Upload
others
View
18
Download
0
Embed Size (px)
Citation preview
1
18F‐FDG‐PET/CT for very early response evaluation predicts CT response in erlotinib treated NSCLC patients – A comparison of assessment methods
Authors:
Joan Fledelius1, Anne Winther‐Larsen2, Azza A. Khalil 3, Catharina M. Bylov4, Karin Hjorthaug 5, Aksel Bertelsen1, Jørgen Frøkiær5 and Peter Meldgaard3
1 Department of Nuclear Medicine, Herning Regional Hospital, Herning, Denmark 2 Department of Clinical Biochemistry, Aarhus University Hospital, Aarhus, Denmark.
3 Department of Oncology, Aarhus University Hospital, Aarhus, Denmark. 4 Department of Radiology, Aarhus University Hospital, Aarhus, Denmark.
5 Department of Nuclear Medicine and PET Centre, Aarhus University Hospital, Aarhus, Denmark.
No disclaimers No financial support Corresponding author: Joan Fledelius, MD Department of Nuclear Medicine, Herning Regional Hospital, DK‐7400 Herning, Denmark. E‐mail: [email protected] Fax: +45 78437361 Telephone: +45 31603316
Word count: 4998
Running title: Early FDG‐PET predicts CT response
Journal of Nuclear Medicine, published on May 10, 2017 as doi:10.2967/jnumed.117.193003by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
2
ABSTRACT
The purpose of this study was to determine which method for early response
evaluation with 18F‐fluoro‐D‐glucose positron emission tomography combined
with whole body computed tomography (18F‐FDG‐PET/CT) performed most
optimal for prediction of response on a later CT‐scan in erlotinib treated non‐
small cell lung cancer (NSCLC) patients.
Methods: 18F‐FDG‐PET/CT scans were performed prior to and after 7‐10 days
of erlotinib treatment in 50 NSCLC patients. The scans were evaluated using a
qualitative approach and various semi‐quantitative methods including
percentage change in Standardized Uptake values, lean body mass corrected
(SUL) SULpeak, SULmax and Total Lesion Glycolysis (TLG).
The PET parameters and their corresponding response categories were
compared to the percentage change in the sum of the longest diameter (SLD)
in target lesions and the resulting response categories from a CT scan
performed after 9‐11 weeks of erlotinib treatment using Receiver Operating
Characteristics (ROC) analysis, linear regression and quadratic weighted kappa.
Results: TLG delineation according to the PET evaluation response criteria in
solid tumors (PERCIST) showed the strongest correlation to SLD (R = 0.564,
p<0.001), compared to SULmax (R=0.298, p=0.039) and SULpeak (R=0.402,
p=0.005). For predicting progression on CT, ROC analysis show area under the
curves (AUCs) between 0.79 and 0.92 with the highest AUC of 0.92 (95%
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
3
confidence interval (CI): 0.84‐1.00) found for TLG (PERCIST). Furthermore,
using a cut‐off of 25% change in TLG (PERCIST) for both partial metabolic
response (PMR) and progressive metabolic disease (PMD), which is the best
predictor of the CT response categories showed a kappa value of 0.53 (95%CI:
0.31‐0.75). This method identifies 41% of the later PDs on CT, with no false
positives. Visual evaluation correctly identified 50% with a kappa value of 0.47
(95%CI: 0.24‐0.70). Conclusion TLG (PERCIST) was the optimal predictor of
response on later CT scans outperforming both SULpeak and SULmax. Using
TLG (PERCIST) with a 25% cut‐off after 1‐2 weeks of treatment allows us to
safely identify 41% of the patients who will not benefit from erlotinib and stop
the treatment at this time.
Key words: FDG‐PET/CT, Lung cancer, PERCIST 1.0, Early response evaluation,
TLG.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
4
INTRODUCTION
Treatment with tyrosine kinase inhibitors (TKIs) has proven to be effective in
NSCLC patients with response rates of 10‐20% in unselected populations.
Subgroups of patients have been identified with good and sometimes
prolonged results (1‐4).
At present, selection of patients is done by detecting mutations in the
epidermal growth factor receptor (EGFR) genes. EGFR mutation positive
patients (EGFR‐mut) have higher response rates than EGFR wild type (EGFR‐
wt) patients (5‐9). However, a subgroup of EGFR‐wt patients (1,5,10) also
benefit from TKI treatment, suggesting that selecting patients depending
solely on mutation‐status will exclude some patients from a potential
treatment benefit.
At our institution, erlotinib treatment is offered to non‐operable EGFR‐
mut, NSCLC patients as first‐line treatment and as second‐ or third‐line
treatment to EGFR‐wt patients. An important challenge with this treatment is
how to evaluate the response, since TKIs are known to have mostly cytostatic
effects (11,12) as opposed to cytotoxic effect. Therefore, a follow‐up CT scan is
routinely performed 9‐11 weeks into the treatment for response evaluation.
Considering that stage III – IV NSCLC patients in general have a short remaining
life expectancy (4,13,14), it is essential to discontinue a futile treatment course
as early as possible.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
5
Preclinical studies have shown that TKI sensitive cells downregulate
their glucose uptake very early after exposure to TKI treatment (1,11).
Clinically 18F‐FDG‐PET/CT performed very early during TKI treatment has
shown promise for predicting both anatomical response and survival
(1,5,15,16), and to predict the histopathologic response (7,17).
The many parameters used for response evaluation with 18F‐FDG‐PET
include visual evaluation, change in standardized uptake value corrected for
body weight (SUV) both for maximum intensity voxels and the mean value in a
standardized volume of 1.2 cm in diameter VOI (SUVmax and SUVpeak), as
well as more complex volume‐based parameters such as TLG with various
ways of delineating the lesions.
PERCIST 1.0 from 2009 (18) uses the SULpeak as the standard, but also
suggest TLG as a supplementary analysis. We have previously shown that
percentage change in TLG is a promising predictor of CT response in a
subgroup of EGFR‐wt patients (10).
The aim of this study is therefore to identify which specific method of a
selection of commonly used methods (including the PERCIST 1.0 methods) is
the optimal for predicting the CT response, early during erlotinib treatment.
We hypothesize that an evaluation of the total disease burden will improve
response evaluation compared to single “hottest” lesion evaluation. We focus
on safely selecting patients after 7‐10 days of treatment, who will not have a
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
6
treatment effect, enabling us to discontinue futile treatment at this early time‐
point.
MATERIALS AND METHODS
The results are reported as according to “Standards for Reporting
Diagnostic Accuracy studies 2015”.
Patients
This retrospective study evaluated 18F‐FDG‐PET/CT scans, in
compliance with the PERCIST 1.0 criteria to ensure comparability of images,
from 50 patients enrolled in prospective single‐center study originally
including 67 consecutive patients with stage III‐IV NSCLC between April 2013
and August 2015. The study was approved by the Central Denmark Region
Committee on Biomedical Research Ethics (no. 1‐10‐72‐19‐12) and subjects
signed an informed consent form. The study was reported to ClinicalTrials.gov
(NCT02043002). All patients were candidates for palliative erlotinib treatment.
Detailed inclusion criteria were previously published (19). The selection of
patients included in the present study is presented in Figure 1.
All patients had an 18F‐FDG‐PET/CT scan performed before (baseline)
and 7‐10 days after initiation of erlotinib treatment (follow‐up). CT scans of
the chest and abdomen were acquired before and after 9‐11 weeks of
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
7
treatment, or earlier on clinical indication. Testing for epidermal growth factor
receptor (EGFR) mutations had been performed in all patients with
adenocarcinoma as part of the routine diagnostic work‐up by use of the
Therascreen EGFR RGQ PCR kit (QIAGEN, Manchester, UK) according to the
manufacturer’s protocol, and based on this, patients were categorized as
either EGFR‐wt or EGFR‐mut.
18F‐FDG‐PET/CT acquisition and evaluation
All 18F‐FDG‐PET/CT scans were performed on a combined PET/CT
scanner (Siemens Biograph TruePoint 40, Siemens Healthcare GMbH,
Erlangen, Germany) at the Department of Nuclear Medicine and PET‐Centre,
Aarhus University Hospital, Denmark, using the same scanner type,
acquisition‐ and reconstruction protocol as previously published in detail (19).
In brief, all patients had a fasting period of at least 6 hours, a blood glucose
concentration of < 11 mM and an uptake time of 60 +/‐ 10 minutes between
injection of 5 MBq pr. kg ±10% of FDG and scan start (3 minutes per bed
position). A whole‐body low‐dose CT scan (50 mAS, 120kVp) was performed.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
8
All 18F‐FDG‐PET/CT scans were evaluated by one experienced Nuclear
Medicine Physician using Siemens Syngovia software (Siemens Healthcare
GMbH, Erlangen, Germany). The evaluator was blinded to the outcome and
the result of the following CT scan. Evaluation of response was performed as
1) Visual evaluation described by Mac Manus et al (20) considering both the
overall change in FDG‐uptake and appearance of new FDG‐avid lesions (visual),
2) The percentage change in SULpeak in the lesion with the highest uptake at
baseline and follow‐up (not necessarily the same lesion) (SULpeak) according
to PERCIST 1.0 (18), and 3) The percentage change in global TLG with various
delineation methods: At SULmean + 2 standard deviations (SD) in a spherical 3
cm ROI in the right lobe of the liver (SULmean(liver)) (TLG (PERCIST)) and at
30% (TLG (30)), 40% (TLG (40)) and 50% of SULmax (TLG (50)). A lesion was
considered evaluable for all semi‐quantitative methods if SULpeak was 1.5 x
SULmean(liver) + 2SD. Delineation was performed semi‐automatically after a
manual rough outlining of each lesion, resulting in an SULmean and a
metabolic tumor volume for the delineated area. TLG for each lesion was
calculated as SULmean x metabolic tumor volume. Global TLG was the sum of
all measurable lesions TLGs.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
9
All the methods allocate patients into 4 different response categories:
1) PMD, 2) stable metabolic disease (SMD), 3) PMR and 4) complete metabolic
response. When using the classification methods, for SULpeak, all patients
were categorized as PMD, if new lesions had appeared, independent of
SULpeak. This was not the case for the TLG categories, since any new lesion
was included in the TLG calculations, if measurable. Multiple cut‐offs for these
response categories were tested. An overview of the methods is presented in
Supplementary Table 1. SULmax was measured for comparison, but not used
for classification into response categories.
CT evaluation
The radiological response was evaluated by one experienced
radiologist on the first CT scan performed after initiation of erlotinib according
to Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 criteria.
Response evaluation included reporting of percentage change in SLD and the
resulting response categories: 1) progressive disease (PD), 2) stable disease
(SD), 3) partial response (PR) and 4) complete response. We chose to
dichotomize into PD vs non‐PD (SD+PR) when appropriate, since this is used as
the basis criteria to decide whether to continue or discontinue erlotinib
treatment.
Statistical analysis
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
10
Comparison of SULpeak, SULmax and the TLGs to the SLD was
performed using linear regression analysis, a significance level of 0.008
(Bonferroni correction for 6 methods) was used.
The predictive accuracy of SULpeak, SUL max and the TLGs was
evaluated by ROC analyses, predicting PD versus non‐PD on CT. The optimal
cut‐off, considering sensitivity and specificity equally important, was identified
by visually locating the data point nearest the top left corner on the ROC
curve.
To compare the 18F‐FDG‐PET response categories to the CT response
categories quadratic weighted kappa analyses were performed (21). As many
patients as practically possible was included, a power calculation was not
performed.
Statistical analyses were performed with SPSS statistics version 23.0 for
Macintosh (IBM SPSS Statistics, Chicago, Il).
RESULTS
Scan times and standardization parameters
The time between scans and treatment start is presented in Table 1.
The indication for an early CT scan was in all cases suspected clinical
progression. 10 showed PD and 4 SD, the four scanned earlier than 4 weeks all
showed PD on CT and PMD on 18F‐FDG‐PET/CT (TLG(PERCIST(25%)). Data on
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
11
injected FDG‐activity, glucose level and uptake time on population basis is
presented in Table 2. A full description of glucose, uptake time and injected
FDG activity is available in Supplementary Table 2.
All patients were analyzable by visual evaluation and SULpeak (and
SULmax), whereas the TLG measurements were not feasible to perform in 3‐7
patients depending on the delineation method. When using the PERCIST
delineation method three patients were not analyzable; in one patient owing
to an uptake at follow‐up close to the background level, another patient had
myriads of very small FDG avid lesions, and the last patient showed intense,
diffuse uptake in the lung tissue impairing delineation of the tumor. TLG (50,
40 and 30) showed in total a further four non‐evaluable patients owing to
relatively low uptake or proximity to the liver (higher background level). For
two patients SLD was not available (non‐measurable lesions). None of the
patients were classified as complete responders on either CT or 18F‐FDG‐
PET/CT.
Comparing of SULpeak, SULmax, and TLGs to SLD
We found that all the measured parameters except SULmax were
significantly correlated to SLD, but TLG (PERCIST) showed the strongest
correlation with R = 0.564 (p < 0.001). Plots for SULpeak, TLG (PERCIST) and
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
12
SULmax are presented in Figure 2, all variables are available in Supplementary
Figure 1.
Dichotomizing the CT response into progression (PD) and non‐PD
resulted in 28 PD and 22 non‐PD patients. For the PET‐methods involving a
continuous variable, the ROC curves for the 43 patients analyzed by all
methods are presented in Figure 3. As can be seen, TLG (PERCIST) has the
highest AUC 0.923 (95% CI:0.840‐1.00) (Table 3), although the confidence
intervals overlap between methods. TLG (30) had the lowest AUC of 0.790
(95%CI: 0.698‐0.949).
The highest sensitivity and specificity for predicting PD on CT when
considering the sensitivity and the specificity equally important (sensitivity of
0.89 and specificity of 0.88) was seen for TLG (PERCIST). For specificity = 1.00,
the sensitivity for TLG (PERCIST) was 0.50, while SULpeak had a slightly higher
sensitivity of 0.58. Again, TLG (30) showed the lowest values. Interestingly,
SULpeak identified more anatomical responders than TLG (PERCIST), but fewer
anatomical progressions, as shown in Figure 4. Plots for SULmax and TLG (30‐
50) are found in Supplementary Figure 2.
Metabolic response categories compared to RECIST
TLG (PERCIST (25%)) resulted in correct classification in 25 of 47
patients (53%) according to the CT response. One patient was classified more
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
13
than one level different (early PMR and later PD on CT). Noticeably, all of the
PMDs identified were classified as PD on later CT scans, identifying 11 of 27
PDs (41%). Even though SULpeak provided the highest sensitivity at specificity
= 1 in the ROC analysis and all patients were analyzable by this method, it only
identified 23 of 50 patients correctly (46%) when the response categories for
the optimal cut‐off (20%) was used. The visual method identified 25 of 50
patients correctly (50%) and both SULpeak and visual found one (the same
patient) more than one level different, possibly owing to bone‐flare. The
results for TLG (PERCIST (25%)), SULpeak (20%) and visual evaluation are
presented in Table 4.
In general, the kappa values were rather low ranging from 0.23 to 0.53,
presented in Table 5. The method with the highest value 0.53 (95%CI: 0.31‐
0.75) was TLG (PERCIST(25%)).
Dichotomizing the 18F‐FDG‐PET classification into PMD and non‐PMD
(PMR+SMD) for comparison to the similar dichotomization for CT (PD versus
non‐PD) resulted for TLG (PERCIST (25%)) in a sensitivity of 0.41, a specificity of
1.00, a positive predictive value of 1.00 and a negative predictive value of 0.56.
For the SULpeak (20%), a slightly lower sensitivity of 0.29, a specificity of 1.00,
a positive predictive value of 1.00 and an negative predictive value of 0.52 was
found. The categorization tables, sensitivity, specificity and predictive values
for all methods are presented in Supplementary Table 3.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
14
DISCUSSION
This study was performed to identify which PET assessment method
was the most optimal for predicting the later CT response early during
erlotinib treatment in unselected NSCLC patients. Early assessment of
progression would enable discontinuation of non‐beneficial treatment after a
few days of treatment (7‐10 days).
The main result of this study was that TLG (PERCIST) had the strongest
correlation to SLD, statistically significant even when applying a significance
level of 0.008. It was the best predictor of PD versus non‐PD on CT, though the
difference between methods were not statistically significant. A 25% change
was optimal cut‐off and avoided false PMDs. It performed better than the
PERCIST suggested 75% increase and 45% decrease, possibly because of the
early time point for evaluation.
There are many assessment methods for evaluating the treatment
response with 18F‐FDG‐PET/CT. The most commonly used parameter is
SUVmax, but SULpeak is also frequently used, especially since this was
recommended as the standard method in the PERCIST 1.0 from 2009(18), and
is also recommended for response evaluation in the latest European
guidelines, together with the mentioning of the increasing interest of TLG(22).
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
15
However, few studies have compared the performance of these methods
simultaneously.
Previously, a comparison of six different parameters including
SUVpeak, SUVmax and SUVmeans using different delineation methods was
performed for residual activity after 1 week of erlotinib treatment (23). In this
study other SUVs than SUVmax were not superior at predicting progression on
CT but they did not consider the percentage change, moreover, they did not
include TLG.
In another study including 34 erlotinib treated patients it was
demonstrated that percentage change in SUVpeak after 1 week of treatment
was predictive of PD vs non‐PD on CT after 6 weeks of treatment (5).
However, there was no comparison to changes in TLG or other parameters.
Moon et al compared SUVmax, SUVpeak and TLG in 52 stage IV NSCLC
patients before and after 4 cycles of platinum‐based chemotherapy (24),
consistent with the results of the present study, they found that change in TLG
was a predictor of PFS while change in SUVpeak and SUVmax were not, thus
supporting our finding, that TLG outperforms SULpeak and SULmax.
We found that a 25% change in TLG separated the response categories
better than the 45/75% cut‐off from PERCIST 1.0. This is consistent with
previous findings by Kahraman demonstrating that a cut‐off between 20 and
30% change is superior for predicting PFS (25). Our results suggest that the
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
16
same is true for predicting PD/non‐PD on CT. When using 25% change in TLG
(PERCIST), we still identified a large group of SMD patients who later showed
PD on CT. In order to reduce the size of this category, we could use the ROC
determined cut‐off value of 9.2 % increase (or 10%), which reduced this group
from 15 to 12. However, since the cut‐off values are determined from the data
one must expect the estimated sensitivity and specificity to be too optimistic,
and the day‐to‐day variation (26‐28) should also be considered. Taking this
into account, we prefer to use the 25% cut‐off.
The visual evaluation method performed well and was comparable to
the more sensitive of the semi‐quantitative methods, SULpeak and TLG
(PERCIST). We have previously demonstrated in a chemotherapy treated
population of locally advanced NSCLC patients, that there is a strong inter‐
observer agreement for this method, but the agreement is stronger for
SULpeak change, however, TLG was not included in that study (29). This
suggests that visual evaluation is a reliable alternative method for evaluation
of the cases that are “non‐evaluable” with the TLG methods.
The strengths of the present study are the strict adherence to
standardization as according to PERCIST, and the head‐on comparison of 6
different methods for evaluating response, including the PERCIST 1.0
recommendations and the analysis on categorization using various cut‐off
levels, not only dichotomizing by ROC optimization.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
17
The main limitation is the large variation in the interval between
treatment and CT scans both at baseline and at follow‐up. We find the
variation in the time intervals for 18F‐FDG‐PET/CT scans acceptable, and
comparable to other studies (5). The very early progressions on CT within 4
weeks after initiating treatment were considered true progressions and as
such should not significantly influence the PET results. The five cases of
prolonged interval (> 4 weeks) between the baseline CT and treatment would
potentially attenuate the CT response and thereby increase the risk for PET to
show a better response than reflected on CT, but in four of the five cases TLG
(PERCIST(25%) agreed with the CT response. For comparison of methods we
consider it usable since it applies to all methods. However, it needs closer
attention in future, since standardization is particularly important in the early
response evaluation setting.
CONCLUSION
The present study demonstrates that using percentage change in
global TLG delineated according to PERCIST tends to be a more sensitive
method for early response evaluation of NSCLC patients during erlotinib
treatment than highest intensity lesion evaluation (SULpeak) and visual
evaluation. The method allows for prediction of a later PD on CT identifying
41% of the PDs. We intend to use this finding in future studies and in the
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
18
clinical setting, supplementing with visual evaluation in the few cases not
evaluable by the TLG method.
Disclosures
We have nothing to disclose.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
19
REFERENCES 1. Tiseo M, Ippolito M, Scarlattei M, et al. Predictive and prognostic value
of early response assessment using 18FDG‐PET in advanced non‐small cell lung cancer patients treated with erlotinib. Cancer Chemoth Pharm. 2014;73:299‐307.
2. Sharma S V, Bell DW, Settleman J, Haber DA. Epidermal growth factor receptor mutations in lung cancer. Nat Rev Cancer. 2007;7:169‐181.
3. Pérez‐Soler R, Chachoua A, Hammond LA, et al. Determinants of tumor response and survival with erlotinib in patients with non‐small‐cell lung cancer. J Clin Oncol. 2004;22(16):3238‐3247.
4. Shepherd FA, Pereira JR, Ciuleanu T, et al. Erlotinib in previously treated non–small‐cell lung cancer. New Engl J Med. 2005;353:123‐132.
5. Zander T, Scheffler M, Nogova L, et al. Early prediction of nonprogression in advanced non‐small‐cell lung cancer treated with erlotinib by using [18F]fluorodeoxyglucose and [18F]fluorothymidine positron emission tomography. J Clin Oncol. 2011;29:1701‐1708.
6. Takahashi R, Hirata H, Tachibana I, et al. Early [18F]fluorodeoxyglucose positron emission tomography at two days of gefitinib treatment predicts clinical outcome in patients with adenocarcinoma of the lung. Clin Cancer Res. 2012;18:220‐228.
7. Aukema TS, Kappers I, Olmos R a V, et al. Is 18F‐FDG PET/CT useful for the early prediction of histopathologic response to neoadjuvant erlotinib in patients with non‐small cell lung cancer? J Nucl Med. 2010;51:1344‐1348.
8. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth factor receptor underlying responsiveness of non–small‐cell lung cancer to gefitinib. New Engl J Med. 2004;350:2129‐2139.
9. Tsao MS, Sakurada A, Cutz JC, et al. Erlotinib in lung cancer ‐ molecular and clinical predictors of outcome. New Engl J Med. 2005;353:133‐144.
10. Winther‐Larsen A, Fledelius J, Demuth C, Bylov CM, Meldgaard P, Sorensen BS. Early change in FDG‐PET signal and plasma cell‐free DNA level predicts erlotinib response in EGFR wild‐type NSCLC patients. Transl Oncol. 2016;9:505‐511.
11. Su H, Bodenstein C, Dumont RA, et al. Monitoring tumor glucose utilization by positron emission tomography for the prediction of treatment response to epidermal growth factor receptor kinase inhibitors. Clin Cancer Res. 2006;12:5659‐5667.
12. Tuma RS. Sometimes size doesn’t matter: reevaluating RECIST and tumor response rate endpoints. J Natl Cancer I. 2006;98:1272‐1274.
13. Fukuoka M, Yano S, Giaccone G, et al. Multi‐institutional randomized
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
20
phase II trial of gefitinib for previously treated patients with advanced non‐small‐cell lung cancer. J Clin Oncol. 2003;21:2237‐2246.
14. Thatcher N, Chang A, Parikh P, et al. Gefitinib plus best supportive care in previously treated patients with refractory advanced non‐small‐cell lung cancer: results from a randomised, placebo‐controlled, multicentre study (Iressa survival evaluation in lung cancer). Lancet. 2005;366:1527‐1537.
15. Sunaga N, Oriuchi N, Kaira K, et al. Usefulness of FDG‐PET for early prediction of the response to gefitinib in non‐small cell lung cancer. Lung Cancer. 2008;59:203‐210.
16. Hachemi M, Couturier O, Vervueren L, et al. [18F]FDG positron emission tomography within two weeks of starting erlotinib therapy can predict response in non‐small cell lung cancer patients. PloS one. 2014;9:1‐10.
17. van Gool MH, Aukema TS, Schaake EE, et al. Timing of metabolic response monitoring during erlotinib treatment in non‐small cell lung cancer. J Nucl Med. 2014;55:1081‐1086.
18. Wahl RL, Jacene H, Kasamon Y, Lodge MA. From RECIST to PERCIST: Evolving considerations for PET response criteria in solid tumors. J Nucl Med. 2009;50(Suppl 1):122‐150.
19. Winther‐Larsen A, Fledelius J, Sorensen BS, Meldgaard P. Metabolic tumor burden as marker of outcome in advanced EGFR wild‐type NSCLC patients treated with erlotinib. Lung Cancer. 2016;94:81‐87.
20. Mac Manus MP, Hicks RJ, Matthews JP, et al. Positron emission tomography is superior to computed tomography scanning for response‐assessment after radical radiotherapy or chemoradiotherapy in patients with non‐small‐cell lung cancer. J Clin Oncol. 2003;21:1285‐1292.
21. Fleiss JL, Levin B, Paik MC. Statistical methods for rates and proportions. 3rd edition. John Wiley & Sons; 2003.
22. Boellaard R, Delgado‐Bolton R, Oyen WJG, et al. FDG PET/CT: EANM procedure guidelines for tumour imaging: version 2.0. Eur J Nucl Med Mol imaging. 2014;42:328‐354.
23. Kobe C, Scheffler M, Holstein A, et al. Predictive value of early and late residual 18F‐fluorodeoxyglucose and 18F‐fluorothymidine uptake using different SUV measurements in patients with non‐small‐cell lung cancer treated with erlotinib. Eur J Nucl Med Mol imaging. 2012;39:1117‐1127.
24. Moon SH, Cho S‐H, Park LC, et al. Metabolic response evaluated by 18F‐FDG PET/CT as a potential screening tool in identifying a subgroup of patients with advanced non‐small cell lung cancer for immediate maintenance therapy after first‐line chemotherapy. Eur J Nucl Med Mol imaging.2013;40:1005‐1013.
25. Kahraman D, Holstein A, Scheffler M, et al. Tumor lesion glycolysis and
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
21
tumor lesion proliferation for response prediction and prognostic differentiation in patients with advanced non‐small cell lung cancer treated with erlotinib. Clin Nucl Med. 2012;37:1058‐1064.
26. Nakamoto Y, Zasadny KR, Minn H, Wahl RL. Reproducibility of common semi‐quantitative parameters for evaluating lung cancer glucose metabolism with positron emission tomography using 2‐deoxy‐2‐[18F]fluoro‐D‐glucose. Mol Imaging and Biol. 2002;4:171‐178.
27. Weber WA, Gatsonis CA, Mozley PD, et al. Repeatability of 18F‐FDG PET/CT in advanced non‐small cell lung cancer: Prospective assessment in 2 multicenter trials. J Nucl Med. 2015;56:1137‐1143.
28. Kramer GM, Frings V, Hoetjes N, et al. Repeatability of quantitative whole body 18F‐FDG PET/CT uptake measures as function of uptake interval and lesion selection in non‐small cell lung cancer patients. J Nucl Med. 2016;57:1343‐1349.
29. Fledelius J, Khalil A, Hjorthaug K, Frøkiær J. Inter‐observer agreement improves with PERCIST 1.0 as opposed to qualitative evaluation in non‐small cell lung cancer patients evaluated with F‐18‐FDG PET/CT early in the course of chemo‐radiotherapy. EJNMMI Research. 2016;6:71.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
1
Figures and figure legends
FIGURE 1. Patient selection
FIGURE 2. Correlation between A) SULpeak, B) TLG (PERCIST) and C) SULmax at 7-10 days and the SLD measured on CT scans performed after 9-11 weeks.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
2
FIGURE 3. ROC curves for SULpeak, SULmax and the various TLG variations. The curves illustrate the parameters’ ability to predict PD on the later CT scan performed after 9‐11 weeks of treatment. The curves represent data from the 43 patients who were analyzable by all methods.
FIGURE 4. Waterfall plots for A) SULpeak, B) TLG (PERCIST) and C) SULmax showing the distribution of CT categories (light grey=PD, red=SD and black=PR). The horizontal reference lines represent the optimal cut‐off for PMR and PMD. For SULmax they represent the 15% change suggested by the EORTC criteria for early evaluation.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
3
Tables
TABLE 1. Time (days) between scans and first day of treatment
Baseline (median (range))
> 4 weeks (n (%))
> 2 weeks (n (%))
Follow‐up (median (range))
Early CT (n (%))
CT 15 (4‐56) 5 (10%) 26 (52%) 77 (20‐85) 14 (28%) FDG‐PET/CT 1 (0‐21) 8 (2‐23)
> 4 (and 2) weeks = number of patients with an interval between baseline CT and first day of treatment above 4 (and 2) weeks. Early CT = number of patients with a follow‐up CT earlier than 9 weeks.
TABLE 2. Compliance with the PERCIST 1.0 standardization criteria
Parameter Baseline Follow‐up Numerical difference
PERCIST 1.0 Adherence to PERCIST 1.0
Injected activity (Mbq) Mean (std) 340 (90) 335 (87) 18 (14) Baseline +/‐ 20% 100% (50/50) Range 197‐609 199‐618 0‐54 Glucose level (mmol/liter)
Mean (std) 6.3 (0.9) 6.4 (0.9) 0.4 (0.5) < 11mM 100% (50/50) Range 4.6‐8.8 4.7‐9.0 0.0‐1.6 Uptake time (minutes) Mean (std) 58.7 (4.0) 59.3 (4.9) 5.1 (4.19) 60 +/‐ 10 min 98% (98/100*) Range 51‐69 48‐72 0‐15 Baseline +/‐ 15 min 100% (50/50)
* The uptake time at both baseline and follow up. Two patients with 48 and 72 minutes uptake time at follow‐up was included because the difference between scans in both cases was within +/‐ 15 minutes. Std is the standard deviation of the means.
TABLE 3. AUCs and sensitivity and specificity from the ROC analyses
95% CI Sens. and spec. equally important
Spec. as high as possible
Method AUC Lower Upper Sens. Spec. Cut-off Sens. Spec. Cut-off
TLG (PERCIST)
0.923 0.842 1.000 0.89 0.88 -6.6 % 0.50 1.00 9.2 %
SULpeak 0.887 0.791 0.983 0.77 0.82 -6.9 % 0.58 1.00 2.1 % SULmax 0.835 0.715 0.954 0.73 0.82 -4.5 % 0.35 1.00 13.2 % TLG 50 0.824 0.698 0.949 0.77 0.82 3.0 % 0.39 1.00 30.0 % TLG 40 0.821 0.696 0.946 0.77 0.71 -1.4 % 0.35 1.00 24.6 % TLG 30 0.790 0.645 0.934 0.73 0.76 0.0 % 0.58 0.94 11.9 %
AUCs optimized by identifying the point on the curve closest to the upper left corner (considering sensitivity and specificity equally important (middle columns)), and considering the specificity of the major importance thereby avoiding any false progressions (right columns). “Cut-off” is the corresponding optimal cut-off value for percentage change
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
4
TABLE 4. TLG (PERCIST (25%)), SULpeak (20%) and visual response compared to RECIST 1.1 CT TLG (PERCIST (25%)) SULpeak (20%) Visual PMR SMD PMD Total PMR SMD PMD Total PMR SMD PMD Total
PR 4 0 0 4 3 2 0 5 4 1 0 5 SD 6 10 0 16 5 12 0 17 5 11 1 17
PD 1 15 11 27 1 19 8 28 1 17 10 28 Total 1 25 11 47 9 33 8 50 10 29 11 50
TABLE 5. Kappa values for all methods Method Kappa (95% CI)
Visual 0.47 (0.24‐0.70) SULpeak 30% 0.39 (0.19‐0.59) SULpeak 25% 0.41 (0.20‐0.62) SULpeak 20% 0.41 (0.20‐0.62) SULpeak 20% 0.42 (0.22‐0.61) TLG (PERCIST) 45/75% 0.25 (n.c.) TLG (PERCIST) 50% 0.29 (0.20‐0.39) TLG (PERCIST) 40% 0.32 (0.16‐0.47) TLG (PERCIST) 30% 0.45 (0.24‐0.67) TLG (PERCIST) 25% 0.53 (0.31‐0.75) TLG (PERCIST) 20% 0.49 (0.29‐0.69) TLG (50) 45/70% 0.23 (n.c.)
TLG (50) 50% 0.27 (0.17‐0.36) TLG (50) 40% 0.25 (0.16‐0.33) TLG (50) 30% 0.35 (0.16‐0.53) TLG (50) 25% 0.33 (0.12‐0.53) TLG (50) 20% 0.38 (0.16‐0.61) TLG (40) 45/70% 0.27 (n.c.) TLG (40) 50% 0.34 (0.16‐0.52) TLG (40) 40% 0.33 (0.16‐0.50) TLG (40) 30% 0.42 (0.18‐0.66) TLG (40) 25% 0.44 (0.20‐0.69) TLG (40) 20% 0.43 (0.20‐0.66) TLG (30) 45/70% 0.24 (n.c.) TLG (30) 50% 0.33 (0.12‐0.54) TLG (30) 40% 0.38 (0.14‐0.62)
TLG (30) 30% 0.32 (0.12‐0.53) TLG (30) 25% 0.38 (0.15‐0.61)
TLG (30) 20% 0.41 (0.18‐0.64)
Quadratic weighted kappa values (95% confidence intervals) for all the methods analyzed including the various cut‐offs for response and progression. n.c.= not calculable by the method used, described by Fleiss et al (21).
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
Supplemental Figure 1
Correlation between %TLGs, %MTVs and %SULmeans at 7-10 days and change in CT
diameter (%SLD after 9-11 weeks.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
Supplemental Figure 2 Waterfall plots for SULmax and TLGs: In order to find PD patients without excluding any non-PD patients, we want to find a method that does not include any reds or blacks (or as far to right as possible) in the progression (left) side of the plots
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
SUPLEMENTARY TABLE 1: An overview of the various evaluation methods. Method Parameter Delineation SMD/PMR SMD/PMD 1 Visual change Sign decrease Sign increase 2a SULpeak 30% decrease 30% increase 2b SULpeak 25% decrease 25% increase 2c SULpeak 20% decrease 20% increase 2d SULpeak 15% decrease 15% increase 3a TLG PERCIST 2 SD above mean SUL liver* 45% decrease 75% increase 3b TLG PERCIST 2 SD above mean SUL liver* 50% decrease 50% increase 3c TLG PERCIST 2 SD above mean SUL liver* 40% decrease 40% increase 3d TLG PERCIST 2 SD above mean SUL liver* 30% decrease 30% increase 3e TLG PERCIST 2 SD above mean SUL liver* 25% decrease 25% increase 3f TLG PERCIST 2 SD above mean SUL liver* 20% decrease 20% increase 4a TLG 50 50% of SULmax of lesion 45% decrease 75% increase 4b TLG 50 50% of SULmax of lesion 50% decrease 50% increase 4c TLG 50 50% of SULmax of lesion 40% decrease 40% increase 4d TLG 50 50% of SULmax of lesion 30% decrease 30% increase 4e TLG 50 50% of SULmax of lesion 25% decrease 25% increase 4f TLG 50 50% of SULmax of lesion 20% decrease 20% increase 5a TLG 40 40% of SULmax of lesion 45% decrease 75% increase 5b TLG 40 40% of SULmax of lesion 50% decrease 50% increase 5c TLG 40 40% of SULmax of lesion 40% decrease 40% increase 5d TLG 40 40% of SULmax of lesion 30% decrease 30% increase 5e TLG 40 40% of SULmax of lesion 25% decrease 25% increase 5f TLG 40 40% of SULmax of lesion 20% decrease 20% increase 6a TLG 30 30% of SULmax of lesion 45% decrease 75% increase 6b TLG 30 30% of SULmax of lesion 50% decrease 50% increase 6c TLG 30 30% of SULmax of lesion 40% decrease 40% increase 6d TLG 30 30% of SULmax of lesion 30% decrease 30% increase 6e TLG 30 30% of SULmax of lesion 25% decrease 25% increase 6f TLG 30 30% of SULmax of lesion 20% decrease 20% increase
* normal mean liver = mean SUL in a 3 cm in diameter ROI in the right lobe of the liver (as according to PERCIST 1.0). The two methods described in PERCIST 1.0 are highlighted in bold. Supplemented with SULmax 25 and SUL max (15).
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
Supplementary Table 2. Data on blood Glucose levels, injected FDG-activity and uptake time per patient and the variation between scans. Patient Blood Glucose (mM) Injected FDG activity (MBq) Uptake time (minutes)
A B Diff.(%) A B Diff. (%) A B Diff. 2 8.1 7.4 8.6 284 312 9.9 58 65 7 4 6.1 7.6 24.6 277 297 7.2 56 65 9 6 7.2 5.9 18.1 237 230 -3.0 56 55 1 7 7.4 7.3 1.4 484 465 -3.9 59 48 11 8 5.5 6.8 23.6 372 377 1.3 55 57 2 9 4.9 5.7 16.3 344 331 -3.8 51 56 5 10 6.8 7.7 13.2 450 421 -6.4 57 61 4 11 7.9 9.0 13.9 415 390 -6.0 51 66 15 12 6.4 6.6 3.1 288 321 11.5 61 60 1 13 7.8 6.9 11.5 475 440 -7.4 69 55 14 14 6.4 6.4 0.0 426 385 -9.6 59 60 1 15 6.2 6.3 1.6 371 323 -12.9 59 61 2 16 7.4 5.8 21.6 323 319 -1.2 54 54 0 17 5.9 6.4 8.5 330 331 0.3 57 60 3 18 6.6 5.9 10.6 356 345 -3.1 60 64 4 19 5.7 5.5 3.5 237 263 11.0 62 63 1 22 6.2 5.6 9.7 547 505 -7.7 58 50 8 24 5.3 5.3 0.0 340 348 2.4 63 59 4 25 4.6 5.8 26.1 340 345 1.5 58 63 5 26 7.1 7.6 7.0 310 287 -7.4 59 62 3 27 6.3 6.3 0.0 406 399 -1.7 56 57 1 28 7.6 7.7 1.3 381 412 8.1 61 58 3 29 4.9 6.4 30.6 434 421 -3.0 65 60 5 30 5.7 6.0 5.3 255 244 -4.3 58 58 0 31 7.4 8.4 13.5 242 259 7.0 60 57 3 32 6.7 6.2 7.5 363 374 3.0 60 67 7 33 8.8 7.8 11.4 430 392 -8.8 55 62 7 34 5.2 6.8 30.8 306 330 7.8 55 55 0 35 6.3 6.7 6.3 485 460 -5.2 51 59 8 36 6.3 6.2 1.6 295 316 7.1 59 58 1 39 6.2 6.3 1.6 323 306 -5.3 57 56 1 40 5.6 5.8 3.6 295 295 0.0 57 60 3 41 5.5 5.9 7.3 253 227 -10.3 57 72 15 45 7.7 7.9 2.6 448 448 0.0 63 54 9 46 5.8 5.7 1.7 214 202 -5.6 68 55 13 47 6.3 5.3 15.9 295 301 2.0 58 66 8 48 5.9 6.7 13.6 302 248 -17.9 62 66 4 49 6.7 6.9 3.0 248 255 2.8 59 57 2 51 5.3 4.9 7.5 286 293 2.4 55 56 1 53 6.4 6.3 1.6 609 618 1.5 65 59 6 54 6.4 7.1 10.9 350 348 -0.6 59 51 8 55 6.2 5.8 6.5 330 287 -13.0 57 55 2 56 6.2 6.4 3.2 240 259 7.9 59 56 3 57 5.9 5.5 6.8 242 245 1.2 59 70 11 59 5.6 5.4 3.6 258 251 -2.7 66 53 13 60 6.3 7.7 22.2 197 199 1.0 66 63 3 62 5.4 4.7 13.0 344 375 9.0 56 63 7 63 6.1 5.7 6.6 252 254 0.8 54 58 4 64 5.9 6.8 15.3 398 404 1.5 61 58 3 65 5.3 5.6 5.7 305 313 2.6 57 61 4
A=baseline scan, B=Follow-up scan, Diff. = difference, for Glucose concentration and FDG-actvity it is percentage difference and for uptake-time, the difference is in minutes.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
Supplementary Table 3. The categorization tables, sensitivity, specificity and predictive values for all methods for PMD/non-PMD predicting PD/non-PD on CT
Sens Spec PPV NPV Visual 0.36 0.95 0.91 0.54 SULpeak 30% 0.25 1.00 1.00 0.52 SULpeak 25% 0.29 1.00 1.00 0.52 SULpeak 20% 0.29 1.00 1.00 0.52 SULpeak 20% 0.29 1.00 1.00 0.52 TLG (PERCIST) 45/70% 0.04 1.00 1.00 0.43 TLG (PERCIST) 50% 0.11 1.00 1.00 0.45 TLG (PERCIST) 40% 0.19 1.00 1.00 0.48 TLG (PERCIST) 30% 0.33 1.00 1.00 0.53 TLG (PERCIST) 25% 0.41 1.00 1.00 0.56 TLG (PERCIST) 20% 0.41 1.00 1.00 0.56 TLG (50) 45/70% 0.11 1.00 1.00 0.44 TLG (50) 50% 0.15 1.00 1.00 0.45 TLG (50) 40% 0.22 1.00 1.00 0.48 TLG (50) 30% 0.33 1.00 1.00 0.51 TLG (50) 25% 0.38 0.89 0.83 0.50 TLG (50) 20% 0.48 0.89 0.87 0.55 TLG (40) 45/70% 0.07 1.00 1.00 0.42 TLG (40) 50% 0.19 1.00 1.00 0.45 TLG (40) 40% 0.22 1.00 1.00 0.46 TLG (40) 30% 0.30 1.00 1.00 0.49 TLG (40) 25% 0.33 1.00 1.00 0.50 TLG (40) 20% 0.33 0.94 0.90 0.49 TLG (30) 45/70% 0.04 0.94 0.50 0.39 TLG (30) 50% 0.19 0.94 0.83 0.43 TLG (30) 40% 0.27 0.94 0.88 0.46 TLG (30) 30% 0.31 0.94 0.89 0.47 TLG (30) 25% 0.31 0.94 0.89 0.47 TLG (30) 20% 0.42 0.94 0.92 0.52
Table A: Sensitivity, specificity, PPV and NPV for all methods comparing progression versus non-progression on FDG-PET (PMD vs non-PMD) to the 9-11 week CT scan (PD vs non-PD). 2x2 tables for all methods are presented below.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
B)
3x3 tables for all methods comparing classification on FDG-PET/CT to the RECIST classification
for the 10 week CT.
Visual PMR SMD PMD Total
PR 4 1 0 5
SD 5 11 1 17
PD 1 17 10 28
Total 10 29 11 50
SULpeak 30% PMR SMD PMD Total
PR 3 2 0 5
SD 5 12 0 17
PD 1 20 7 28
Total 9 34 7 50
SULpeak 25% PMR SMD PMD Total
PR 3 2 0 5
SD 5 12 0 17
PD 1 19 8 28
Total 9 33 8 50
SULpeak 20% PMR SMD PMD Total
PR 3 2 0 5
SD 5 12 0 17
PD 1 19 8 28
Total 9 33 8 50
SULpeak 15% PMR SMD PMD Total
PR 3 2 0 5
SD 7 10 0 17
PD 1 19 8 28
Total 11 31 8 50
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG (PERCIST)
45/70%
PMR SMD PMD Total
PR 2 2 0 4
SD 4 12 0 16
PD 0 26 1 27
Total 6 40 1 47
TLG(PERCIST)50% PMR SMD PMD Total
PR 2 2 0 4
SD 3 13 0 16
PD 0 24 3 27
Total 5 39 3 47
TLG(PERCIST)40% PMR SMD PMD Total
PR 2 2 0 4
SD 5 11 0 16
PD 1 21 5 27
Total 8 34 5 47
TLG(PERCIST)30% PMR SMD PMD Total
PR 3 1 0 4
SD 6 10 0 16
PD 1 17 9 27
Total 10 28 9 47
TLG(PERCIST)25% PMR SMD PMD Total
PR 4 0 0 4
SD 6 10 0 16
PD 1 15 11 27
Total 11 25 11 47
TLG(PERCIST)20% PMR SMD PMD Total
PR 4 0 0 4
SD 8 8 0 16
PD 2 14 11 27
Total 14 22 11 47
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(50)45/70% PMR SMD PMD Total
PR 2 2 0 4
SD 4 11 0 15
PD 2 22 3 27
Total 8 35 3 46
TLG(50)40% PMR SMD PMD Total
PR 2 2 0 4
SD 5 10 0 15
PD 4 17 6 27
Total 11 29 6 46
TLG(50)25% PMR SMD PMD Total
PR 3 1 0 4
SD 5 8 2 15
PD 4 13 10 27
Total 12 22 12 46
TLG(50)50% PMR SMD PMD Total
PR 2 2 0 4
SD 5 10 0 15
PD 2 21 4 27
Total 9 33 4 46
TLG(50)30% PMR SMD PMD Total
PR 3 1 0 4
SD 5 10 0 15
PD 4 14 9 27
Total 12 25 9 46
TLG(50)20% PMR SMD PMD Total
PR 3 1 0 4
SD 5 8 2 15
PD 4 10 13 27
Total 12 19 15 46
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(40)45/70% PMR SMD PMD Total
PR 2 2 0 4
SD 3 11 0 14
PD 0 25 2 27
Total 5 38 2 45
TLG(40)40% PMR SMD PMD Total
PR 2 2 0 4
SD 3 11 0 14
PD 1 20 6 27
Total 6 33 6 45
TLG(40)25% PMR SMD PMD Total
PR 3 1 0 4
SD 3 11 0 14
PD 1 17 9 27
Total 7 29 9 45
TLG(40)50% PMR SMD PMD Total
PR 2 2 0 4
SD 3 11 0 14
PD 0 22 5 27
Total 5 35 5 45
TLG(40)30% PMR SMD PMD Total
PR 3 1 0 4
SD 3 11 0 14
PD 1 18 8 27
Total 7 30 8 45
TLG(40)20% PMR SMD PMD Total
PR 3 1 0 4
SD 5 8 1 14
PD 1 17 9 27
Total 9 26 10 45
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(30)45/70% PMR SMD PMD Total
PR 2 1 0 3
SD 3 10 1 14
PD 0 25 1 26
Total 5 36 2 43
TLG(30)40% PMR SMD PMD Total
PR 2 1 0 3
SD 3 10 1 14
PD 0 19 7 26
Total 5 30 8 43
TLG(30)25% PMR SMD PMD Total
PR 3 0 0 3
SD 4 9 1 14
PD 2 16 8 26
Total 9 25 9 43
TLG(30)50% PMR SMD PMD Total
PR 2 1 0 3
SD 3 10 1 14
PD 0 21 5 26
Total 5 32 6 43
TLG(30)30% PMR SMD PMD Total
PR 2 1 0 3
SD 3 10 1 14
PD 2 16 8 26
Total 7 27 9 43
TLG(30)20% PMR SMD PMD Total
PR 3 0 0 3
SD 5 8 1 14
PD 3 12 11 26
Total 11 20 12 43
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
C) 2x2 tables for CT PD/non-PD versus Methods PMR/non-PMR (left column) and PMD/non-PMD (right column) true positives are marked in red. The corresponding Sensitivity, specificity, PPV and NPV below each table for predicting PD on CT on predicting PD on CT. When using non-PMR as a sign of progression sensitivity is high range: 0.85-1.00, but specificity is rather low range: 0.25-0.50, best for method 3.
Using PMD as sign of progression the specificity is high (0.89-1.00), but the sensitivity is low:
0.04-0.48.
Visual Non-PMR PMR Total
Non-PD 13 9 22
PD 27 1 28
Total 44 10 50 Sensitivity = 0.96 Sensitivity = 0.36 Specificity = 0.41 Specificity = 0.95 PPV = 0.66 PPV = 0.91 NPV = 1.00 NPV = 0.54
SULpeak 30% Non-PMR PMR Total
Non-PD 14 8 22
PD 27 1 28
Total 41 9 50 Sensitivity = 0.96 Sensitivity = 0.25 Specificity = 0.36 Specificity = 1.00 PPV = 0.66 PPV = 1.00 NPV = 0.89 NPV = 0.52
SULpeak 25% Non-PMR PMR Total
Non-PD 14 8 22
PD 27 1 28
Total 41 9 50 Sensitivity = 0.96 Sensitivity = 0.29 Specificity = 0.36 Specificity = 1.00 PPV = 0.66 PPV = 1.00 NPV = 0.89 NPV = 0.52
Visual Non-PMD PMD Total
Non-PD 21 1 22
PD 18 10 28
Total 39 11 50
SULpeak 30% Non-PMD PMD Total
Non-PD 22 0 22
PD 21 7 28
Total 43 7 49
SULpeak 25% Non-PMD PMD Total
Non-PD 22 0 22
PD 20 8 28
Total 42 8 50
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
SULpeak 20% Non-PMR PMR Total
Non-PD 14 8 22
PD 27 1 28
Total 41 9 50 Sensitivity = 0.96 Sensitivity = 0.29 Specificity = 0.36 Specificity = 1.00 PPV = 0.66 PPV = 1.00 NPV = 0.89 NPV = 0.52
SULpeak 15% Non-PMR PMR Total
Non-PD 12 10 22
PD 27 1 28
Total 39 11 50 Sensitivity = 0.96 Sensitivity = 0.29 Specificity = 0.45 Specificity = 1.00 PPV = 0.69 PPV = 1.00 NPV = 0.91 NPV = 0.52
TLG(PERCIST)
45/70%
Non-PMR PMR Total
Non-PD 14 6 20
PD 27 0 27
Total 41 6 47 Sensitivity = 1.0 Sensitivity = 0.04 Specificity = 0.30 Specificity = 1.00 PPV = 0.66 PPV = 1.00 NPV = 1.0 NPV = 0.43
SULpeak 20% Non-PMD PMD Total
Non-PD 22 0 22
PD 20 8 28
Total 42 8 50
SULpeak 15% Non-PMD PMD Total
Non-PD 22 0 22
PD 20 8 28
Total 42 8 50
TLG(PERCIST)
45/70%
Non-PMD PMD Total
Non-PD 20 0 20
PD 26 1 27
Total 46 1 47
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(PERCIST)50% Non-PMR PMR Total
Non-PD 15 5 20
PD 27 0 27
Total 44 5 47 Sensitivity = 1.00 Sensitivity = 0.11 Specificity = 0.25 Specificity = 1.00 PPV = 0.64 PPV = 1.00 NPV = 1.00 NPV = 0.45
TLG(PERCIST)40% Non-PMR PMR Total
Non-PD 13 7 20
PD 26 1 27
Total 39 8 47 Sensitivity = 0.96 Sensitivity = 0.19 Specificity = 0.35 Specificity = 1.00 PPV = 0.67 PPV = 1.00 NPV = 0.88 NPV = 0.48
TLG(PERCIST)30% Non-PMR PMR Total
Non-PD 11 9 20
PD 26 1 27
Total 37 10 47 Sensitivity = 0.96 Sensitivity = 0.33 Specificity = 0.45 Specificity = 1.00 PPV = 0.70 PPV = 1.00 NPV = 0.90 NPV = 0.53
TLG(PERCIST)50% Non-PMD PMD Total
Non-PD 20 0 20
PD 24 3 27
Total 44 3 47
TLG(PERCIST)40% Non-PMD PMD Total
Non-PD 20 0 20
PD 22 5 27
Total 42 5 47
TLG(PERCIST)30% Non-PMD PMD Total
Non-PD 20 0 20
PD 18 9 27
Total 38 9 47
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(PERCIST)25% Non-PMR PMR Total
Non-PD 10 10 20
PD 26 1 27
Total 36 11 47 Sensitivity = 0.96 Sensitivity = 0.41 Specificity = 0.50 Specificity = 1.00 PPV = 0.72 PPV = 1.00 NPV = 0.91 NPV = 0.56
TLG(PERCIST)20% Non-PMR PMR Total
Non-PD 8 12 20
PD 25 2 27
Total 33 14 47 Sensitivity = 0.93 Sensitivity = 0.41 Specificity = 0.60 Specificity = 1.00 PPV = 0.76 PPV = 1.00 NPV = 0.86 NPV = 0.56
TLG(50)45/70% Non-PMR PMR Total
Non-PD 13 6 19
PD 25 2 26
Total 38 8 46 Sensitivity = 0.93 Sensitivity = 0.11 Specificity = 0.32 Specificity = 1.00 PPV = 0.68 PPV = 1.00 NPV = 0.75 NPV = 0.44
TLG(PERCIST)25% Non-PMD PMD Total
Non-PD 20 0 20
PD 16 11 27
Total 36 11 47
TLG(PERCIST)20% Non-PMD PMD Total
Non-PD 20 0 20
PD 16 11 27
Total 36 11 47
TLG(50)45/70% Non-PMD PMD Total
Non-PD 19 0 19
PD 24 3 27
Total 43 3 46
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(50)50% Non-PMR PMR Total
Non-PD 12 7 19
PD 25 2 27
Total 37 9 46 Sensitivity = 0.93 Sensitivity = 0.15 Specificity = 0.37 Specificity = 1.00 PPV = 0.67 PPV = 1.00 NPV = 0.78 NPV = 0.45
TLG(50)40% Non-PMR PMR Total
Non-PD 12 7 19
PD 23 4 27
Total 35 11 46 Sensitivity = 0.85 Sensitivity = 0.22 Specificity = 0.37 Specificity = 1.00 PPV = 0.66 PPV = 1.00 NPV = 0.64 NPV = 0.48
TLG(50)30% Non-PMR PMR Total
Non-PD 11 8 19
PD 23 4 27
Total 34 12 46 Sensitivity = 0.85 Sensitivity = 0.33 Specificity = 0.42 Specificity = 1.00 PPV = 0.68 PPV = 1.00 NPV = 0.67 NPV = 0.51
TLG(50)50% Non-PMD PMD Total
Non-PD 19 0 19
PD 23 4 27
Total 42 4 46
TLG(50)40% Non-PMD PMD Total
Non-PD 19 0 19
PD 21 6 27
Total 40 6 46
TLG(50)30% Non-PMD PMD Total
Non-PD 19 0 19
PD 18 9 27
Total 37 9 46
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(50)25% Non-PMR PMR Total
Non-PD 11 8 19
PD 23 4 27
Total 34 12 46 Sensitivity = 0.85 Sensitivity = 0.38 Specificity = 0.42 Specificity = 0.89 PPV = 0.68 PPV = 0.83 NPV = 0.67 NPV = 0.50
TLG(50)20% Non-PMR PMR Total
Non-PD 11 8 19
PD 23 4 27
Total 34 12 46 Sensitivity = 0.85 Sensitivity = 0.48 Specificity = 0.42 Specificity = 0.89 PPV = 0.68 PPV = 0.87 NPV = 0.67 NPV = 0.55
TLG(40)45/70% Non-PMR PMR Total
Non-PD 13 5 18
PD 27 0 27
Total 40 5 45 Sensitivity = 1.00 Sensitivity = 0.07 Specificity = 0.28 Specificity = 1.00 PPV = 0.68 PPV = 1.00 NPV = 1.00 NPV = 0.42
TLG(50)25% Non-PMD PMD Total
Non-PD 17 2 19
PD 17 10 27
Total 34 12 46
TLG(50)20% Non-PMD PMD Total
Non-PD 17 2 19
PD 14 13 27
Total 31 15 46
TLG(40)45/70% Non-PMD PMD Total
Non-PD 18 0 18
PD 25 2 27
Total 43 2 45
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(40)50% Non-PMR PMR Total
Non-PD 13 5 18
PD 27 0 27
Total 40 5 45 Sensitivity = 1.00 Sensitivity = 0.19 Specificity = 0.28 Specificity = 1.00 PPV = 0.68 PPV = 1.00 NPV = 1.00 NPV = 0.45
TLG(40)40% Non-PMR PMR Total
Non-PD 12 6 18
PD 26 1 27
Total 38 7 45 Sensitivity = 0.96 Sensitivity = 0.22 Specificity = 0.33 Specificity = 1.00 PPV = 0.68 PPV = 1.00 NPV = 0.86 NPV = 0.46
TLG(40)30% Non-PMR PMR Total
Non-PD 12 6 18
PD 26 1 27
Total 38 7 45 Sensitivity = 0.96 Sensitivity = 0.30 Specificity = 0.33 Specificity = 1.00 PPV = 0.68 PPV = 1.00 NPV = 0.86 NPV = 0.49
TLG(40)50% Non-PMD PMD Total
Non-PD 18 0 18
PD 22 5 27
Total 40 5 45
TLG(40)40% Non-PMD PMD Total
Non-PD 18 0 18
PD 21 6 27
Total 39 6 45
TLG(40)30% Non-PMD PMD Total
Non-PD 18 0 18
PD 19 8 27
Total 37 8 45
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(40)25% Non-PMR PMR Total
Non-PD 12 6 18
PD 26 1 27
Total 38 7 45 Sensitivity = 0.96 Sensitivity = 0.33 Specificity = 0.33 Specificity = 1.00 PPV = 0.68 PPV = 1.00 NPV = 0.86 NPV = 0.50
TLG(40)20% Non-PMR PMR Total
Non-PD 10 8 18
PD 26 1 27
Total 36 9 45 Sensitivity = 0.96 Sensitivity = 0.33 Specificity = 0.44 Specificity = 0.94 PPV = 0.72 PPV = 0.90 NPV = 0.89 NPV = 0.49
TLG(30)45/70% Non-PMR PMR Total
Non-PD 12 5 17
PD 26 0 26
Total 38 5 43 Sensitivity = 1.00 Sensitivity = 0.04 Specificity = 0.29 Specificity = 0.94 PPV = 0.68 PPV = 0.5 NPV = 1.00 NPV = 0.39
TLG(40)25% Non-PMD PMD Total
Non-PD 18 0 18
PD 18 9 27
Total 36 9 45
TLG(40)20% Non-PMD PMD Total
Non-PD 17 1 18
PD 18 9 27
Total 35 10 45
TLG(30)45/70% Non-PMD PMD Total
Non-PD 16 1 17
PD 25 1 26
Total 41 2 43
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(30)50% Non-PMR PMR Total
Non-PD 12 5 17
PD 26 0 26
Total 38 5 43 Sensitivity = 1.00 Sensitivity = 0.19 Specificity = 0.29 Specificity = 0.94 PPV = 0.68 PPV = 0.83 NPV = 1.00 NPV = 0.43
TLG(30)40% Non-PMR PMR Total
Non-PD 12 5 17
PD 26 0 26
Total 38 5 43 Sensitivity = 1.00 Sensitivity = 0.27 Specificity = 0.29 Specificity = 0.94 PPV = 0.68 PPV = 0.88 NPV = 1.00 NPV = 0.46
TLG(30)30% Non-PMR PMR Total
Non-PD 12 5 17
PD 24 2 25
Total 36 7 43 Sensitivity = 0.92 Sensitivity = 0.31 Specificity = 0.29 Specificity = 0.94 PPV = 0.67 PPV = 0.89 NPV = 0.71 NPV = 0.47
TLG(30)50% Non-PMD PMD Total
Non-PD 16 1 17
PD 21 5 26
Total 37 6 43
TLG(30)40% Non-PMD PMD Total
Non-PD 16 1 17
PD 19 7 26
Total 35 8 43
TLG(30)30% Non-PMD PMD Total
Non-PD 16 1 17
PD 18 8 26
Total 34 9 43
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
TLG(30)25% Non-PMR PMR Total
Non-PD 10 7 17
PD 24 2 26
Total 34 9 43 Sensitivity = 0.92 Sensitivity = 0.31 Specificity = 0.41 Specificity = 0.94 PPV = 0.71 PPV = 0.88 NPV = 0.78 NPV = 0.47
TLG(30)20% Non-PMR PMR Total
Non-PD 9 8 17
PD 23 3 25
Total 32 11 43 Sensitivity = 0.88 Sensitivity = 0.42 Specificity = 0.47 Specificity = 0.94 PPV = 0.72 PPV = 0.92 NPV = 0.73 NPV = 0.52
TLG(30)25% Non-PMD PMD Total
Non-PD 16 1 17
PD 18 8 26
Total 34 9 43
TLG(30)20% Non-PMD PMD Total
Non-PD 16 1 17
PD 15 11 26
Total 31 12 43
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from
Doi: 10.2967/jnumed.117.193003Published online: May 10, 2017.J Nucl Med. Jørgen Frøkiær and Peter MeldgaardJoan Fledelius, Anne Winther-Larsen, Azza Ahmed Khalil, Catharina Mølgaard Bylov, Karin Hjorthaug, Aksel Bertelsen,
A comparison of assessment methods−treated NSCLC patients F-FDG-PET/CT for very early response evaluation predicts CT response in Erlotinib18
http://jnm.snmjournals.org/content/early/2017/05/10/jnumed.117.193003This article and updated information are available at:
http://jnm.snmjournals.org/site/subscriptions/online.xhtml
Information about subscriptions to JNM can be found at:
http://jnm.snmjournals.org/site/misc/permission.xhtmlInformation about reproducing figures, tables, or other portions of this article can be found online at:
and the final, published version.proofreading, and author review. This process may lead to differences between the accepted version of the manuscript
ahead of print area, they will be prepared for print and online publication, which includes copyediting, typesetting,JNMcopyedited, nor have they appeared in a print or online issue of the journal. Once the accepted manuscripts appear in the
. They have not beenJNM ahead of print articles have been peer reviewed and accepted for publication in JNM
(Print ISSN: 0161-5505, Online ISSN: 2159-662X)1850 Samuel Morse Drive, Reston, VA 20190.SNMMI | Society of Nuclear Medicine and Molecular Imaging
is published monthly.The Journal of Nuclear Medicine
© Copyright 2017 SNMMI; all rights reserved.
by on April 5, 2020. For personal use only. jnm.snmjournals.org Downloaded from