlation found between SUV values (corrected or uncorrected for lung density) and patients’ functional and clinical evolution at 1 year.

In conclusion, the estimated [18F]-FDG uptake with no correction for lung density is tightly related to the lung volumes measurements in IPF; this relation is strongly influenced by the lung density since it disappears when a correction for lung density is applied. To evaluate the intrinsic activity of the disease with [18F]-FDG at the pulmonary tissue level, this correction appears necessary. This correction reveals a relationship between [18F]-FDG uptake in the pulmonary tissue and gas transfer across the disturbed alveolar-capillary barrier in IPF. Despite a relationship with the GAP classification, we found no correlation between [18F]-FDG uptake and most indices of clinical evolution at 1 year. Further studies in larger populations and with a longer follow-up are warranted to establish the prognostic value of corrected SUV values in IPF.