Professional Documents
Culture Documents
GENERAL THORACIC
Background. Integrated [18F]fluorodeoxyglucose positron 329), both SUV-T and SUV-N were predictors of survival.
emission tomography– computed tomography (PET-CT) As maximum SUV of the primary mass increased, sur-
scan is a widely used modality in the evaluation of lung vival decreased (hazard ratio, 1.05; p < 0.001). As maxi-
cancer. Our goal was to determine the ability of the stan- mum SUV of locoregional lymph nodes increased, sur-
dard uptake value (SUV) of the primary tumor (SUV-T) and vival also decreased (hazard ratio, 1.06; p < 0.001).
regional lymph nodes (SUV-N) to predict survival. Furthermore, among patients with no mediastinal dis-
Methods. From January 2005 through June 2007, 584 ease identified by PET-CT scan, increased SUV-T contin-
consecutive patients undergoing integrated PET-CT scan ued to predict poor survival (hazard ratio, 1.06; p ⴝ 0.001).
for suspected lung cancer were studied. Results of inte- Conclusions. Local and regional maximum SUVs de-
grated PET-CT scans, including the maximum SUV-T fined by integrated PET-CT scanning have a strong
and SUV-N, were recorded. A patient was defined as correlation with survival in patients with non–small cell
having a positive PET scan if the maximum SUV (T or N) lung cancer. An elevated SUV is known preoperatively
was greater than 2.5. Overall survival was documented and may assist clinicians in stratifying patients at in-
from clinical records and the Social Security Death Index. creased overall risk preoperatively.
Cox regression analysis was used to evaluate the corre-
lation between SUV and survival. (Ann Thorac Surg 2009;88:911– 6)
Results. Among patients with a positive PET scan (n ⴝ © 2009 by The Society of Thoracic Surgeons
GENERAL THORACIC
Fig 1. Positron emission tomography (PET) stage versus survival. Fig 2. Standardized uptake value of the primary mass (SUV-T) ver-
sus survival.
Mediastinum-Negative
Clinical stage I or II disease was found in190 patients
(58%) on the basis of their PET-CT scans. Among this
PET Stage Hazard Ratio p Value Confidence Interval Hazard Ratio p Value Confidence Interval
was also a significant correlation between SUV-N and value, with values above 2.5 considered positive and
survival (hazard ratio, 1.11; p ⫽ 0.01). values below 2.5 considered negative. The results of this
study argue that SUV should instead be used as a
gradient, and higher values should potentially alter over-
Comment
all treatment plan. Decisions about whether to perform
The goal of our study was to understand the ability of mediastinoscopy before resection [12], the need for adju-
PET-CT scan to predict overall outcome. Our results vant therapy [13], and the frequency of postoperative
show that PET-CT scan can in fact act as a prognosticator surveillance all may be affected by preoperative SUV
for long-term survival. There are many different aspects levels.
of PET-CT scan that were reviewed in this study. Our study has shown that survival decreases as SUV of
Overall PET stage was seen to predict survival in our the primary tumor or locoregional lymph nodes in-
study. This finding has been seen previously [4] and is in creases. An important point that remains to be discov-
part related to the poor overall outcome in patients ered, however, is the mechanism of failure in these
identified with advanced disease, especially in patients patients. One potential mechanism is that tumors with
with M1 disease [11]. higher SUV values have a more advanced stage at
Because patients with M1 disease have such guarded surgery than predicted by the preoperative PET stage,
outcomes, we performed separate analyses of the role of implying that as the SUV increases, accuracy decreases.
SUV versus survival excluding these patients. Even after Another potential mechanism is earlier local recurrence
excluding patients with M1 disease, there was still a of disease, implying that tumors with higher SUV values
significant correlation between SUV and survival, both in are more locally aggressive. Yet another possible mech-
the primary tumor and in locoregional lymph nodes. This anism is an increased propensity for distant metastasis.
correlation was also significant in the group of patients Prospective studies are required to determine the abso-
with pathologic evidence of malignant disease. Impor-
tantly, these analyses were performed adjusting for mass
size to prevent potential confounding from a variable
already known to be associated with worse survival.
These findings are important in that they can perhaps
guide treatment plan based on these values, as the SUV
levels are known preoperatively.
We also thought it was important to analyze the
correlation of SUV with survival within each clinical
stage. Although the only individual stage with a statisti-
cally significant association of SUV and survival was for
patients with PET stage II, the group of patients with no
clinical evidence of disease in the mediastinum (PET
stages I and II) and the group of patients with no clinical
evidence of metastatic disease (PET stages I, II, and III)
each had significant associations. These two groups are
composed of the patients most likely to benefit from
appropriate preoperative risk stratification and potential
surgical resection. Whereas the long-term survival in
patients identified with M1 disease is unlikely to be
affected considerably, alterations in management based
on SUV can potentially impact long-term outcome in
patients with stage I, II, or III disease identified clinically. Fig 4. Survival in mediastinum-negative patients. (SUV ⫽ stan-
Many centers now use SUV primarily as an “all or none” dardized uptake value.)
Ann Thorac Surg OKEREKE ET AL 915
2009;88:911– 6 STANDARD UPTAKE VALUE AND SURVIVAL IN NSCLC
lute causes for decreased survival in patients with higher 2. Antoch G, Stattaus J, Nemat A, et al. Non–small cell lung
SUV values. cancer: dual-modality PET/CT in preoperative staging. Ra-
diology 2003;229:526 –33.
Although we believe that SUV should be used as a
3. Lauer M, Murthy S, Blackstone E, Okereke I, Rice T.
gradient, we attempted to find a cutoff value, above and [18F]Fluorodeoxyglucose uptake by positron emission to-
below which there were significant differences in sur- mography for diagnosis of suspected lung cancer: impact of
vival. We were able to achieve this both for SUV-T and verification bias. Arch Intern Med 2007;167:161–5.
SUV-N, with values of 12.5 and 9.7, respectively. We 4. Cerfolio R, Bryant A, Ohja B, Bartolucci A. The maximum
believe that these cutoff points can be useful as a refer- standardized uptake values on positron emission tomogra-
phy of a non-small cell lung cancer predict stage, recurrence,
ence for clinicians, and may eventually be able to be and survival. J Thorac Cardiovasc Surg 2005;130:151–9.
incorporated into a staging system. Further prospective 5. Dunagan D, Chin R, McCain T, et al. Staging by positron
studies are required, however, before this goal can be emission tomography predicts survival in patients with
achieved. This cutoff would be especially practical in non-small cell lung cancer. Chest 2001;119:333–9.
6. Dhital K, Saunders C, Seed P, O’Doherty M, Dussek J.
GENERAL THORACIC
patients with no evidence of mediastinal disease preop-
[18F]Fluorodeoxyglucose positron emission tomography and
eratively. Better ability to stratify these patients would its prognostic value in lung cancer. Eur J Cardiothorac Surg
lead to more accurate prediction of long-term outcome 2000;18:425– 8.
and more appropriate treatment preoperatively. Our 7. Gupta N, Graeber G, Bishop H. Comparative efficacy of
results argue that patients with a high SUV would po- positron emission tomography with fluorodeoxyglucose in
tentially profit from a more aggressive treatment plan, evaluation of small, intermediate and large lymph node
lesions. Chest 2000;117:773– 8.
including mediastinoscopy before resection of the pri- 8. Patz E, Lowe V, Hoffman J, et al. Focal pulmonary abnormal-
mary tumor and adjuvant chemotherapy, regardless of ities: evaluation with F-18 fluorodeoxyglucose PET scanning.
final pathologic results. Radiology 1993;188:487–90.
The goals of this study were to determine the ability of 9. Mountain C. Revisions in the international system for stag-
PET-CT to predict survival in patients with suspected ing lung cancer. Chest 1997;111:1710 –7.
10. Lin D, Wei L, Ying Z. Checking the Cox model with cumu-
lung cancer, and to understand which specific aspects of
lative sums of Martingale-based residuals. Biometrika 1993;
PET-CT were important. Preoperative PET stage does 80:557–72.
predict long-term outcome, and SUV values predict sur- 11. Ginsberg M, Grewal R, Heelan R. Lung cancer. Radiol Clin
vival. Further studies are required to determine mecha- N Am 2007;45:21– 43.
nisms of failure related to higher SUV values. 12. Detterbeck F. Invasive mediastinal staging of lung cancer:
ACCP evidence-based clinical practice guidelines (2nd ed).
Chest 2007;132(Suppl):202S–20S.
References 13. Scott W, Howington J, Feigenberg S, Movsas B, Pisters K.
Treatment of non-small cell lung cancer stage I and stage II:
1. Cancer Statistics 2007. American Cancer Society. Available ACCP evidence-based clinical practice guidelines (2nd edi-
at: www.cancer.org. Accessed June 18, 2009. tion). Chest 2007;132(Suppl):234S– 42S.
DISCUSSION
DR DAVID R. JONES (Charlottesville, VA): I had two quick DR MICHAEL J. LIPTAY (Chicago, IL): I enjoyed the paper very
questions for you. How many patients had bronchioloalveolar much. I have a question along similar lines when the tumor SUV
cell carcinoma given that they typically have a low SUV (stan- is high and the hilar and mediastinal nodes are negative. Do you
dardized uptake value) value? What do you think if you ex- change your strategy at all?
cluded those, would it have any impact on your analysis? Secondly, did you find in the pathologic analysis of the
Second, what do we do now with the patient who is medias- surgical specimens that the nodes had a higher incidence of
tinoscopy negative, but who has a tumor with an SUV of 15 and being positive; that is, high SUV in the tumor may predict nodal
a node of 12, which would fall into your high-risk group? Should positivity because it’s a more aggressive tumor?
we give those patients induction chemo? Is your group starting
to look at that or considering it? DR OKEREKE: Yes, that’s a great question. In fact, that’s what
we investigated. It did not specifically say, and there was no
DR OKEREKE: I think those are both great questions. I think the statistical significance in the difference of concordance between
second one goes toward the heart of the relevance of this study. the two in that situation. But I think that’s what potentially
To answer the first question, approximately 10% to 15% of our further analysis would need to consider.
patients had bronchoalveolar pathology. As such, I would say
that bronchoalveolar probably does not have a significant im- DR JOSHUA R. SONETT (New York, NY): Excellent presenta-
pact toward the overall results of this study. This is my tion. Did you look at all at the size of the tumor relative to the
approximation. SUV and quantitate SUV for the size?
As to the second question, I think that once again the rele- So does a 1-cm tumor that throws off an SUV of 10 versus a
vance of this study is, especially in mediastinal-negative patients 10-cm tumor that throws off an SUV of 10 have different clinical
with a very high SUV, should we do neoadjuvant strategies? I outcomes, as per gram of tumor the smaller tumor appears more
think based on this, the next study should be a prospective active?
randomized trial looking at SUV and potentially separating
patients into two groups, one with neoadjuvant and one without. DR OKEREKE: Yes, we did. Yes, I didn’t include that. The size
I think that will be the next step. of the tumor was correlated with survival, which is not too
916 OKEREKE ET AL Ann Thorac Surg
STANDARD UPTAKE VALUE AND SURVIVAL IN NSCLC 2009;88:911– 6
unexpected. Secondly, these results were independent of mass imaging. And so for patients who were N1 positive by PET, what
size. was the concordance in pathologic specimens?
DR LIPTAY: No. But did you ever think of dividing by the size DR OKEREKE: Right. N1 disease was very limited preopera-
of the tumor, so an SUV, like a Dlco (lung diffusing capacity for tively, but I think it’s a good point.
carbon monoxide), for the size of that tumor?
DR JONES: Did you do a multivariate analysis of SUVmax?
DR OKEREKE: Oh, I see. No, we did not. You’re saying it’s a marker of a poor prognosis if it’s above or
below a certain number. But is it an independent predictor of a
DR LIPTAY: So a 1-cm tumor that throws off an SUV of 5 may poor prognosis?
be worse than a 5-cm tumor that throws off an SUV of 12. It may Did you look at tumor size, pathologic stage, or other kinds of
bespeak early to the bad biologic character of it, but it didn’t get histopathologic variables such as grade or lymphovascular
big enough to throw off an SUV of 12. invasion?
GENERAL THORACIC
DR OKEREKE: No, I think that’s a good point. DR OKEREKE: To answer your question, yes and no. We did it
based on mass size as well.
DR LIPTAY: But you have that data, so it might be nice for us to
see it in the future.
DR JONES: And was it a multivariate analysis?
DR PAUL DE LEYN (Leuven, Belgium): Thank you for your very
DR OKEREKE: Yes.
nice presentation. As you know, SUV is not at all standardized.
It might be standardized in one institution for one PET (positron
emission tomography) scanner, but you cannot transmit the data DR LEDFORD POWELL (Newport Beach, CA): A quick ques-
for other PET scanners. tion. I enjoyed the presentation. You actually answered a lot of
Can we use the data from your center to select patients for questions that I had had about PET scans previously.
adjuvant therapy in centers with other PET scanners? I did want to know whether or not you noticed a correlation
between the mediastinal-negative patients that had high pri-
DR OKEREKE: Yes. Though it is called Standard Update Value, mary tumor SUVs and their tumor biology, meaning if they had
it’s quite nonstandard across different PET scanners. a low SUV, did they have moderately differentiated adenocarci-
We eliminated that problem at our institution because we had noma? And if they had a higher SUV, did they have a more
a single PET-CT scanner which was used for all 584 patients. But aggressive tumor?
I think trying to apply a cutoff value among different institutions And should we be considering the aggressiveness of the
may have some degree of difference. tumor when we’re looking at whether or not these patients need
That being said, there has been some work by Dr Cerfolio. to get chemotherapy?
That’s one of the papers that we’ve referenced. His cutoff value
by a different methodology was about 10. We had a cutoff value DR OKEREKE: Yes, I think potentially, and, no, we did not
for the primary tumor of 12.5. notice that.
I think that there will be some gray area, but it still should not
eliminate the possibility of trying to create some type of ran- DR JEAN-FRANCOIS LEVI (Neuilly-sur-Seine, France): Did
domized trial looking at neoadjuvant strategies based on SUV. I you notice if the decrease of SUV after chemotherapy is a better
think that’s a great point, though. prognostic or not?
DR SETH D. FORCE (Atlanta, GA): I have a quick question. I DR OKEREKE: No, that wasn’t included in our study. No, we
think that one of the previous questions may have addressed did not include patients who underwent neoadjuvant chemora-
this, but I didn’t see kind of a report of the accuracy of your PET diation in this study.