Preoperative Maximum Standardized Uptake Value Associated with Recurrence Risk In Early Lung Cancer

Abstract

Background:

We aimed to investigate the maximum standardized uptake value (SUVmax) as a predictor of recurrence and timing of recurrence after resection of early-stage non-small cell lung cancer.

Methods:

We retrospectively reviewed patients from a single institution who underwent lobectomy for stage I-IIa non-small cell lung cancer from 2013–2018. Exclusion criteria included preoperative therapy and neuroendocrine histology. We collected recurrence and follow-up data, as well as preoperative SUVmax. A receiver operator characteristic curve was used to identify the optimal SUVmax for predicting recurrence. Kaplan-Meier curves and Cox Regression analyses were used to identify predictors of freedom from recurrence (FFR).

Results:

The study included 238 patients, 30(12.6%) of whom developed recurrence. The receiver operator characteristic curve had an area-under-the-curve of 0.671 and identified 4.93 as the optimal SUVmax cut-off. Patients were stratified into groups based on this value; each group included 119 patients. High SUVmax was associated with larger tumor size, poor differentiation, lymphovascular invasion, and shorter FFR. The proportion of patients without recurrence at 5 years in the low- and high-SUVmax groups were 92.4% and 73.4%, respectively (p<0.001). On univariate analysis, poor differentiation (HR:2.35, 95%CI:1.04–5.31; p=0.04), lymphovascular invasion (HR:3.19;95%CI:1.37–7.44;p=0.007), visceral pleural invasion (HR:2.33;95%CI:1.05–5.20;p=0.04), and SUVmax≥4.93 (HR:4.51;95%CI:1.84–11.03;p=0.001) predicted FFR. On multivariable analysis, only SUVmax≥4.93 remained significant (HR:5.36, 95%CI:1.50–19.17; p=0.01).

Conclusions:

SUVmax is independently associated with risk of recurrence after resection of early-stage lung cancer. SUVmax may be a valuable tool in stratifying patients with early-stage lung cancer for adjuvant therapy and surveillance frequency.

Graphical Abstract

Recurrence rates after surgical resection in early-stage non-small cell lung cancer (NSCLC) range from 30–60%¹. Recurrence timing demonstrates a tri-modal distribution, with the first peak occurring seven to nine months after surgery, and the second and third peaks occurring in the second and fourth years of surveillance, respectively^2,3. Early identification of recurrence provides the best opportunity for potentially curative treatment, and is best achieved with appropriately timed postoperative surveillance. However, optimal surveillance paradigms are poorly defined and current surveillance patterns vary, lacking uniformity among institutions and practitioners^4,5.

Preoperative staging workup in NSCLC often includes an F-fluorodeoxyglucose positron emission tomography (FDG-PET) scan. The maximum standardized uptake value (SUVmax) is a standard component of PET scan reporting and indicates the level of metabolic activity in bodily tissues. SUVmax demonstrates a bimodal distribution in multiple cancers, including NSCLC⁶. Higher SUVmax values have been associated with more poorly differentiated tumors, as well as higher expression of genes related to glycolysis and proliferation, with gene expression profiles resembling undifferentiated stem cells^6,7. Higher SUVmax is also associated with elevated risk of lymph node (LN) metastasis and worse prognosis after resection⁶.

The goal of this study was to investigate the prognostic significance of preoperative SUVmax in relation to the risk and timing of recurrence in patients who undergo lobectomy for early-stage node-negative NSCLC. We also evaluate the role of SUVmax in stratifying recurrence risk in the context of other high-risk features, and consider the implications for patient stratification for adjuvant therapy and tailored surveillance regimens.

Patients and Methods

We identified patients with pathologic stage I-IIa (T1-T2, N0) NSCLC who underwent lobectomy between January 2013 and July 2018⁸. Exclusion criteria included treatment with preoperative therapy, neuroendocrine histology, and unavailable preoperative PET or follow-up imaging. Patients who underwent surgical resection for recurrence of a previous tumor treated with curative-intent stereotactic body radiation therapy were also excluded. Patients with synchronous primary tumors treated by stereotactic body radiation therapy, rather than resection, or by sub-lobar resection, were excluded.

We collected demographic, clinicopathologic, and recurrence data. Preoperative SUVmax was obtained from preoperative PET reports performed at our institution or another institution when available, or from the summary of the PET report included in clinical notes for scans performed at another institution. For patients who had a preoperative PET without reported SUVmax, original images were reviewed by a thoracic radiologist to provide the SUVmax values retrospectively (CS). If patients had multiple synchronous stage I-IIa primary lung cancers that were treated by surgical resection, the highest SUVmax value was included.

Surgical Approach and Surveillance

All patients in this study underwent anatomic lobectomy with appropriate dissection of hilar and mediastinal nodal stations.

The current National Comprehensive Cancer Network (NCCN) guidelines recommend postoperative surveillance for stage I-IIa NSCLC to include history and physical examination with computed tomography (CT) scan of the chest every 6 months for the first 2–3 years, followed by low-dose non-contrast-enhanced CT of the chest annually thereafter⁹. The current surveillance practice at our institution is consistent with these guidelines and includes a clinic visit accompanied by CT of the chest every 6 months for two years, and then annually until 5 years after resection, provided there are no atypical findings or concerns that lead the clinician to recommend heightened observation. The patients are then transitioned to our survivorship clinic where they are followed by the Thoracic Surgery Nurse Practitioner on at least an annual basis, unless concerning findings are noted. Patients may also elect to transition their continued surveillance imaging to their primary care physician or oncology provider closer to home.

For this study, patients were considered to have recurrent disease if findings on surveillance imaging were suggestive of recurrence, with either biopsy-based pathologic confirmation or PET imaging findings compelling for recurrence. Local recurrences were defined as those occurring at the surgical resection margin, in the ipsilateral hilum, or ipsilateral mediastinum^2,10. Distant recurrences included all other sites of subsequent disease, including within an ipsilateral lobe of the lung or the ipsilateral pleura, as well as more remote metastases. A diagnosis of second lung primary (SLP) was considered when the confirmed histology differed from the first primary cancer, or in cases of the same histology, when the clinical presentation was most consistent with an SLP rather than a recurrence, based on imaging findings, lack of nodal involvement, and disease-free interval². Only first recurrence or SLP events were considered for the purpose of this study. Freedom from recurrence (FFR) was defined as the time from surgical resection to the date of first imaging demonstrating the subsequently confirmed recurrence. Patients without recurrence were censored at their most recent follow-up date or date of death. Due to the clinical subjectivity in distinguishing a SLP from a recurrence, patients who developed only a SLP, without recurrence of the resected tumor, were excluded from the recurrence analyses, but were included in the overall survival (OS) analysis. OS was calculated from date of resection to date of death, or most recent follow-up date.

Statistical Analysis

A receiver operator characteristic (ROC) curve was constructed to identify the SUVmax value with optimal sensitivity and specificity for recurrence, identified by the value with the greatest Youden’s J index. Patients were stratified by SUVmax according to this value. Groups were compared with standard statistical analyses. Kaplan-Meier survival curves were created and groups compared with log-rank test. Factors predictive of FFR were identified by univariate and multivariable Cox Regression analyses. All comparison tests were two-tailed and a p-value<0.05 was considered statistically significant. All statistical analysis was performed with IBM SPSS Statistics for Windows, Version 26.0 (IBM Corp. Released 2019. Armonk, NY, USA).

This study was approved by the University of Texas MD Anderson Cancer Center Institutional Review Board and was granted a waiver of informed consent. .

Results

We identified 238 patients who met our inclusion criteria. The demographic and clinicopathologic characteristics of the patient cohort are included in Table 1. The median (interquartile range, IQR) SUVmax for the cohort was 4.93(2.7–9.5). In patients whose disease recurred (N=30, 12.6%), the median SUVmax was 7.5(5.1–11.4) compared to 4.3(2.4–8.8) in those without recurrence (p=0.01).

Table 1.

Demographic and clinicopathologic features of the study cohort.

Variable	N=238	N(%)
Age (mean +/− SD), years		66.9 +/− 8.4
Male Sex		107(45.0%)
Race/Ethnicity	White	206(86.6%)
	Black	13(5.5%)
	Hispanic	6(2.5%)
	Asian	12(5.0%)
	Native American	1(0.4%)
Smoking Status	Never Smoker	53(22.3%)
	Previous/Current Smoker	185(77.7%)
Histology	Adenocarcinoma	184(77.3%)
	Squamous	45(18.9%)
	Large Cell	6(2.5%)
	Mucoepidermoid	2(0.8%)
	Adenosquamous	1(0.4%)
Tumor Size (median, IQR), cm		2.5(1.8–3.1)
Tumor Size Category	≤2cm	88(37.0%)
	>2–3cm	86(36.1%)
	>3–5cm	64(26.9%)
Pathologic T stage	T1	118(49.6%)
	T2	120(50.4%)
Tumor Differentiation	Well-differentiated	56(23.5%)
	Moderately-differentiated	135(56.7%)
	Poorly-differentiated	36(15.1%)
	Undifferentiated	1(0.4%)
	Unknown	10(4.2%)
Lymphovascular Invasion		32(13.4%)
Visceral Pleural Invasion		81(34.0%)
SUVmax (median, IQR)		4.9(2.7–9.5)
Recurrence	No recurrence	184(77.3%)
	Second Primary	24(10.1%)
	Recurrence	30(12.6%)
Recurrence Location	Locoregional	5(20.0%)
	Distant	16(53.3%)
	Locoregional and distant	8(26.7%)
Imaging follow-up time (median, IQR), months		49.4(26.8–63)
Follow-up time (median, IQR), months		52.5(33.5–68.0)

IQR, interquartile range; SD, standard deviation

An ROC curve was constructed for SUVmax with recurrence as the outcome (Figure 1). The area-under-the-curve (AUC) was 0.671. The highest Youden’s J index was identified at SUVmax of 4.93.

Figure 1.

Receiver-operator-characteristic (ROC) curve for recurrence based on maximum standardized uptake value (SUVmax). The area-under-the-curve is 0.671 (95%CI: 0.583–0.758, p=0.003).

Patients with a preoperative SUVmax<4.93 (low-SUVmax group, N=119) were compared to those with SUVmax≥4.93 (high-SUVmax group, N=119, Table 2). Patients in the high-SUVmax group displayed more frequent smoking history, squamous cell histology, larger tumors, and greater rates of poor tumor differentiation and lymphovascular invasion (LVI). The recurrence rate in the high-SUVmax group was 20.2% (24/119) compared to 5.0% (6/119) in the low-SUVmax group (p=0.001). The recurrence locations were similar. The rates of SLP were also similar, at approximately 10% in each group.

Table 2.

Demographic and clinicopathologic features of patients with low and high maximum standardized uptake values (SUVmax).

Variable		Low SUVmax (n=119)	High SUVmax (n=119)	P-value
Age (mean ± SD), years		66.5 ± 8.7	67.2 ± 8.1	0.52
Male Sex		49(41.2%)	58(48.7%)	0.30
Race/Ethnicity	White	98(82.4%)	108(90.8%)	0.03
	Black	6(5.0%)	7(5.9%)
	Hispanic	5(4.2%)	1(0.8%)
	Asian	10(8.4%)	2(1.7%)
	Native American	0(0.0%)	1(0.8%)
Smoking Status	Never Smoker	40(33.6%)	13(10.9%)	<0.001
	Previous/Current Smoker	79(66.4%)	106(89.1%)
Diabetes Mellitus		19 (16.0%)	17 (14.3%)	0.86
Renal Disease		0 (0.0%)	1 (0.8%)	1.00
Histology	Adenocarcinoma	113(95.0%)	71(59.7%)	<0.001
	Squamous	4(3.4%)	41(34.5%)
	Large Cell	0(0.0%)	6(5.0%)
	Mucoepidermoid	2(1.7%)	0(0.0%)
	Adenosquamous	0(0.0%)	1(0.8%)
Tumor Size (median, IQR), cm		2.1(1.5–2.8)	2.7(2.0–3.5)	<0.001
Tumor Size Category	≤2cm	58(48.7%)	30(25.2%)	<0.001
	>2–3cm	39(32.8%)	47(39.5%)
	>3–5cm	22(18.5%)	42(35.3%)
Pathologic T stage	T1	70(58.8%)	48(40.3%)	0.01
	T2	49(41.2%)	71(59.7%)
Tumor Differentiation	Well-differentiated	47(39.5%)	9(7.6%)	<0.001
	Moderately-differentiated	65(54.6%)	70(58.8%)
	Poorly-differentiated/Undifferentiated	4(3.4%)	33(27.7%)
	Unknown	3(2.5%)	7(5.9%)
Lymphovascular Invasion		6(5.0%)	26(21.8%)	<0.001
Visceral Pleural Invasion		33(27.7%)	48(40.3%)	0.07
SUVmax (median, IQR)		2.7(1.7–3.5)	9.5(6.8–14.5)
Recurrence	No recurrence	102(85.7%)	82(68.9%)	<0.001
	Second Primary	11(9.2%)	13(10.9%)
	Recurrence	6(5.0%)	24(20.2%)
Recurrence Location	Locoregional	1(16.7%)	5(20.8%)	0.84
	Distant	4(66.7%)	12(50.0%)
	Locoregional and distant	1(16.7%)	7(29.2%)
Imaging follow-up time (median, IQR), months		50.5(27.5–66.1)	49.2(25.9–61.3)
Follow-up time (median, IQR), months		53.6(31.9–68.7)	51.5(34.7–68.0)

IQR, interquartile range; SD, standard deviation; SUVmax, maximum standardized uptake value

Freedom From Recurrence

The median (IQR) imaging follow-up time was 50.5(27.5–66.1) months in the low-SUVmax group and 49.2(25.9–61.3) months in the high-SUVmax group. The 3- and 5-year FFR rates for the low-SUVmax group were 96.9% and 92.4%, respectively, compared to 80.9% and 73.4% in the high-SUVmax group (p<0.001) (Figure 2).

Figure 2.

Kaplan-Meier curves depicting (A) freedom from recurrence (FFR) and (B) overall survival (OS) in patients with a low or high maximum standardized uptake value (SUVmax), p<0.001 for both.

On univariate cox regression analysis, poorly/undifferentiated tumors (compared to well- or moderately-differentiated, HR:2.35, 95%CI:1.04–5.31; p=0.04), LVI (HR:3.19, 95%CI:1.37–7.44; p=0.007), visceral pleural invasion (HR:2.33, 95%CI:1.05–5.20; p=0.04), and high SUVmax (HR:4.51, 95%CI:1.84–11.03; p=0.001) were significantly associated with FFR (Table 3). Due to the low rate of recurrence in our early-stage cohort, our multivariable model was limited by degrees of freedom. Thus, we included variables with a p<0.05 in the multivariable analysis. Only SUVmax≥4.93 (HR:5.36, 95%CI:1.50–19.17; p=0.01) was associated with FFR on multivariable analysis (Table 3).

Table 3.

Univariate and multivariable cox regression proportional hazards models for freedom from recurrence and overall survival.

Freedom From Recurrencea	Univariate Analysis			Multivariable Analysis
	HR	95% CI	p	HR	95% CI	p
Age	1.00	0.96–1.05	0.92
Female Sex (v Male)	1.17	0.56–2.42	0.68
Race/Ethnicity White	ref		0.97
Black	0.65	0.09–4.79	0.67
Hispanic	1.16	0.16–8.54	0.89
Asian	1.47	0.35–6.19	0.60
Smoking history	0.87	0.37–2.03	0.75
Histology Adenocarcinoma	ref		0.50
Squamous Cell	0.74	0.26–2.15	0.59
Other	2.07	0.49–8.74	0.33
Tumor Size (cm)	1.33	0.92–1.91	0.13
Pathologic T stage – T2 v T1	2.11	0.99–4.52	0.054
Poor/Undifferentiated Tumor	2.35	1.04–5.31	0.04	1.47	0.57–3.76	0.42
Lymphovascular Invasion	3.19	1.37–7.44	0.007	1.25	0.44–3.57	0.68
Visceral Pleural Invasion	2.33	1.05–5.20	0.04	1.94	0.80–4.73	0.15
SUVmax ≥ 4.93	4.51	1.84–11.03	0.001	5.36	1.50–19.17	0.01
Overall Survival	Univariate Analysis			Multivariable Analysis
	HR	95% CI	p	HR	95% CI	p
Age	1.09	1.02–1.16	0.008	1.09	1.03–1.16	0.006
Female Sex (v Male)	0.62	0.26–1.49	0.29
Race/Ethnicity White	ref		1.0
Black	0.77	0.10–5.73	0.80
Smoking historyb	29.19	0.30–2873.1	0.15
Histology Adenocarcinoma	ref		0.08
Squamous Cell	2.83	1.14–7.04	0.03
Other	1.89	0.24–14.66	0.54
Tumor Size (cm)	1.44	0.93–2.21	0.10
Pathologic T stage – T2 v T1	2.00	0.80–5.03	0.14
Poor/Undifferentiated Tumor	3.56	1.45–8.71	0.005	2.30	0.91–5.78	0.08
Lymphovascular Invasion	1.88	0.62–5.74	0.27
Visceral Pleural Invasion	2.56	0.99–6.62	0.053
SUVmax ≥ 4.93	8.93	2.07–38.47	0.003	7.30	1.62–32.82	0.01

^aexcluded patients with SLP only from recurrence analysis.

^ball deceased patients had a positive smoking history.

SUVmax, maximum standardized uptake value

Figure 3 includes Kaplan-Meier FFR curves for the SUVmax groups, stratified by the presence of individual high-risk pathologic features. Additionally, patients were classified as having no NCCN-defined high-risk pathologic features (favorable-pathology group, n=101), or at least one high-risk feature (poor differentiation, LVI, visceral pleural invasion, or tumor≥4cm; unfavorable-pathology group, n=113). Figure 4 illustrates the FFR curves in these two groups based on SUVmax. In the favorable-pathology group, 6 patients (5.9%) developed recurrence, compared to 24 patients (21.2%) in the unfavorable group. Inclusion of high-SUVmax as a high-risk feature reclassifies 32 patients from the initial favorable-pathology group into the unfavorable-pathology group, 5 (15.6%) of whom developed recurrence. A single patient (1.4%) in the adjusted favorable-pathology group (no high-risk features and low SUVmax) with SUVmax of 4.6 developed recurrence 27.99 months after resection.

Figure 3.

Kaplan-Meier curves depicting freedom of recurrence (FFR) in low and high maximum standardized uptake value (SUVmax) groups, stratified by (A-B) pathologic T stage (C-D) tumor histologic differentiation; (E-F) the presence and absence of lymphovascular invasion (LVI); and (G-H) the presence and absence of visceral pleural invasion.

Figure 4.

Kaplan-Meier curves depicting freedom from recurrence (FFR) based on SUVmax in patients with (A) favorable-pathologic features (p=0.007) and (B) unfavorable-pathologic features based on National Comprehensive Cancer Network Guidelines (p=0.054). Unfavorable-pathology group includes patients that demonstrate one or more high-risk features, including poor tumor differentiation, lymphovascular invasion, visceral pleural invasion, or tumor size ≥ 4cm. Favorable-pathology group includes patients that do not demonstrate any of these features.

Distribution of Recurrence Timing and Prognosis

Patients who recurred were stratified into three groups to represent the tri-modal distribution for timing of recurrence: 1) <12 months after resection, 2) 12 to 36 months after resection, and 3) >36 months after resection. The median SUVmax of the groups were 9.16(7.8–10.0), 6.56(5.1–11.4), and 7.55(3.5–16.25), respectively (p=0.58, Figure 5).

Figure 5.

Histogram of post-resection timing of recurrence in the low and high maximum standardized uptake value (SUVmax) groups.

There were no differences in post-recurrence survival by SUVmax (p=0.952, Supplementary Figure 1).

Overall Survival

With regard to OS, at 3 and 5 years after resection, 98.1% and 98.1% of the low-SUVmax group was alive, compared to 89.9% and 81.8%, respectively, in the high-SUVmax group (p=0.005, Figure 2).

On univariate cox regression analysis, poorly/undifferentiated tumor (HR:3.56, 95%CI:1.45–8.71; p=0.005), age at operation (HR:1.09, 95%CI:1.02–1.16; p=0.008) and SUVmax≥4.93 (HR:8.93, 95%CI:2.07–38.47; p=0.003) were associated with OS (Table 3). Again, limited by degrees of freedom, only variables with a p<0.05 were included in the multivariable analysis, in which only age (HR:1.09, 95%CI:1.03–1.16; p=0.006) and SUVmax≥4.93 (HR:7.30, 95%CI:1.62–32.82; p=0.01), remained significant.

Patients with SLP only

The proportion of patients without SLP at 3 and 5 years post-resection in the low-SUVmax group were 95.5% and 90.9%, respectively, compared to 91.6% and 87.2% in the high-SUVmax group (p=0.631, Supplementary Figure 2). There was no difference in the SLP-free time, as calculated by Kaplan-Meier analysis.

Comment

In early-stage node-negative NSCLC, high SUVmax is associated with increased tumor size, smoking history, poor tumor differentiation, and LVI. Patients with tumors demonstrating preoperative SUVmax≥4.93 also experienced shorter FFR compared to those with preoperative SUVmax<4.93. Importantly, we’ve shown that for patients with early-stage, node-negative NSCLC treated with lobectomy, SUVmax is more strongly associated with recurrence than pathologic T stage, tumor histologic differentiation, the presence of LVI, or the presence of visceral pleural invasion. Preoperative SUVmax≥4.93 was also independently associated with decreased OS.

Multiple investigators have identified associations between preoperative PET parameters, specifically SUVmax, and recurrence and survival in NSCLC. However, the SUVmax cut-off identified to predict recurrence has varied widely, ranging from 2.4 to 20¹¹. Furthermore, many studies have included patients of all stages, which can influence the SUVmax cut-off values and recurrence rates. Our study included a homogenous group of pathologic stage I-IIa (T1–T2, N0) NSCLC, for which the postoperative treatment guidelines are the same, and identified a significantly increased hazard for recurrence and death in patients with a preoperative SUVmax≥4.93. Other studies limited to stage I-II lung cancer identified similar findings, though the reported SUVmax cut-off values span a wide range even in this early-stage patient group.^11–21 The variability in SUVmax cut-off values may be related to differences in PET imaging scanners or software used for measuring SUVmax and limits comparison of cut-off values between studies. While, we were able to include a larger cohort of patients, compared to most prior studies in early-stage disease, and to provide a modernized analysis by using the 8^th edition TNM staging recommendations⁸, we do not intend to propose a specific SUVmax cut-off value based on our findings to be applied broadly to clinical practice. Rather, our findings suggest that patients with very low tumor avidity have significantly lower risk of recurrence, particularly in the absence of other high-risk features.

We utilized the Youden’s J index to identify a SUVmax cut-off to optimize both sensitivity and specificity. While a lower SUVmax cut-off could improve sensitivity (at the expense of specificity) and identify a larger portion of patients who experienced recurrence, the AUC of 0.671 in the ROC curve demonstrates that SUVmax alone is not adequate to discriminate between high and low recurrence risk. Thus, consideration of SUVmax in the context of other high-risk factors is important.

Under the current NCCN guidelines, the presence of certain high-risk features in resected stage Ib and IIa patients, such as poor differentiation, vascular invasion, wedge resection, tumors >4cm, visceral pleural involvement, and unknown lymph node status, are indications for consideration of adjuvant therapy, which is not routinely recommended otherwise⁹. While our multivariable FFR model is limited by few degrees of freedom due to the low recurrence rate in this patient cohort, our stratified recurrence analyses, though not a perfect substitute, can provide some insight into the importance of SUVmax in the context of each of the other pathologic variables. Based on these analyses, SUVmax significantly impacts rates of recurrence in patients with well to moderate histologic differentiation, a finding that is echoed in patients without LVI, those with and without visceral pleural invasion, and those with pathologic T2 tumors (and approached significance for those with T1 tumors). In the absence of the NCCN high-risk features, the inclusion of SUVmax as a high-risk feature identified additional high-risk patients, as 15.6% of those with a high SUVmax and no other high-risk characteristic, developed recurrence. The impact of SUVmax on recurrence rates, independent of these other traditional pathologic prognostic indicators is important for stratifying additional patients as high-risk, with consideration to receive adjuvant therapy, for more frequent postoperative surveillance, and for inclusion in future clinical trials. Patients lacking classic prognostic markers may be erroneously classified as “low-risk” in the case of patients with a high preoperative SUVmax. Tumors with high SUVmax have gene expression profiles resembling undifferentiated stem cells, with a presumably greater malignant potential, even in the absence of identifiable high-risk pathologic features.⁶

We did not find a difference in SUVmax based on timing of recurrence. Patients in both SUVmax groups experienced recurrence at each time mode of the trimodal recurrence distribution, indicating that the commonly accepted surveillance of five years is justified in both groups. However, patients with low SUVmax, at lower risk of recurrence overall, may not require as frequent surveillance as patients with a high SUVmax, particularly in the absence of other poor prognostic features. On the contrary, patients with high SUVmax, may benefit from more frequent surveillance for a longer period of time, given their increased risk. We are unable to comment on the potential benefit of surveillance greater than five years. While none of the patients in our study with a follow-up time greater than five years experienced recurrence, it is a possible that smaller and less metabolically active tumors may recur at time periods further out from resection, though differences in timing of recurrence based on SUVmax are not evident in our findings.

We also identified a strong association between SUVmax and OS. The association of SUVmax with OS may be related to the increased risk of recurrence in patients with high SUVmax, which consequently drives decreased survival. SUVmax is readily available preoperatively and can provide information for counseling regarding the likelihood of other high-risk features and recurrence as well as overall prognosis, allowing patients to make a more informed decision about their care and treatment regimen.

Our study has several limitations. The selection bias inherent in a retrospective study design is unable to be fully controlled for. The homogeneous group of patients with pathologic stage T1–T2, N0 NSCLC only, excluding those who underwent neoadjuvant therapy, is a strength of our study, but is unlikely to control for all potential confounding variables. The low rates of recurrence and death in this early-stage cohort limited the degrees of freedom in our multivariable analyses and inclusion of all variables that could possibly influence the outcomes of interest. Further, the retrospective nature of the study precludes the inclusion of tumor mutational data in our analysis, which may be a valuable consideration for future studies that utilize SUVmax for stratifying patient risk. We also were unable to include data regarding tumor radiologic growth patterns, which has the potential to affect recurrence and long-term outcomes.

Additionally, the limitations of SUVmax as a representation for metabolic activity should be acknowledged. SUVmax represents the FDG-avidity of a single pixel on the PET scan and does not represent the entirety of the tumor. Other studies have identified associations between more sophisticated PET parameters such as metabolically active tumor volume and total lesion glycolysis^11,22–24, and outcomes in NSCLC. We felt that SUVmax was a more clinically relevant variable due to routine reporting in PET scans and its ready availability for clinicians’ use for real-time prognostication. A further limitation that should be considered is the variability of SUVmax between different PET imaging scanners and software programs, which limits the ability to generalize a specific SUVmax cutoff value, or compare cutoff values between different institutional studies. Our study included PET scans performed at both ours and other institutions, which may make our results more generalizable. While we identified a SUVmax cutoff value for the purpose of recurrence and survival analysis, we do not propose a specific SUVmax as a global cutoff value, rather that we should consider that patients with very low avidity on PET imaging are at lower risk for recurrence based on our findings. The variability of SUVmax values between institutions does not preclude the application of this conclusion to clinical practice.

In conclusion, a high preoperative SUVmax is a strong prognostic indicator for recurrence in patients with resected early-stage node-negative NSCLC. Patients with early-stage tumors demonstrating high SUVmax should be considered high-risk for recurrence. Our data suggest that, particularly in the absence of other high-risk features, SUVmax can help further stratify patients for adjuvant therapy and a more tailored approach to post-resection surveillance. Our study can serve as the basis for future studies to prospectively validate these findings and to evaluate possible changes in surveillance practices based on SUVmax.

Abbreviations

AUC

Area-under-the-curve

CT

Computed tomography

FDG-PET

F-fluorodeoxyglucose positron emission tomography

FFR

Freedom From Recurrence

LN

Lymph node

LVI

Lymphovascular invasion

NCCN

National Comprehensive Cancer Network

NSCLC

Non-small cell lung cancer

OS

Overall Survival

ROC

Receiver Operating Characteristic

SLP

Second Lung Primary

SUVmax

Maximum Standardized Uptake Value