Metabolic active tumour volume quantified on [¹⁸F]FDG PET/CT further stratifies TNM stage IV non-small cell lung cancer patients

Abstract

Purpose

This study aimed to assess whether the whole body metabolic active tumour volume (MTV_WB), quantified on staging [¹⁸F]FDG PET/CT, could further stratify stage IV non-small cell lung cancer (NSCLC) patients.

Methods

A group of 160 stage IV NSCLC patients, submitted to staging [¹⁸F]FDG PET/CT between July 2010 and May 2020, were retrospectively evaluated. MTV_WB was quantified. Univariate and multivariate Cox regressions were carried out to assess correlation with overall survival (OS). C-statistic was used to test predictive power. Kaplan–Meier survival curves with Log-Rank tests were performed to compute statistical differences between strata from dichotomized variables and to calculate the estimated mean survival times (EMST). Survival rates at 1 and 5 years were calculated.

Results

MTV_WB was a statistically significant predictor of OS on univariate (p < 0.0001) and multivariate analyses (p < 0.0001). The multivariate model with MTV_WB (Cindex ± SE = 0.657 ± 0.024) worked significantly better as an OS predictor than the cTNM model (Cindex ± SE = 0.544 ± 0.028) (p = 0.003). An EMST of 29.207 ± 3.627(95% CI 22.099–36.316) months and an EMST of 10.904 ± 1.171(95% CI 8.609–13.199) months (Log-Rank p < 0.0001) were determined for patients with MTV_WB < 104.3 and MTV_WB ≥ 104.3, respectively. In subsamples of stage IVA (cut-off point = 114.5) and IVB patients (cut-off point = 191.1), statistically significant differences between EMST were also reported, with p-values of 0.0001 and 0.0002, respectively. In both substages and in the entire cohort, patients with MTV_WB ≥ cut-off points had lower EMST and survival rates.

Conclusion

Baseline MTV_WB, measured on staging [¹⁸F]FDG PET/CT, further stratifies stage IV NSCLC patients. This parameter is an independent predictor of OS and provides valuable prognostic information over the 8th edition of cTNM staging.

Background

Lung cancer is the leading cause of cancer-related mortality worldwide, constituting the type of cancer most commonly diagnosed in both genders combined (Ferlay et al. 2013). These tumours can be broadly divided into two main histological categories: small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC), the latter encompassing 85% of all lung cancer cases (Duma et al. 2019). For optimal management of NSCLC, it is crucial to classify patients with tumour, node and metastasis (TNM) staging. These tumours are frequently not diagnosed until metastatic disease is present (Duma et al. 2019; Kocher et al. 2015) and the 5-year overall survival (OS) rates are still extremely poor in stage IV patients (William et al. 2009).

Comprehensive staging remains the most important tool for prognostic purposes (Thakur and Gadgeel 2016). However, it may not always provide the most reliable prediction since each stage comprises a highly heterogeneous population. For instance, the novel 8th edition of TNM has recently divided stage IV into two separate categories—IVA and IVB—since the prior single stage IV category did not account for the prognostic differences in these two new substages (Goldstraw 2016). Therefore, prognostic factors besides the clinical TNM (cTNM) staging system must be studied in order to better stratify patients and improve decision-making when selecting risk-adapted therapies. Other widely known patient-specific prognostic markers are age, gender, performance status and weight loss (Brundage et al. 2002).

Currently, [¹⁸F]FDG PET/CT represents a key component of the diagnostic algorithm of NSCLC, as well as of the evaluation of response to therapy and detection of recurrent disease (Groheux et al. 2016; Hanin et al. 2008; Kandathil et al. 2018). Maximum standardized uptake value (SUV_max), which is defined as the value of the voxel showing the highest uptake, is the main parameter used in clinical practice to measure [¹⁸F]FDG uptake (Groheux et al. 2016). Moreover, there have been several successful studies which demonstrated that SUV_max yields prognostic information (Hanin et al. 2008; Cerfolio et al. 2005). Despite that, some concerns arise when measuring SUV_max, such as random noise causing a single ‘hot’ pixel rather than an accurate altered uptake in the body (Obara and Pu 2013).

Furthermore, since SUV_max only constitutes a semi-quantitative measure, quantitative parameters including metabolic active tumour volume (MTV) and total lesion glycolysis (TLG) have been assessed to accurately reflect tumour burden. MTV consists in the metabolic active tumour volume measured on PET/CT with a segmentation technique, whilst TLG is calculated by multiplying MTV by SUV_mean. Similarly to SUV_max, these volumetric parameters have been extensively studied for their ability to predict disease progression in NSCLC (Obara and Pu 2013; Liao et al. 2012a; Kim et al. 2012; Chen et al. 2012) and in other types of cancer (La et al. 2009; Berkowitz et al. 2008; Roedl et al. 2008). Lee et al. were the first to demonstrate that MTV was a prognostic factor independent from other established markers in lung cancer (Lee et al. 2007). They also hypothesized that this result implied that some prognostic markers, especially stage, could easily depend on more significant underlying factors, such as tumour burden (Lee et al. 2007). Some studies even postulated that MTV was superior to SUV_max (Liao et al. 2012a, b) in terms of prognostic value and, more importantly, not inferior to cTNM staging itself (Lapa et al. 2017).

Effectively, our centre has previously conducted a study regarding this matter, in which the prognostic significance of MTV of the whole body (MTV_WB) was compared to the stratifying power of cTNM staging in a cohort of all stages (Lapa et al. 2017). In fact, this previous work proved that MTV_WB further stratified NSCLC patients and proposed a new index containing both cTNM staging and MTV_WB. However, these results were achieved before the new division of stage IV, thus we found it relevant to verify whether MTV_WB would still carry prognostic importance in this group of patients, even considering this division.

To expand on the aforementioned hypothesis, the main goal of this study was to demonstrate whether MTV_WB could further stratify stage IV NSCLC patients, over the standard 8th edition cTNM staging system, considering that mortality in this stage remains so high and optimal prognostic algorithms are needed to enrol these patients in certain clinical trials. This was achieved by testing its stratifying power and comparing its OS predictive ability with that of conventional cTNM staging.

Materials and methods

Study design

This retrospective study was conducted in the Department of Nuclear Medicine of the Centro Hospitalar e Universitário de Coimbra (CHUC) in February 2021. Ethical approval for the study was obtained from the CHUC Ethics Committee and principles from the Declaration of Helsinki were fully met. Informed consent was not required for this type of retrospective analysis.

Study population

We conducted a retrospective review of the medical records of patients diagnosed with NSCLC in our institution between 2010 and 2020. We selected the 160 consecutive patients based on the following inclusion criteria: (1) all patients had histological confirmation of the disease, (2) were submitted to [¹⁸F]FDG PET/CT at our institution for initial staging and (3) were stage IV at diagnosis. These PET/CT scans were performed from July 2010 to May 2020. Exclusion criteria comprised: (1) presence of brain metastases (diagnosed by magnetic resonance) and (2) presence of other past or concurrent malignancies.

The PET/CT scans were performed before any therapeutic intervention. After attribution of cTNM staging and histological characterization of the lung tumour, patients were treated according to the most appropriate therapeutic strategies for their clinical condition, in accordance with the good practice guidelines at the time of treatment.

[¹⁸F]FDG PET/CT acquisition protocol

This was a unicentre study and the [¹⁸F]FDG PET/CT scans were conducted according to the institution’s existing protocol: patients completed a 6 h fast and their glycaemic levels were below 144 mg/dL prior to intravenous [¹⁸F]FDG administration. The administered activities were calculated based on the European Association of Nuclear Medicine Guidelines for tumour imaging on [¹⁸F]FDG PET/CT—minimum [¹⁸F]FDG (MBq) recommended for systems that apply a PET bed overlap of ≤ 30% = 14 (MBq.min.bed − 1 kg − 1) × patient weight (kg)/emission acquisition duration per bed position (min.bed − 1) (Boellaard et al. 2015). Images were then acquired, also according to these guidelines, after the recommended 60-min interval, with a range of 55–75 min (Boellaard et al. 2015). The variations observed in the administered activities and biodistribution times were associated with the usual conditions of clinical practice (Graham et al. 2011). Patients were positioned in dorsal decubitus and whole-body images were acquired using a General Electric Discovery ST PET/CT scanner (GE Healthcare, Waukesha, WI, USA). The acquisition parameters of CT for attenuation correction and anatomic mapping were as follows: 120 kV, smart mA (with current values between 10 and 200 mA and noise index 35), pitch 1.5:1, rotation 0.5 s and slice thickness 3.75 mm. The PET emission study was obtained in 3-D mode with an acquisition time of 3 min per table position, as per the manufacturer’s recommendations. The collected data were reconstructed with a Field Of View diameter of 70 cm and 256 × 256 matrix using the VUE Point 3-D iterative reconstruction algorithm, with two iterations, 35 subsets and 4 mm full width at half maximum post-reconstruction filter.

Data collection

All data was transcribed into a randomized database sheet. Age at the time of PET/CT, gender and the cTNM stage assigned to each patient were recorded. To be consistent throughout our study, we reviewed the group of patients whose staging was previously performed with the AJCC 7th edition of the TNM staging and grouped them into stage IVA and IVB according to the AJCC 8th edition guidelines (Goldstraw 2016). Histological types were also recorded.

The [¹⁸F]FDG PET/CT scans were retrospectively evaluated on a dedicated post-processing workstation (Advanced Windows 4.4 GE Medical Systems, Milwaukee, USA). Each patient’s lesions were delineated and evaluated using the Volume Computer Assisted Reading (PET_VCAR) software (version vxtl_8_3_65). The PET_VCAR software generated whole body 3-D regions of interest, based on the pre-defined threshold SUV value of 2.5. Regions corresponding to physiological uptake and/or uptake in benign lesions were manually excluded according to the consensus between two nuclear medicine specialists. After this initial post processing step, 3-D regions of interest, corresponding to the primary lung tumour, lymph node metastasis and all distant metastatic lesions, were obtained. A quantitative analysis was performed by the software to calculate MTV_WB. Additionally, SUV_max was also logged.

The primary endpoint of the study was OS. OS was calculated from the date of the initial baseline staging PET/CT scan to the date of death from any cause, based on the follow-up and records described above. The patients last known to be alive were censored at the date of the end of the study (February 11, 2021).

Statistical analysis

The values of the quantitative data were presented with minimum–maximum (mean ± standard deviation) or median (interquartile range), categorical data with absolute and relative frequency—n (%) −, and OS times with the estimated mean.

Univariate analyses using Cox proportional hazards regression were run for the total study population to assess the relationship between all the variables logged and survival. MTV_WB, cTNM staging, SUV_max and the other patient specific factors such as age at the time of PET/CT, gender and histological type were also submitted to a multivariate Cox regression.

The OS time predictive abilities of cTMN staging and MTV_WB were evaluated using Harrell-Concordance indexes (higher values indicating better discriminatory power). Multivariate Cox regressions adjusted for age, gender and histological type were run for both cTNM staging and MTV_WB separately. C indexes of each multivariate model were then determined with R software package ‘SurvComp’, that also provided standard errors, confidence intervals and p values for both. Afterwards, the OS predictive abilities of both multivariate models were compared using the same package (Harrell et al. 1996).

An optimal cut-off point for MTV_WB was computed for stage IV patients using the ‘cutp’ function of the R software package ‘SurvMisc’ (Contal and O’Quigley 1999). The cut-off point with the lowest p values was selected. The total study population was then divided into two groups based on the cut-off point chosen. Kaplan–Meier analysis with the Log-Rank test was used to compare estimated mean survival time between the two groups. The 1-year and 5-year survival rates were computed and compared between subjects above and below the cut-off point. Estimated mean survival times and survival rates at 1 and 5 years of stage IVA and IVB patients were also computed and compared for validation purposes.

The same procedures described above were used to select the best cut-off points for MTV_WB in each patient subgroup, defined by cTNM stages IVA and IVB, and to compare the estimated mean survival times, as well as the 1-year and 5-year survival rates, between subjects who were above and below the cut-off points.

A two-tailed p values of less than 0.05 was considered statistically significant for all tests performed. Analyses were performed using SPSS software (version 27; Armonk, NY, USA: IBM Corp) and R software (R Foundation for Statistical Computing, Vienna, Austria).

Results

Patient characteristics

The descriptive analysis of age at the time of PET/CT, gender, histological type, SUV_max and MTV_WB in the 160 stage IV patients, divided in IVA and IVB, is shown in Table 1. There were 114 (71.3%) men and 46 (28.7%) women, aged 34–88 years (mean ± SD = 66.0 ± 10.829). Histological findings consisted mainly in adenocarcinoma (n = 102; 63.7%). The remaining types are described thoroughly in Table 1. Stage-wise, 70 (43.8%) patients had stage IVA, whilst the other 90 (56.3%) patients had stage IVB NSCLC. Overall measurements of SUV_max ranged from 3.0–45.6 (mean ± SD = 14.6 ± 6.621). MTV_WB ranged from 0.2 to 1181.0 cm³ (mean ± SD = 196.8 ± 234.419). Follow-up time ranged from 0.39 to 104.25 months (mean ± SD = 16.691 ± 17.840).

Table 1.

Characterization of the enrolled patients

Characteristics	Values
Age at PET/CT (years)	Mean, 66.0; SD, 10.829; range, 34–88
Gender
Male	114 (71.3%)
Female	46 (28.7%)
Histological type
Adenocarcinoma	102 (63.7%)
Epidermoid carcinoma	26 (16.3%)
Adenosquamous carcinoma	11 (6.9%)
Pleomorphic carcinoma	10 (6.3%)
Sarcomatoid carcinoma	4 (2.5%)
Adenomucinous carcinoma	7 (4.4%)
cTNM stage IV
IVA	70 (43.8%)
IVB	90 (56.3%)
SUVmax	Mean, 14.6; SD, 6.621; range, 3.0–45.6
MTVWB (cm3)	Mean, 196.8; SD, 234.419; range, 0.2–1181.0
Follow-up time (months)	Mean, 16.691; SD, 17.840; range, 0.39–104.25

cm³ cubic centimetre, cTNM, clinical tumour, node, metastasis, MTV_WB metabolic active tumour volume of the whole body, PET/CT positron emission tomography/computed tomography, SD standard deviation, SUV_max maximum standardized uptake value

The distribution of MTV_WB values in the stage IVA and IVB subsamples is depicted in Table 2. The median value obtained from IVA patients (65.9) is considerably lower than in stage IVB patients (168.6). Stage IVA patients had measurements from 0.2 to 1181.0 cm3 (mean ± SD = 128.0 ± 191.9), whilst stage IVB patients ranged from 12.1 to 1093.2 cm3 (mean ± SD = 250.3 ± 251.0).

Table 2.

Descriptive analysis of MTV_WB in the subsamples stage IVA and IVB

	Mean ± SD	Min	Max	Median	IQR
cTNM stage IVA	128.0 ± 191.9	0.2	1181.0	65.9	24.1–141.9
cTNM stage IVB	250.3 ± 251.0	12.1	1093.2	168.6	83.2–327.9

cTNM clinical tumour, node, metastasis, IQR interquartile range, Min minimum, Max maximum, MTV_WB metabolic active tumour volume of the whole body, SD standard deviation

MTV_WB as a predictor of overall survival

In the univariate analyses, gender, cTNM stage, SUV_max and MTV_WB were statistically significant (Table 3). Age and histological type were not; however, they were included in the multivariate model, since they are relevant baseline factors. In the multivariate analysis (Table 3), age remained significant (p = 0.004), as well as stage (p = 0.012). Notably, SUV_max was not an independent predictor of survival (p = 0.256), contrary to MTV_WB, which remained significant (p < 0.0001) and proved to be an independent OS predictor.

Table 3.

Univariate and multivariate analysis

	Univariate model			Multivariate model
	HR	CI (95%)	p value	HR	CI (95%)	p value
Age	1.014	0.997–1.031	0.104	1.029	1.009–1.049	0.004
Gender
Male	Reference			Reference
Female	0.622	0.423–0.914	0.016	0.608	0.399–0.925	0.020
Histological type
Adenocarcinoma	Reference			Reference
Epidermoid Ca	1.115	0.708–1.754	0.639	1.021	0.633–1.647	0.931
Adenosquamous Ca	0.977	0.490–1.943	0.946	1.406	0.667–2.959	0.370
Pleomorphic Ca	2.622	1.301–5.286	0.007	1.183	0.547–2.561	0.668
Sarcomatoid Ca	4.097	1.462–11.474	0.007	3.244	1.078–9.757	0.036
Adenomucinous Ca	0.944	0.383–2.328	0.900	1.394	0.536–3.627	0.496
cTNM stage IV
IVA	Reference			Reference
IVB	1.693	1.193–2.403	0.0029	1.622	1.111–2.369	0.012
SUVmax	1.026	1.005–1.048	0.016	1.015	0.989–1.042	0.256
MTVWB	1.002	1.001–1.003	< 0.0001	1.002	1.001–1.003	< 0.0001

Association of overall survival with age, gender, histological type, cTNM stage, SUV_max and MTV_WB Statistically significant results are presented in bold

Ca. carcinoma, CI confidence interval, cTNM clinical tumour, node, metastasis, HR hazard ratio, MTV_WB, metabolic active tumour volume of the whole body, SUV_max maximum standardized uptake value

Given these findings, predictive abilities of two multivariate models adjusted for gender, age and histological type, one containing cTNM and the other MTV_WB, were compared using C-statistic. The model with cTNM staging was not a statistically significant predictor (p = 0.123), in contrast to the model containing MTV_WB (p < 0.0001). The predictive ability of the model with MTV_WB was significantly better than the one with cTNM staging (p = 0.003) (Table 4).

Table 4.

Comparison of cTNM staging and MTV_WB overall survival predictive abilities

	C index			Comparison
	C index ± SE	CI (95%)	p value	p value
Multivariate model with cTNM	0.544 ± 0.028	0.488–0.599	0.123	0.003
Multivariate model with MTVWB	0.657 ± 0.024	0.609–0.704	< 0.0001

Statistically significant results are presented in bold

CI confidence interval, cTNM, clinical tumour, node, metastasis, MTV_WB metabolic active tumour volume of the whole body, SE standard error

MTV_WB with a cut-off point as a predictor of overall survival in stage IV patients

To employ MTV_WB in clinical practice, we dichotomized the variable with a calculated cut-off point. The value of 104.3 (p < 0.0001) was identified as the optimal cut-off point for the whole sample. Patients with MTV_WB < 104.3 had an estimated mean survival time of 29.207 ± 3.627 months (95% CI 22.099–36.316), while those with MTV_WB ≥ 104.3 had an estimated mean survival time of 10.904 ± 1.171 months (95% CI: 8.609–13.199). There was a statistically significant difference in the estimated mean survival times, in months, (Log-Rank Chi-Square = 27.165; p < 0.0001) between the two groups of patients (Fig. 1).

Fig.1.

Kaplan–Meier curves comparing overall survival between groups as a function of MTV_WB in the total study population

The probability of survival above or below the MTV_WB cut-off point at 1 and 5 years after diagnosis was calculated. The 1-year survival rate was 39 ± 6% (mean ± standard error) for patients with MTV_WB < 104.3 and only 12 ± 4% for patients with MTV_WB ≥ 104.3. The 5-year survival rate was 9 ± 4% for patients with MTV_WB < 104.3 and there were no survivors in the group of patients with MTV_WB ≥ 104.3 (Table 5).

Table 5.

Survival rate (%) (mean ± standard error) according to the cut-off point defined for MTV_WB at each cTNM stage and substages IVA and IVB, and according to cTNM staging (IVA and IVB)

FU	IV		IVA		IVB		Stage IV
	< 104.3	≥ 104.3	< 114.5	≥ 114.5	< 191.1	≥ 191.1	IVA	IVB
1 year	39 ± 6	12 ± 4	45 ± 8	15 ± 7	27 ± 7	7 ± 4	34 ± 6	17 ± 4
5 year	9 ± 4	0	11 ± 5	0	0	0	7 ± 4	0

Statistically significant results are presented in bold

cTNM clinical tumour, node, metastasis, FU follow-up, MTV_WB metabolic active tumour volume of the whole body

MTV_WB with a cut-off point as a predictor of overall survival in each subsample – stage IVA and IVB

Although we proved that MTV_WB can further stratify stage IV patients, this may be evident considering that this stage has been recently separated into two substages. Therefore, it made sense to search for cut-off points in each subsample. Additionally, as expected, there was a statistically significant difference in estimated mean survival times between stage IVA and IVB patients (p = 0.003). The survival curves for these substages are shown in Fig. 2. The estimated mean survival times as a function of cTNM stage are depicted in Table 6 and survival rates at one and 5 years of stage IVA and IVB patients are shown in Table 5.

Fig. 2.

Kaplan–Meier curves comparing overall survival between groups as a function of cTNM stages—IVA and IVB—in the total study population

Table 6.

Estimated mean survival time, in months, according to cTNM staging and according to the cut-off point defined for MTV_WB in each cTNM stage

Stage	EMST ± SE	CI (95%)	p*	MTVWB	EMST ± SE	CI (95%)	p*
IVA	26.374 ± 3.657	19.205–33.542	0.003	< 114.5	33.951 ± 4.965	24.219–43.682	0.0001
				≥ 114.5	10.467 ± 1.680	7.174–13.760
IVB	13.833 ± 1.464	10.964–16.702		< 191.1	18.412 ± 2.242	14.018–22.806	0.0002
				≥ 191.1	8.552 ± 1.326	5.953–11.151

Statistically significant results are presented in bold

CI confidence interval, cTNM clinical tumour, node, metastasis, EMST estimated mean survival time, MTV_WB metabolic active tumour volume of the whole body, SE standard error

*Log-rank test

The identified optimal MTV_WB cut-off points were 114.5 (p = 0.02) for stage IVA and 191.1 (p = 0.005) for stage IVB. The estimated mean survival times of stage IVA and IVB patients, as a function of the respective MTV_WB are depicted in Table 6. Patients with values above the cut-off point at each category had worse prognosis. There was a statistically significant difference in estimated mean survival times, in months, between patient groups of both subsamples. Survival curves are represented in Figs. 3 and 4 (p = 0.0001 for stage IVA and p = 0.0002 for stage IVB).

Fig. 3.

Kaplan–Meier curves comparing overall survival between groups as a function of MTV_WB in stage IVA patients

Fig. 4.

Kaplan–Meier curves comparing overall survival between groups as a function of MTV_WB in stage IVB patients

The 1-year and 5-year survival rates for the groups above and below the MTV_WB cut-off points at each substage were also determined. Patients with MTV_WB values above the cut-off points had lower survival rates than patients with MTV_WB values below them (Table 5).

Discussion

A growing body of literature has been consistently proving that MTV adds stratification power in NSCLC patients (Obara and Pu 2013; Liao et al. 2012a; Kim et al. 2012). However, this parameter has yet to be included in official staging guidelines. Thus, we found it relevant to further investigate this premise, particularly in patients with metastatic lesions, whose survival depends on the optimization of stratification and, consequently, the selection of risk-adapted therapies. In fact, our results demonstrate that MTV_WB can subcategorize patients, specifically these patients with metastasized cancer. Firstly, MTV_WB was an independent predictor in multivariate analysis. Secondly, apart from the fact that MTV_WB proved to be a better OS predictor in comparison to cTNM staging through C-statistic, both survival curves as a function of an optimal MTV_WB cut-off point in stage IVA and IVB patients had statistically significant differences. Interestingly, it can also be noted that stage IVB patients below their computed cut-off point had a higher estimated mean survival time than stage IVA patients with values above their own cut-off point. We hypothesize that this could indicate that the updated cTNM is still not sufficient to classify these cases.

For validation purposes, we compared the 5-year survival rates of our cohort to the documented ones present in the proposals for new guidelines concerning the classification and cTNM staging of lung cancer. Our patients had a 5-year survival rate of 7 ± 4% (mean ± standard error) in stage IVA and 0% in stage IVB. These guidelines show similar survival rates: 10% and 0%, respectively (Goldstraw 2016). For this reason, we believe our sample was representative. Moreover, our cohort had similar numbers of stage IVA (70) and IVB (90) patients, allowing for a concise analysis of both groups.

Recently, Pu et al. conducted an analysis of a quite large sample of NSCLC patients of all stages, divided according to the new guidelines, where they validated a novel MTV_WB risk stratification system (Pu et al. 2018), reporting results consistent with ours. Accordingly, Pellegrino and colleagues also concluded that MTV_WB was an independent predictor of OS in all stages (Pellegrino et al. 2019). In spite of this, they raised an important question about the absence of consensus on the optimal technique for MTV delineation. Our centre used a threshold of 2.5 for SUV, in accordance with previous studies (Pellegrino et al. 2019; Im et al. 2015). Nonetheless, it should be noted that not all centres perform these measurements as described. We concur and corroborate that it is necessary to validate a certain method and threshold for more reproducible results. It is noteworthy to mention that this choice of threshold intensity value did not affect the consistency of our measurements, since the same value was used for the whole sample (Lee et al. 2007).

Moreover, there have been three previous studies regarding MTV and only stage IV patients, albeit considering the old cTNM staging guidelines. Firstly, a study with 92 consecutive patients with newly diagnosed stage IV NSCLC indicated that MTV was a prognostic marker at the whole-body tumour burden level, and at the primary tumour level as well (Liao et al. 2012a). However, our sample was larger, and their results were obtained before the new guidelines had divided stage IV patients. This supports our hypothesis that this novel division is not sufficient, considering that our results were still concordant, and thus MTV_WB would improve stratification. In contrast, two other studies reported that only MTV from primary lung lesions at PET/CT for staging purposes had prognostic value (Yoo et al. 2012; Lee et al. 2016). Yet, Yoo et al. measured MTV of primary lung tumour and MTV-torso instead of MTV_WB in accordance with their country’s standard protocol (Yoo et al. 2012) and Lee and colleagues evaluated MTV at primary lesion level separately from MTV at the node and metastasis level (Lee et al. 2016); therefore, comparisons cannot be accurately drawn.

Some of these studies assessed both MTV and the other volumetric value—TLG. Supposedly, TLG could be more promising since it combines volumetric and metabolic information. However, previous works did not report superiority of TLG (Yoo et al. 2012; Zhang et al. 2013), and Zhang et al. added that, in future clinical practice, MTV_WB would be sufficient for measuring metabolic tumour burden in NSCLC, as there is no demonstrable advantage of TLG of the whole body (TLG_WB) over MTV_WB (Zhang et al. 2013). In any case, calculation of TLG is provided by the PET/CT station after MTV quantification; therefore, they could be easily logged together.

Additionally, our results substantiate that SUV_max may not be an independent predictor of OS, highlighting the importance of volumetric parameters. As Huang and colleagues posited, SUV_max may not be the most accurate predictor, given that it only reflects a single-pixel value of maximal metabolic activity that will show greater response to treatment, and, thus, less impact on outcome (Huang et al. 2014). Conversely, MTV_WB will reflect metabolic changes throughout the entire tumour mass and, on top of that, it considers the whole-body tumour burden, yielding more precise information on prognosis.

It should be emphasized that therapeutic approaches have evolved drastically throughout this 10-year period. Stage IV patients in 2010 were mostly treated with empirical cytotoxic therapies (Herbst et al. 2018) whilst, more recently, immunotherapy and molecularly targeted therapies (Herbst et al. 2018; Arbour and Riely 2019) were introduced, aiming to improve survival. This might have compromised our analysis since OS could have been considerably better in recent times, even with larger tumour volumes, undermining the value of MTV_WB. In spite of that, our results remained significant. Additionally, our results were also not affected by the fact that different therapeutic schedules were chosen for each patient. In this way, we may conjecture that this marker is highly predictive, regardless of the selected treatment and/or schedule.

Despite the claimed advantages MTV_WB brings, its measurement can still be time-consuming. Our previous work presented a mean time of 5 min per patient (Lapa et al. 2017), yet this included early-stage patients with low MTVs. In this study, we determined a mean time of roughly 25 min per case. However, with the advent of deep learning networks and computer-aided automatic processing, the need for handcrafted radiomic features of images will be eliminated and this process will become more efficient (Avanzo et al. 2020; Zhong et al. 2018).

Our study does have some limitations. Firstly, we excluded patients with brain metastasis since [¹⁸F]FDG PET/CT does not accurately characterize them. This might disregard a non-negligible number of patients that can reach up to 26% of stage IV NSCLC cases (Waqar et al. 2018). Secondly, we were not able to retrieve the specific causes of death of each patient, considering some of them were unknown. However, the 5-year OS rate of stage IV patients remains extremely low in both substages (Goldstraw 2016). Hence, it is safe to assume that OS would be extremely close if not equal to disease-specific survival. In addition, the performance status at the time of PET/CT was also not documented since it was absent from some records. Nonetheless, poor performance status may simply depend on high tumour burden (Lee et al. 2007), which is already assessed by MTV_WB.

Finally, the retrospective nature of this study encompasses already well-known drawbacks, and further prospective studies with larger cohorts should be performed to confirm our results. More importantly, multicentre projects should be carried out, in order to establish optimal MTV_WB cut-off points, suitable for the general NSCLC population. This would constitute a simple method of introducing this parameter into more comprehensive staging algorithms, allowing for improved planning of clinical trials and individualized therapeutic strategies.

Conclusion

All things considered, the baseline metabolic active tumour volume of the whole body, measured on [¹⁸F]FDG PET/CT for staging purposes, further stratifies stage IV NSCLC patients. This parameter is an independent predictor of overall survival and provides valuable prognostic information over the 8th edition of cTNM staging. We suggest standardizing measurements between centres, as well as finding optimal cut-off points within stage IVA and IVB patients and incorporating them in official staging guidelines.

Abbreviations

[¹⁸F]FDG PET/CT

2-Deoxy-2-[18F]fluoro-D-glucose positron emission tomography/computed tomography

3-D

Three dimensional

AJCC

American Joint Committee on Cancer

Ca.

Carcinoma

CHUC

Centro Hospitalar e Universitário de Coimbra

CI

Confidence interval

C Index

Harrell’s Concordance Index

cm³

Cubic centimetre

CT

Computed tomography

cTNM staging

Clinical tumour, node, metastasis staging

EMST

Estimated mean survival time

FU

Follow-up

GE

General electric

HR

Hazard ratio

IMB

International Business Machines Corporation

IQR

Interquartile range

kg

Kilogram

kV

Kilovolt

mA

Milliampere

Max

Maximum

MBq

Megabecquerel

MBq.min.bed⁻¹ kg⁻¹

Megabecquerel.minute.bed⁻¹ kg⁻¹

Min

Minimum

min.bed⁻¹

Minute.bed⁻¹

mm

Millimetre

MTV

Metabolic active tumour volume

MTV_WB

Metabolic active tumour volume of the whole body

n

Number of individuals

NSCLC

Non-small cell lung carcinoma

NY

New York

OS

Overall survival

p

p Value

PET

Positron emission tomography

PET/CT

Positron emission tomography/computed tomography

PET_VCAR

Positron emission tomography with volume computer assisted reading

SD

Standard deviation

SE

Standard error

SPSS

Statistical package for the social sciences

SUV

Standardized uptake value

SUV_max

Maximum standardized uptake value

SUV_mean

Mean standardized uptake value

TLG

Total lesion glycolysis

TLG_WB

Total lesion glycolysis of the whole body

TNM staging

Tumour, node, metastasis staging

USA

United States of America

VUE

Virtually unenhanced

vxtl

Verxatile

WI

Wisconsin

Funding

No funding was received to assist with the preparation of this manuscript.

Availability of data and materials

The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Conflict of interest

The authors have no conflict of interest to declare that are relevant to the content of this article.

Ethics approval

Approval was obtained from the Ethics Committee of Centro Hospitalar e Universitário de Coimbra. The procedures used in this study adhere to the tenets of the Declaration of Helsinki.

Informed consent

The need for informed consent was waived.