FDG-PET maximum standardised uptake value is associated with variation in survival: Analysis of 498 lung cancer patients

1. Introduction 

The majority of lung cancer patients have disseminated disease. Positron emission tomography with [18F]-fluoro2-deoxy-glucose (FDG-PET) is recommended to detect metastases so that futile surgery or radical radiotherapy is avoided. FDG uptake in the primary tumour is also thought to correlate with poor survival on the basis of several studies [1], [2], [3], [4], [5], [6], [7], [8]. All but one [5] of these studies were based on comparisons of survival between groups of patients with “high” or “low” SUV defined by applying cut-offs in the maximum standardised uptake value (SUV) of FDG, the cut-off values used ranging from 15 to 5. Some authors have concluded that maximum SUV is a predictor of survival independent of other factors including tumour size [2], stage [3], [6], [8] and performance status [8]. With the increasing use of adjuvant and neoadjuvant chemotherapy, PET SUV has a potential role in selecting patients for disease modifying drugs.

Many of these studies included only patients that went on to receive surgery [2], [5], [6] or curative radiotherapy [8] and were thus likely to have been dominated by patients with earlier stage disease (probably reflecting less aggressive tumours). Those studies that included a broader population of patients [1], [3], [4], [7] involved the use of cut-off values that were retrospectively chosen on inspection of the data. In this much larger study, we sought to establish, in an unbiased manner, the extent to which maximum SUV is associated with variation in survival in a broader population with lung cancer than just those patients who had curative treatment.

2. Data sources and methods 

Patients thought to be potential candidates for curative treatment were scanned in the fasting state 90min after injection of 350MBq of FDG on 951 ECAT scanners (CTI Siemens, Knoxville, TN). Segmented attenuation corrected scans were obtained of the chest using Germanium-68 rod sources for the transmission scans as part of a whole body imaging protocol. Transmission and emission scans were acquired for 5 and 10min per bed position (axial field of view=10cm), respectively. Images were reconstructed using iterative reconstruction. Resolution at full width half maximum in the reconstructed datasets was 8mm. The datasets were viewed in orthogonal planes and regions of interest drawn around the region of the tumour with maximum SUV calculated according to the formula:

A locally collated database of all FDG-PET scans performed at Guy's and St. Thomas’ Hospitals was used to record patient identifiers, the date on which the FDG-PET scan was performed, the measurement of maximum SUV taken from this scan and whether the patient was known to have had a subsequent resection. For each patient represented in this database for whom an FDG-PET scan was performed between January 2000 and December 2002, the relevant cancer registry was asked in October 2004 to provide the registered diagnosis, the date of diagnosis and the date of death (if applicable). The information provided by the cancer registry was added to the record in the FDG-PET database.

From the resulting database, we selected all records relating to patients that had a diagnosis of primary lung cancer and for whom a measurement of maximum SUV was available from an FDG-PET scan performed less than 91 days prior to or following the date of diagnosis. In cases for which more than one such record was available for a given patient, we selected the earlier measurement of maximum SUV.

The selected patients were divided into quintiles of maximum SUV. For each quintile, a Kaplan–Meier plot was produced showing cumulative survival versus the time since the FDG-PET scan was performed. To minimise the effect of the inevitable lag between deaths in the community occurring and the cancer registries being notified of such deaths [9], we chose a censoring date (December 31, 2002) well before the date of the enquiry to the cancer registries. The log rank method was used to establish the statistical significance of the association between quintile of maximum SUV and survival. The cumulative survival at 12, 18 and 24 months was calculated for each quintile and for the cohort as a whole.

3. Results 

Records for 498 patients met the selection criteria outlined above. The median age of the patients was 66 years (inter-quartile range 59–73). Of the 498 patients, 306 (61%) were male and 192 (39%) were female and 209 were known to have subsequently had surgical resection. Demographic data are presented for each quintile in maximum SUV in Table 1.

Table 1. Demographic data and the proportion of patients known to have had a resection within each quintile of maximum standardised uptake value

	Quintile	N	Male (%)	Known resection (%)	Median age (IQR)
	1st	99	49 (49)	52 (53)	65 (59–72)
	2nd	102	57 (56)	45 (44)	66 (57–73)
	3rd	97	64 (66)	38 (39)	66 (58–73)
	4th	101	68 (67)	41 (41)	66 (58–72)
	5th	99	68 (69)	33 (33)	68 (62–74)
	All	498	306 (61)	209 (42)	66 (59–73)

The cumulative survival at 12, 18 and 24 months for each quintile and for the cohort as a whole is given in Table 2. Fig. 1 shows the cumulative survival following FDG-PET scan for each quintile of maximum SUV. Maximum SUV is clearly associated with survival (p=0.0003, log rank test for trend). The number of patients included in the analysis at different time-points is given in Appendix A.

Table 2. Cumulative (Kaplan–Meier) survival for each quintile of maximum standardised uptake value (SUV)

	Quintile	N	Range of maximum SUV	Cumulative survival
				12 months (%)	18 months (%)	24 months (%)
	1st	99	0.4–6.4	79	76	61
	2nd	102	6.5–9.6	73	59	57
	3rd	97	9.7–12.8	69	56	53
	4th	101	12.9–16.6	64	55	52
	5th	99	16.8–48.1	61	43	37
	All	498	0.4–48.1	69	58	52

View Large Image Download to PowerPoint

Fig. 1. Kaplan–Meier survival curves for each quintile of maximum standardised uptake value (SUV).

4. Discussion 

The results presented in this paper establish that higher maximum SUV is associated with shorter survival. It has previously been demonstrated that uptake of FDG in lung cancer is correlated with tumour proliferation and doubling time [10] and our data suggest that the maximum SUV may be useful in gauging the biological aggressiveness of the tumour. It is our view that the measurements of maximum SUV in this case series did not influence decisions concerning treatment and so the association between maximum SUV and survival is unlikely to be self-fulfilling.

Several previous studies have each reported differences in survival between groups of patients determined by the use of single cut-off values chosen by the researchers [1], [2], [3], [4], [6], [7], [8]. In some of these studies [1], [3], [4], [6], [7], [8], the cut-off values were retrospectively chosen on inspection of the data, a fact that undermines the statistical validity of these studies [11]. One feature of our study is the grouping of patients by quintile of maximum SUV, which avoids potential bias in the choice of cut-off values, albeit creating boundary values that look unnatural.

There is variability in the measurement of maximum SUV depending on the metabolic status of the patient, the size of the tumour, timing of the scan acquisition in relation to injection of tracer and the reconstruction parameters used. Whilst all of the patients included in our study had scans performed in the same institution according to the same set of protocols, such variability would suggest that is unwise to recommend the use of arbitrary cut-offs (natural or otherwise) in clinical management.

It should be noted that one limitation of our study is that we have not established the extent to which the association between maximum SUV and survival is independent of other factors known to be associated with survival such as histology, stage, performance status and tumour size. The data presented in Table 1 suggest that, as might be expected, maximum SUV is not completely independent of factors that influenced the decision to perform curative surgery.

With further research, it may be possible to incorporate tumour biology as estimated using FDG-PET into better and more patient-specific estimates of survival and a more refined understanding of differential response to treatment. Such information would be useful in helping to identify patients suitable for the complex and burdensome combination therapies that, along with earlier diagnosis, currently seem to offer the best chance of improving outcomes among lung cancer patients. That said, a major decline in lung cancer deaths will only be achieved by smoking cessation.

5. Conclusion 

There is a clear association between high tumour uptake of [18F]-fluoro2-deoxy-glucose measured using positron emission tomography and poor survival. If future research indicates that this association is independent of other known predictors of survival, positron emission tomography may have a role in identifying those patients with most to gain from multi-modality therapy.