Usefulness of the PET/CT to predict the progression and mortality risk in patients with diffuse interstitial lung disease

Objective: to assess the capacity of the PET/CT to predict pulmonary function deterioration and increased mortality risk in patients with idiopathic interstitial pneumonia (IIP) and to establish a possible SUV max cutoff which identies these patients. Material and methods: retrospective study between January 2007-December 2020. Inclusion criteria: patients > 18 years of age, diagnosed with IIP by PET/CT and pulmonary function test, with less than 6 months’ difference between the two tests. A study was made of the outcome variables associated with the PET/CT, the pulmonary function test measured at 2 stages (initially at the time of the PET/CT and at the end of follow-up), the mortality risk (using the GAP index) and the relationship between them all. Other variables of interest observed were age, sex, department requesting the PET/CT, indication, and the presence of lung cancer (LC). The statistical analysis was performed using the SPSS program. Results: 39 patients were analysed: 34 males (87%), with an age of 75 ± 8 years (mean ± DS). The mean ± SD of the SUV max was 2.57 ± 1.17, with a statistically insignicant difference (p = 0.670) between patients with and without lung cancer. LC was conrmed in 21 cases (54%). There is a small inverse correlation between the SUV max and the initial and nal predicted FVC% (r= -0.154, r= -0.252), together with a medium correlation for the initial and nal predicted DLCO% (r= -0.523, r= -0.514). The mean ± SD of the GAP index was 3.77 ± 1.08. There is a medium correlation between the SUV max and the mortality risk of stages I and II (r = 0.468). By means of ROC curve analysis, an SUV max of 2.2 was established to predict the fall of the FVC below 80%, of 1.9 to predict the fall of the DLCO below 60%, and of 2.15 to predict the progression from GAP stage I to II in mortality risk. Conclusions: There is an inverse correlation between the SUV max and the pulmonary function, together with a direct relationship between the SUV max and


Introduction
Diffuse interstitial lung diseases (DILD) constitute a heterogeneous group of clinicopathologic entities which present similar clinical, radiological, physiological and histological manifestations which diffusely affect the alveolar-interstitial space and the pulmonary vasculature [1].They are classi ed as: DILD of known cause or associated with other entities; or DILD of unknown cause. The latter group includes the idiopathic interstitial pneumonias (IIP) which constitute the most signi cant group (Table 1) [2]. In 2002, a panel of experts consisting of members of the European Respiratory Society (ERS) and the American Thoracic Society (ATS) drew up a consensus statement to de ne and classify IIPs [3]. This consensus was updated in 2013 with the inclusion of new clinical entities which had previously not been contemplated [4]. Among the IIPs, the brotic subgroup, which consists of idiopathic pulmonary brosis (IPF) and non-speci c interstitial pneumonias (NSIP), is characterized at a clinical level by a progressive dyspnoea on effort, persistent dry cough and velcro crackles in the physical examination. From a radiological point of view, a diffuse interstitial lung pattern stands out. Although this can be seen on a chest X-ray, maximum de nition is achieved by means of a high-resolution computed tomography (HRCT). This provides such precise patterns that it is sometimes su cient for a speci c diagnosis to be made with some con dence, especially in the case of IPF [5] as stated in the diagnostic standards of 2013 [6] and 2018 [7]. Regarding pulmonary function tests (PFT), these patients have a restrictive ventilatory impairment with a decrease in both the forced vital capacity (FVC) and the diffusion capacity of the lungs for carbon monoxide (DLCO). Nowadays, the progression of brotic DILDs is based on the clinical and radiological deterioration, together with decreased pulmonary function.
The GAP (Gender, Age and Physiology) index was created in 2012 and proven to predict the mortality risk in patients with IPF using a multidimensional index based on the combination of 4 variables (sex, age, predicted FVC % and predicted DLCO %). The total score ranges from 0 to 8 points and strati es the risk over 3 levels: stage I from 0-3 points; stage II from 4-5 points; and stage III from 6 to 8 points. The mean mortality of the stages I, II and III after one year is 5.6%, 16.2% and 39.2%; after 2 years 10.9%, 29.9% and 62.1%, and after 3 years 16.3%, 42.1% and 76.8%, respectively [8].
Furthermore, the positron emission tomography (PET) is a nuclear medicine diagnostic imaging technique which originated in the 1960s [9]. This test consists of injecting the patient with a tracer known as a radiopharmaceutical or radiotracer, which is the binding of a drug or physiological substance (with known pharmacokinetics, pharmacodynamics and biodistribution) with a radioactive positron-emitting isotope which indicates the location of this drug after scanning with a camera. However, this tracking lacks precise anatomical reference. To solve this problem, in 1994, Townsend and his colleagues proposed the fusion of the PET system with the computerized tomography (CT) [10] giving rise to the multimodal PET/CT which integrates the metabolic information of the PET with the morphology from the CT [11]. The most commonly used radiotracer is 18 F Fluoro-deoxy-D-glucose ( 18 F-FDG), a glucose analogue which enters the cells, be they tumorous or not, through the membrane receptors and is used as a metabolic marker [12]. The 18 F-FDG uptake can be determined, either through subjective visual observation of the images or by means of the semiquantative assessment of the concentration of the radiopharmaceutical, known as SUV max (Standardized Uptake Value), in a region of interest (ROI) within the PET image. The SUV max in a given tissue is calculated using the following formula: There are other measures of metabolic activity such as Metabolic Lung Volume (MLV) or the Total Lesion Glycolysis (TLG), however, the most common and easiest parameter to measure in clinical practice is the SUV max .
Some studies have proposed the usefulness of the PET/CT in assessing patients with DILD [13]. The avidity of 18 F-FDG in these patients is due to the active in ammatory process secondary to the migration of broblasts and to the accumulation of in ammatory cells (endothelial cells and macrophages) in the alveolar interstitium [14]. In 2016, Nobashi et al. published the rst study which associated the SUV max measured by PET/CT with the clinical and analytical deterioration and decreased pulmonary function in patients with DILD, as well as assessing its implication on the prognosis [15]. However, the available evidence is still too scarce to consider it as a useful test in the diagnostic process. Based on the hypothesis that the PET/CT could be useful for predicting the deterioration of the pulmonary function and the mortality risk in patients with IIP, the objectives of this study are: to study, in the patients with IIP, the relationship between the 18 F-FDG uptake both with the pulmonary function and the mortality risk; to assess the differences in metabolic activity in the PET/CT of patients with or without lung cancer (LC); and nally, to establish a possible SUV cutoff which would classify the IIP patients with a decreased pulmonary function and an increased mortality risk.

Material And Methods
An observational retrospective study was performed between the Multidisciplinary DILD Department, the Nuclear Medicine Department and the Public Health and Preventative Medicine Department in a tertiary hospital in Madrid (Spain). The patients were taken from the database of the DILD Department and the study was carried out between January 2007 and December 2020. Participating patients were over 18 years of age and diagnosed with IIP, who had undergone a PET/CT and PFT, with less than 6 months' time between the two tests.
Patients with DILD of known cause or associated with other entities were excluded, as were those with idiopathic DILD which did not meet the diagnosis of brotic IIP (IPF and NSIP).
Outcome variables associated with the PET/CT, the PFT, and the mortality risk were studied, together with the relationship between them.
The variables associated with the PET/CT were: SUV max (measured in the interstitial pattern area of pulmonary parenchyma with greater metabolic uptake and free of tumour lesions) and the site (pulmonary lobe) where this measurement was made. The measurements were made at the time of the study by a radiologist and nuclear medicine physician.
The variables associated with the PFT (predicted FVC% and predicted DLCO%) were collected on two occasions. The 'initial' measurements were essential to the study and had to be taken at the time of the PET/CT within +/-6 months, and the ' nal' measurements correspond to the last ones made on the patient during follow-up.
The GAP index was used to study the mortality risk using the initial PFT which coincide with the PET/CT test.
Other variables of interest studied were age, sex, department requesting the PET/CT, indication for performing it and the presence of LC.
In the descriptive study of the data, the mean ± SD was used on symmetric variables and the median and the interquartile range [IQR] on asymmetric variables for the quantitative variables. Frequency distribution was used for the qualitative variables. The association between the quantitative variable outcomes, such as the SUV max and the mortality risk, was studied by means of the Pearson correlation coe cient or Spearman's rank correlation coe cient in the cases where there was a small sample size. To associate quantitative variables between groups, the non-parametric tests of Mann-Whitney or Kruskal-Wallis were used for groups of two categories or more, respectively. ROC curve analysis was used to obtain a global measure of the test accuracy for the set of possible SUV max cutoff points which would allow classi cation of the patients with a decrease in pulmonary function (predicted FVC ≤ 80% and predicted DLCO ≤ 60% ) and the mortality risk. The statistical analysis was performed using the IBM SPSS Statistics v21 software.

Results
There are 323 patients in our database with brotic IIP (IPF and NSIP) and 39 of them ful l the inclusion criteria: 34 males (87%) and 5 females (13%), with an age of 75 ± 8 years (mean ± SD). The population characteristics are described in Table 2.
In all cases, the indication for performing the PET/CT was the suspicion or pathology of a tumour, with LC con rmed in 21 patients (54%).
The SUV max measurement was taken in the right inferior lobe in 20 patients (51.3%), in the left inferior lobe in 15 (38.5%), in the left superior lobe in 3 (7.7%) and the right superior lobe in 1 (2.5%).
The mean ± SD of the SUV max of all the studies was 2.57 ± 1.17. There is a difference between the mean SUV max of IIP patients with LC (2.42 ± 0.84) compared to the IIP patients without LC (2.76 ± 1.5) although this is not statistically signi cant (p = 0.670). Figure 1 shows the images of a patient with IIP and associated LC.
All the patients have initial PFTs (the DLCO could not be measured in one patient due to a tracheostomy) and 21 (54%) underwent nal PFTs. The drop in nal tests was due to the passing of 8 patients, a change of residence in one case, and in 9 cases, to recent diagnoses for which there has been no time to provide a follow-up visit. The mean ± SD of the PFTs were: Initial predicted FVC% 93.53 ± 17.65, initial predicted DLCO% 53.45 ± 20.60, nal predicted FVC% 86.95 ± 22.45 and nal predicted DLCO% 40.09 ± 19.56. Table 3 shows the correlation between the SUV max and the initial (Pearson correlation) and nal PFTs (Spearman's correlation). There is a small inverse correlation in the initial PFTs between the SUV max and the initial predicted FVC% (r= -0.154) and a medium one between the SUV max and the predicted initial DLCO% (r= -0.523). Regarding the nal PFTs, there was also a small inverse correlation between the SUV max and the predicted nal FVC% (r= -0.252) and medium between the SUV max and the predicted nal DLCO% (r= -0.514).
Concerning the GAP index, used to stratify the mortality risk, the mean ± SD was 3.77 ± 1.08. Out of the whole series, 16 patients (41%) were at stage I, 21 patients (53.9%) at stage II, and 2 patients (5.1%) at stage III. The mean ± SD SUV max for each stage was 2.08 ± 0.70 for stage I; 2.66 ± 0.82 for II; and 5.50 ± 3.25 for stage III, with statistically signi cant differences (p = 0.031) between stages I and II. The differences with stage III could not be analysed due to the scarce number of patients in this stage. There is a medium direct correlation (Pearson's correlation) between the SUV max and the mortality risk between stages I and II (r = 0.468).
By means of ROC curve analysis, a cutoff has been established for the SUV max at 2.

Discussion
Most of the patients in the study population are males of advanced age, as is normal in patients with brotic IIP.
The high prevalence of LC in our study (54%) is due to all PET/CT tests being requested on suspicion and/or staging of a tumour pathology. It is worth highlighting the known relationship between IPF and LC, as IPF increases the risk of developing a LC by 7-20% according to the series [16].
With regard to the metabolic activity of the pulmonary parenchyma in the PET/CT of our population, the average values obtained both on the whole (2.57 ± 1.17) and in the patients with DILD without LC (2.76 ± 1.5) fall within the range of the previously published studies where the SUV max varies from 2.46 ± 0.76 in the study by Nobashi 15 to 3.7 ± 2.5 according to Justet [17]. A limitation of these results is the dispersion of the values, as shown by the large standard deviation observed in all the studies.
Attention is drawn to the lower mean value ± SD of the SUV max of patients with LC (2.42 ± 0.84) compared to those without it (2.76 ± 1.5), although this difference is not statistically signi cant (p = 0.670). This lower metabolic activity has already been described in the study by Yamamichi [18] which involved 120 patients with IPF and LC. Here, the mean SUV max in the IPF area was 1.88 ± 0.76, which is lower than in previously cited studies without LC.
A total of 89% of the SUV max measurements have been made in inferior lobes due to the relationship with the apico-basal gradient of involvement in the majority of brotic IIP cases.
Concerning the PFTs, the predicted FVC% values measured are more conserved than the predicted DLCO%, which are moderately diminished. A reason for this is that many patients also have associated pulmonary emphysema, which is why they have these falsely conserved lung volumes.
In our study, there is a small inverse correlation between the SUV max and the initial (r= -0.154) and nal predicted FVC% (r= -0.252), together with a medium correlation between the SUV max and the initial (r= -0.523) and nal predicted DLCO% (r= -0.514). This inverse correlation has already been reported in the study by Lee et al. [19] published in 2014, involving 8 patients, although in this case, the correlation was medium both for the predicted FVC% and for the predicted DLCO% (r= -0.6 and r= -0.7, respectively). What stands out when comparing this study to ours is the similarity with regard to the predicted DLCO% and the difference with the predicted FVC%, which could be explained by the sample size (8 versus 39) or by the good conservation of the predicted FVC% of our patients (be it due to the presence of associated pulmonary emphysema or because the IIPs have a variable and unpredictable course [20]).
Concerning the mortality risk and GAP index calculation, the initial PFTs were used, which should be carried out together with the PET/CT with a difference of +/-6 months. This ensures time bias is avoided. It is striking that 23 patients (58.9%) in our study are in risk groups II and III, with a mortality after 3 years of 42.1 and 76.8% respectively. The mean ± SD of the SUV max increases in tandem with the GAP stage: 2.08 ± 0.70 for stage I; 2.66 ± 0.82 for stage II; and 5.50 ± 3.25 for stage III. This is a statistically signi cant difference (p = 0.031) between stages I and II, but comparison with stage III was not possible due to the small sample size.
We have aimed to de ne a cutoff for the SUV max to identify/classify patients with IIP who would have a decreased pulmonary function and an increased mortality risk. Although we initially considered dividing the sample between patients with and without LC (given that the mean SUV max was different for both groups), we nally considered both samples to be homogeneous as the difference is not statistically signi cant (p = 0.670) and we can thus increase the sample size. By means of ROC curve analysis, an SUV max of 2.2 was established to classify the patients with a predicted FVC% below 80%, of 1.9 for patients with a predicted DLCO% below 60%, and of 2.15 to predict progression from stage I to II in the mortality risk. (Figure 2). These cutoffs could have implications for the prognosis of IIPs and have an in uence on the treatment of patients with associated LC. To this end, in the aforementioned study by Yamamichi et al., a cutoff of 1.69 is established to predict the risk of acute exacerbation in patients with DILD and LC after thoracic surgery.
The limitations of our study lie in the retrospective design, the fact it was performed in a single centre, and the small sample size. The study population is not greater because we are dealing with an observational study of healthcare practice in the present climate and currently, the PET/CT is not indicated in either the diagnosis or follow-up of brotic IIPs.
However, its strengths are the vast database of the Multidisciplinary DILD Department (which collects a large number of patients assessed despite these disorders being considered as rare or in the minority [21, Page 8/15 22,23,24]) and its focus on brotic IIPs which are those which have the worse prognosis.

Conclusions
There is an inverse correlation between the SUV max and the pulmonary function (small for the predicted FVC% and medium for the predicted DLCO%) and a medium direct correlation between the SUV max and the mortality risk according to the GAP index. Therefore, the PET/CT may be useful for predicting the deterioration in pulmonary function and the mortality risk. However, multicentre studies are required with a greater number of patients to establish an optimal SUV max cutoff which allows prognosis and treatment decisions to be made both for brotic IIPs and for those associated with LC. Availability of data and materials

List Of Abbreviations
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.