Prospective Comparison of [18F]FluorodeoxyglucosePositron Emission Tomography with ConventionalAssessment by Computed Tomography Scans andSerum Tumor Markers for the Evaluation of ResidualMasses in Patients with Nonseminomatous Germ CellCarcinoma

Christian Kollmannsberger, M.D.1

Karin Oechsle, M.D.1

Bernhard M. Dohmen, M.D.2

Anna Pfannenberg, M.D.3

Roland Bares, M.D.2

Claus D. Claussen, M.D.3

Lothar Kanz, M.D.1

Carsten Bokemeyer, M.D.1

1 Department of Hematology/Oncology, Universityof Tuebingen, Tuebingen, Germany.

2 Department of Nuclear Medicine, University ofTuebingen, Tuebingen, Germany.

3 Department of Radiology, University of Tuebin-gen, Tuebingen, Germany.

Address for reprints: Carsten Bokemeyer, M.D.,Department of Hematology/Oncology, University ofTuebingen Medical Center, Otfried-Mueller-Strasse 10, 72076 Tuebingen, Germany; Fax:�49-7071-293675; E-mail: carsten.bokemeyer@med.uni-tuebigen.de

Received March 2, 2001; revision received July 9,2001; accepted January 20, 2002.

BACKGROUND. To assess the ability of [18F]fluorodeoxyglucose (F-18 FDG) positron

emission tomography (PET) to predict the viability of residual masses after che-

motherapy in patients with metastatic nonseminomatous germ cell tumors (GCT),

PET results were compared in a blinded analysis with computed tomography (CT)

scans and serum tumor marker changes (TUM) as established methods of assess-

ment.

METHODS. Independent reviewers who were blinded to each other’s results eval-

uated the PET results and corresponding CT scan and TUM results in 85 residual

lesions from 45 patients. All patients were treated within prospective clinical trials

and received primary/salvage, high-dose chemotherapy with autologous blood

stem cell support for primary poor prognosis disease or recurrent disease. PET

results were assessed both visually and by quantifying glucose uptake (standard-

ized uptake values). Results were validated either by histologic examination of a

resected mass and/or biopsy (n � 28 lesions) or by a 6-month clinical follow-up

after evaluation (n � 57 lesions).

RESULTS. F-18 FDG PET showed increased tracer uptake in 32 of 85 residual

lesions, with 29 true positive (TP) lesions and three false positive (FP) lesions.

Fifty-three lesions were classified by PET as negative (no viable GCT), 33 lesions

were classified by PET as true negative (TN), and 20 lesions were classified by PET

as false negative (FN). In the blinded reading of the corresponding CT scan and

TUM results, 38 residual lesions were assessed correctly as containing viable

carcinoma and/or teratoma. Forty-six lesions were classified as nonsuspicious by

CT scan/TUM (33 TN lesions and 14 falsely classified lesions). PET correctly

predicted the presence of viable carcinoma in 5 of these 14 and the absence of

viable carcinoma in 3 of these 14 lesions. Resulting sensitivities and specificities for

the prediction of residual mass viability were as follows: PET, 59% sensitivity and

92% specificity; radiologic monitoring, 55% sensitivity and 86% specificity; and

TUM, 42% sensitivity and 100% specificity. The positive and negative predictive

values for PET were 91% and 62%, respectively. The diagnostic efficacy of PET did

not improve when patients with teratomatous elements in the primary tumor were

excluded from the analysis. In patients with multiple residual masses, a uniformly

increased residual F-18 FDG uptake in all lesions was a strong predictor for the

presence of viable carcinoma.

CONCLUSIONS. F-18 FDG PET imaging performed in conjunction with conventional

staging methods offers additional information for the prediction of residual mass

2353

© 2002 American Cancer Society

histology in patients with nonseminomatous GCT. A positive PET is highly predic-

tive for the presence of viable carcinoma. Other useful indications for a PET

examination include patients with multiple residual masses and patients with

marker negative disease. Cancer 2002;94:2353– 62.

© 2002 American Cancer Society.

DOI 10.1002/cncr.10494

KEYWORDS: germ cell carcinoma, residual masses, positron emission tomography,tumor markers, computed tomography scan.

The development of cisplatin-based combinationchemotherapy has dramatically improved the

prognosis of patients with metastatic nonseminoma-tous germ cell tumors (NSGCT), resulting in a long-term cure rate of 70 – 80%.1,2 After the completion ofchemotherapy, residual tumor lesions are found in upto 40% of patients with advanced NSGCT, despite thenormalization of serum tumor markers.3 The differen-tiation between viable carcinoma, mature teratoma,and necrosis is possible only with histologic examina-tion of the resected specimen. Thus, the surgical re-section of all residual masses, if technically possible, isthe recommended standard of care. Approximately50% of residual lesions consist of necrosis only,whereas mature teratoma is found in 30% of residualmasses, and the remaining 20% of residual lesionscontain viable immature tumor cells.4 In the lattergroup of patients, secondary resection may improvethe prognosis of the patient depending on the biologicaggressiveness of the disease, the completeness of sur-gery, and the effectiveness of additional therapy. Theresection of mature teratoma also is important, be-cause subsequent problems caused by local tumorgrowth or late disease recurrence with possible malig-nant, nongerminal transformation can be avoided.However, in the case of necrotic tissue, patients do notbenefit from secondary surgery yet are exposed tosurgery-related morbidity and mortality.5,6 Therefore,it is of great interest to identify those patients whorequire a secondary postchemotherapy resection andthose who do not in order to optimize individual treat-ment. To date, no diagnostic tool has been developedthat reliably predicts in vivo the viability of the resid-ual mass. Unfortunately, radiologic criteria derivedfrom computed tomography (CT) scans or magneticresonance imaging (MRI) have failed to differentiatereliably between viable carcinoma, mature teratoma,and necrosis/scar tissue from residual lesions in pa-tients with NSGCT. Serum tumor markers also may bemisleading, because viable tumor may be found insome patients with normalized serum tumor markers,and necrosis may be found in some patients with aprolonged elevation of serum tumor markers.4,7

Positron emission tomography (PET) imaging us-

ing 2-[18F]fluoro-2-deoxy-D-glucose (F-18 FDG) is anew diagnostic technique that allows the visualizationand semiquantitative calculation of regional glucosemetabolism within the body.8,9 Because tumor cellsare characterized by a higher glucolytic rate than nor-mal tissue cells, PET exploits this difference by assess-ing the rate and quantity of F-18 FDG uptake by thetumor. PET has been demonstrated as a valuable im-aging technique for the accurate staging of lympho-ma.10 –12 In patients with testicular carcinoma, PEThas demonstrated a high sensitivity and specificity forthe detection of metastases at initial diagnosis.13,14

Thus, the primary objective of the current studywas to evaluate prospectively the ability of PET topredict the presence of viable carcinoma, mature ter-atoma, or necrosis in residual masses from patientswith metastatic NSGCT. Because both tumor markerdecline and radiologic imaging currently serve as es-tablished methods to assess the response to chemo-therapy and the viability of residual masses, the po-tential value of PET in evaluating residual masses wascompared with those methods.

MATERIALS AND METHODSPatients and TreatmentPatients with either newly diagnosed, metastatic, poorprognosis NSGCT according to the International GermCell Cancer Collaborative Group classification or pa-tients with recurrent disease after cisplatin-based che-motherapy and at least one residual mass measuring� 1 cm in greatest dimension on a CT scan wereeligible for inclusion in the PET protocol. All patientswere treated within one of two prospective Germanmulticenter, high-dose chemotherapy trials betweenSeptember 1995 and October 1999.15–17 Eligibility cri-teria were similar for both high-dose chemotherapytrials and consisted of the following: germ cell carci-noma of any primary tumor site, Karnofsky perfor-mance status � 50%, normal kidney function, absenceof severe heart or liver disease, and written informedconsent. Patients with poor prognosis germ cell carci-noma at initial diagnosis received sequential, first-line, high-dose chemotherapy plus autologous stemcell support. The treatment protocol consisted of one

2354 CANCER May 1, 2002 / Volume 94 / Number 9

cycle of standard-dose etoposide, ifosfamide, and cis-platin (VIP) chemotherapy (cisplatin 20 mg/m2, eto-poside 75 mg/m2, and ifosfamide 1200 mg/m2 dailyfor 5 days) followed by three cycles of high-dose VIPchemotherapy (cisplatin 20 mg/m2, etoposide 300g/m2, and ifosfamide 2000 –2400 mg/m2 daily for 5consecutive days every 3 weeks for a total of 3 cycles).Patients with recurrent disease were treated with threecycles of standard-dose paclitaxel, ifosfamide, and cis-platin (TIP) chemotherapy followed by one cycle ofthiotepa, etoposide, and carboplatin (TEC) high-dosechemotherapy. Standard-dose TIP chemotherapyconsisted of paclitaxel 175 mg/m2 given on Day 1 andifosfamide 1200 mg/m2 and cisplatin 20 mg/m2, bothadministered on Days 2– 6 of a 22-day cycle. The TEChigh-dose regimen contained thiotepa 150 mg/m2,etoposide 600 mg/m2, and carboplatin 500 mg/m2,with all drugs given daily over 3 days. All patientsreceived autologous peripheral blood stem cell sup-port and granulocyte-colony stimulating factor afterhigh-dose chemotherapy according to treatment pro-tocol and institutional practice.

All patients were treated at Tuebingen UniversityMedical Center. The current study was approved bythe Ethical Committee of the University of Tuebingen.All patients were required to provide written informedconsent.

Tumor Response EvaluationAll patients underwent extensive staging proceduresthat included either a spiral CT scan with orally andintravenously administered contrast medium or anMRI of the chest, abdomen, and brain; determinationof serum tumor marker levels (�-human chorionicgonadotrophin, �-fetoprotein, and lactate dehydroge-nase) as well as a baseline PET image prior to the startof chemotherapy. Serum tumor marker levels weredetermined prior to each chemotherapy cycle, and aCT scan or MRI of the tumor lesions was performedafter every second cycle. After the completion of che-motherapy, all patients underwent a PET examinationas well as a CT scan or MRI to assess residual masses.The PET examination and the CT scan were performedat least 3 weeks after treatment. All staging procedureswere performed within 3 weeks of one another toallow adequate comparisons. If it was feasible techni-cally, all residual masses were resected after the com-pletion of chemotherapy. Follow-up examinations,which included spiral CT scans and serum tumormarker level evaluations, were performed in 3-monthintervals after the completion of therapy.

All CT scans were reviewed by an independent,board-certified radiologist at the Department of Radi-ology of the University of Tuebingen who was not

aware of the PET findings. Most CT scans were per-formed at this department using standard, state-of-the-art techniques (spiral CT scanning with a slicethickness of 8 mm, a table feed of 12 mm, and incre-ments of 7 mm; oral and intravenous contrast media).Clinical response was classified according to the mod-ified World Health Organization criteria.18 The criteriafor the assessment of viability on CT scans werechanges in tumor size compared with the tumor sizeon initial CT scan and the degree of contrast enhance-ment. A decrease in tumor size � 50% and/or dimin-ished or absent contrast medium uptake were consid-ered radiologic signs of a nonviable lesion. Progressivelesions, lesions with a reduction � 50% in size, andlesions with persistent/increased contrast mediumuptake were rated viable. Compared with baselinevalues prior to therapy, a serum tumor marker decline� 90% after chemotherapy was classified as a favor-able response predicting a nonviable residual lesion. Atumor marker decrease � 90% or increased tumormarker levels were considered unfavorable responsesand were predictive of a viable residual mass.

PET ImagingA dedicated PET scanner (ADVANCE; General Elec-trics Medical Systems, Milwaukee, WI) was used, pro-viding an axial field of view (FOV) of 14.6 cm androtating 68Ga/68Ge line sources for measured attenu-ation correction. Emission data were corrected forrandom events, attenuation (restricted to FOVs con-taining residual tumor masses in patients with early-stage tumors), and scattering. Thirty-five slices (4.25mm) per FOV were reconstructed iteratively or byfiltered back projection with a 128 � 128 pixel matrix(4.3 � 4.3 mm pixel size), resulting in a final resolutionof about 8 mm (full width at half maximum). Thereconstructed images were printed on film (MatrixLR3300 P Laser Imager; Agfa-Gevaert, Mortsel, Bel-gium) using a black-and-white color table represent-ing standard uptake values (SUVs) of 0 – 8 (SUV 8,black) to facilitate visual image analysis. In patientswho underwent whole body scans, coronal slices (8.6mm) also were documented.

All patients fasted for a minimum of 12 hoursprior to PET imaging. Blood glucose levels werechecked for each patient before the intravenous ad-ministration of 250 –500 megabecquerels of F-18 FDG.Forty-five to 60 minutes after the F-18 FDG injection,static PET scans were recorded for 5–15 minutes perFOV (depending on reconstruction algorithm, injectedradioactivity, and patient weight), typically coveringthe head, the trunk, and the upper legs to the midfe-mur. Transmission scanning (3–20 minutes per FOV,depending on patient weight, transmission source ac-

F-18 FDG PET in Patients with NSGCT/Kollmannsberger et al. 2355

tivity, and whether segmentation was applied; sino-gram windowing) was performed prior to F-18 FDGinjection or after emission scanning. After tracer in-jection, patients received 1 L of fluid orally or intrave-nously and 20 mg furosemide intravenously to mini-mize image artifacts from residual radioactivity in theurinary tract.

PET Image AnalysisEach PET image was reviewed by an experienced nu-clear medicine physician from the Department of Nu-clear Medicine at the University of Tuebingen whowas blinded to both the serum tumor marker courseand the interpretation of CT scan and/or MRI results.PET images were classified as either positive or nega-tive by visual assessment for all patients and wereclassified subsequently by semiquantitative analysisusing SUVs in 75 of 85 patients.8,19 No SUVs werecalculated in 10 patients. Residual tumor lesions withan SUV � 2 were considered positive. On visual as-sessment, focally increased F-18 FDG uptake exceed-ing that of the surrounding tissue and/or contralateralbody regions was interpreted as viable tumor tissue. Inaddition, tomographic whole body coronal imageswere reviewed for F-18 FDG accumulations outside ofknown lesions.

PET and conventional assessment (CT scan andtumor marker decline) results also were correlatedwith histologic findings in residual masses from pa-tients who underwent secondary resection after thecompletion of chemotherapy. If no resection of resid-ual tumor masses was performed, then the clinicalcourse of the patient was used. The absence of tumorprogression on CT scans or tumor marker increaseswithin 6 months after chemotherapy were consideredindicators of a nonviable residual lesion. Significancelevels for differences between median SUVs were cal-culated using the Fisher exact test.

For the evaluation of the predictive ability of PETimaging, each mass was assessed separately. Similarly,the comparison of PET images with CT scans wasbased on the comparison of each residual mass. Theassessment of serum tumor marker decline was basedon each patient, because tumor markers are the samefor all residual lesions in a patient. When comparingPET images with the combination of CT scan/serumtumor markers, each residual mass was assessed to-gether with the corresponding serum tumor markerdecline to allow an exact comparison.

RESULTSEighty-five residual lesions in 45 patients who re-ceived high-dose chemotherapy for NSGCT (32 pa-tients for recurrent disease and 13 patients for poor

prognosis disease at initial diagnosis) were evaluatedwith F-18 FDG PET before and after treatment. Patientcharacteristics are listed in Table 1. The most commonmetastatic sites were the retroperitoneal lymph nodesand the lungs. Eight patients presented with an ex-tragonadal primary tumor. Baseline PET images priorto the start of chemotherapy showed an increasedF-18 FDG uptake in the tumors from all patients. Afterthe completion of chemotherapy, 21 patients (47%)showed a marker negative partial response, and 8

TABLE 1Patient Characteristics (N � 45 patients)

Characteristic No. (%)

Median age in yrs (range) 33 (21–57)Primary tumor localization

Gonadal 37 (82)Extragonadal 8 (18)

Primary histologyMalignant teratoma 6 (13)Embryonal carcinoma 7 (16)Yolk sac 2 (4)Chorion-carcinoma 3 (7)Mixed histology 24 (53)

With teratoma 13Without teratoma 11

Other/not known 3 (7)Location of metastases

Retroperitoneal lymph nodes 28 (62)Lungs 24 (53)Mediastinum 12 (27)Liver 10 (22)Other 11 (24)

Patients with elevated serum tumor markers�-HCG 22 (49)AFP 24 (53)LDH 15 (33)None 3 (7)

Treatment situationFirst-line treatment 13 (29)Treatment for first recurrence 32 (71)

Response to therapyPR marker negative 21 (47)PR marker positive 8 (18)PR (no initially elevated TUM) 1 (2)Stable disease 11 (24)Progression 4 (9)

Histologic findings/secondary resection (n � 28 localizations,n � 18 patients)

Viable carcinoma 13 (46)Mature teratoma 3 (11)Necrosis 12 (43)

Course of disease after the end of therapyNo disease progression (for at least 6 months) 21 (47)Disease progression (within 6 months) 24 (53)

Median follow-up in months (range) 27 (6–62)

HCG: human chorionic gonadotropin; AFP: �-fetoprotein; LDH: lactate dehydrogenase; PR: partial

response; TUM: serum tumor marker.

2356 CANCER May 1, 2002 / Volume 94 / Number 9

patients (18%) showed a partial remission withoutmarker normalization. Fifteen patients (33%) had sta-ble disease or progressive disease within 4 weeks afterthe completion of chemotherapy. One patient (2%)who was without initial marker elevation had a partialresponse.

A histologic specimen was available after second-ary residual mass resection in 8 patients with a total of28 residual lesions; whereas, in 27 patients with a totalof 57 lesions, no resection could be performed. There-fore, the clinical course over the subsequent 6 monthsafter evaluation was monitored in the latter patients.Histologic examination revealed viable carcinoma in12 lesions (43%), mature teratoma in 3 lesions (11%),and necrosis in 13 lesions (46%). Overall, 21 patients(47%) remained free of disease progression for at least6 months after high-dose chemotherapy, whereas 24patients (53%) experienced disease recurrence withinthe same period. Sixteen patients experienced diseaserecurrence beyond 6 months follow-up (7 patientsafter 6 –12 months, 7 patients after 12–18 months, and2 patients after � 18 months). These patients alsoshowed markedly increased tumor markers at the timeof recurrence, indicating the presence of viable carci-noma rather than mature teratoma. The median fol-low-up of all patients was 27 months (range, 6 – 62months).

Evaluation of Residual Masses by PETOverall, viability was assessed correctly by PET imag-ing in 62 of 85 residual masses (73%). A negative PETwas found by visual assessment in 53 residual masses(62%). In 20 of these lesions, either tumor progressionwas observed within 6 months after treatment (18lesions), or mature teratoma was found on histologicexamination (2 lesions), indicating false negative PETfindings. Thirty-two lesions (38%) in 22 patients wereconsidered positive by the PET reviewer. Nineteenpatients with a total of 29 lesions either experienceddisease recurrence within 6 months after high-dosechemotherapy, or the histology of the resected resid-ual tumor mass after high-dose chemotherapy stillrevealed the presence of viable carcinoma/mature ter-atoma. Three residual masses in three different pa-tients were considered positive, but the patients had afavorable outcome. In two patients, the histology ofthe resected mass revealed inflammation; whereas, inthe third patient, necrosis was found. Thus, the sen-sitivity and specificity of PET for the prediction ofresidual mass viability after the completion of chemo-therapy were 59% and 92%, respectively. The positivepredictive value of PET for the presence of viablecarcinoma/teratoma was 91%, whereas the negativepredictive value of PET was 62% (Table 2).

Quantification of F-18 FDG uptake by calculatingSUVs was performed for direct comparison in 75 re-sidual lesions (Table 3). The average SUV of true pos-itive lesions was 2.6 (range, 0.9 –10), and the averageSUV of true negative lesions was 1.2 (range, 0.9 –5.8; P� 0.05). The median SUV of lesions that containednecrosis was 1.5 (range, 0.9 –5.8), and, in the two re-sected lesions that contained mature teratoma, themedian SUVs were 1.5 and 1.7 (P � 0.05). Thus, thesensitivity (68%) and negative predictive value (67%)improved slightly by the estimation of SUVs comparedwith visual assessment alone (sensitivity, 59%; nega-tive predictive value, 62%). In contrast, the specificity(83%) and positive predictive value (84%) decreasedslightly.

Twenty-eight of 45 patients presented with resid-

TABLE 2Comparison of the Sensitivity, Specificity, and Negative and PositivePredictive Values of Positron Emission Tomography (Visualassessment), Computed Tomography Scans/Magnetic ResonanceImaging, and Serum Tumor Markers for the Assessmentof Viability of Residual Masses

Parameter

PET (n � 85lesions)

CT scans/MRI(n � 85lesions)

Tumor markers(n � 42 patientswith a total of78 lesions)a

% 95% CI % 95% CI % 95% CI

Sensitivity 59 44–73% 55 40–69% 42 23–63%Specificity 92 78–98% 86 71–95% 100 79–100%Negative predictive value 62 48–75% 58 44–72% 52 33–70%Positive predictive value 91 75–98% 84 67–95% 100 71–100%

95% CI: 95% confidence interval.a Three patients with a total of 7 residual lesions and without serum tumor marker elevation at the time

of initial diagnosis were excluded.

TABLE 3Comparison of the Sensitivity, Specificity, and Negative and PositivePredictive Values of Visual Assessment and Quantitative Analysis(Standard uptake value) of Positron Emission Tomography inPatients with Residual Masses after High-Dose Chemotherapyfor Nonseminomatous Germ Cell Tumorsa

Parameter

Visual assessmentQuantitativeanalysis

% 95% CI % 95% CI

Sensitivity 59 44–73% 68 49–93%Specificity 92 78–98% 83 63–95%Negative predictive value 62 48–75% 67 47–83%Positive predictive value 91 75–98% 84 64–95%

95% CI: 95% confidence interval.a Seventy-five of 85 lesions were evaluated in this comparative analysis.

F-18 FDG PET in Patients with NSGCT/Kollmannsberger et al. 2357

ual masses at multiple localizations. Whereas, in 24 ofthese patients, the residual masses within each patientdisplayed a uniform clinical behavior (either regres-sion or progression), the residual masses in 4 patientsdeveloped differently from one another. PET indicateduniform behavior of all residual masses in 19 of thesepatients: Four patients had truly positive results, 10patients had truly negative results, and 5 patients hadfalse negative results. No false positive overall behav-ior was suggested by PET in patients with a uniformglucose uptake in all residual masses.

Comparison of PET Imaging with CT Scan/MRI for thePrediction of Therapy ResponseComparisons of PET imaging with CT scan/MRI per-formed at the same time interval after chemotherapywere available for all 45 patients, who had a total of 85residual lesions (Table 2). Overall, CT scan/MRI cor-rectly predicted viability in 58 residual lesions (68%).CT scan/MRI correctly indicated viable carcinoma/teratoma in 27 residual lesions (20 in patients withstable disease and 7 in patients with tumor progres-sion during chemotherapy). Fifteen of these lesionswere also PET positive; whereas, in 12 residual masses,PET results were considered negative by the reviewers.

In 31 residual lesions without progression duringfollow-up and/or with decreased contrast mediumuptake, CT scan/MRI correctly predicted a necroticresidual mass. Whereas only 28 of these lesions werealso PET negative, an increased F-18 FDG uptake wasfound in 3 lesions. Histologic examination after resec-tion of these three lesions showed inflammation intwo lesions and necrosis in the third lesion (seeabove).

CT scans/MRI after therapy were not able to pre-dict correctly the viability of 27 lesions (32%). Twenty-two of these residual masses regressed more than 50%without contrast medium enhancement on CT scan/MRI, indicating a nonviable lesion, although they stillprogressed during follow-up. In 14 of these lesions,PET results correctly predicted a viable residual mass.However, eight lesions also had false negative PETresults. Five lesions remained progression free orshowed necrosis on histologic examination, despitethe fact that CT scans showed no tumor lesion re-sponse. PET correctly assessed all five lesions as ne-crosis. Thus, the respective sensitivity and specificityof CT scans/MRI (stable disease or progressive diseasepredicting viable residual masses) were 55% and 86%.The positive and negative predictive values of CTscans/MRI were 84% and 58%, respectively (Table 2).

Comparison of PET with Serum Tumor Marker Declinefor the Prediction of Residual Mass ViabilityThe comparison of PET imaging with serum tumormarker decline during therapy included 42 patientswith a total of 78 lesions. Because three patients witha total of seven lesions never exhibited elevated mark-ers during the course of their disease, they were notincluded in the analysis. Residual mass viability waspredicted correctly by serum tumor marker declinealone in 27 patients (64%) with 56 lesions (72%). In 11patients with 27 lesions, tumor markers remained el-evated or even increased after therapy, and all of thesepatients failed treatment. In 8 of these patients with atotal of 20 residual lesions, increased F-18 FDG uptakewas found in at least one residual lesion after treat-ment. Conversely, PET falsely predicted a necrotic re-sidual mass in three patients with seven lesions. Tu-mor marker decline correctly indicated the absence ofviable tumor in the other 16 patients with 29 lesions.PET also was normalized in 27 of these 29 residualmasses yet remained positive at the time of restagingafter treatment in 2 lesions (2 patients). The decline (7patients with 9 lesions) or normalization (8 patientswith 13 lesions) of tumor markers indicated the pres-ence of necrosis in 15 other patients with a total of 22residual masses, but all of these patients experienceddisease progression. In comparison, PET correctlypredicted the viability in 9 patients with a total of 13lesions and falsely indicated the absence of viabletumor in the remaining 6 patients with 9 lesions.There was no false positive serum tumor marker ele-vation. The sensitivity and specificity using tumormarker elevation alone (based on the number of pa-tients) for the indication of residual tumor viabilitywere 42% and 100%, respectively. The positive predic-tive value was 100%, and the negative predictive valuewas 52% (Table 2).

Comparison of PET with the Combined Assessment byCT Scan, MRI, and Serum Marker DeclineFor the blinded comparison of PET results with thecombined assessment by CT scan/MRI plus serumtumor marker decline, the data from all 45 patientswith a total of 85 lesions were included (Table 4). Of 85residual lesions, viability was predicted correctly bythe combination of CT scan and MRI plus tumormarker decline in 71 masses (83%). Thirty-eight le-sions progressed or were stable on CT scans/MRI dur-ing therapy, with serum tumor markers remainingelevated or even increasing in these patients. Eitherthese patients experience disease recurrence duringfollow-up, or histology of the resected mass revealedviable carcinoma/teratoma. PET imaging performed

2358 CANCER May 1, 2002 / Volume 94 / Number 9

at the time of restaging after the end of therapy wasalso positive in 24 of these lesions, whereas it falselyindicated the absence of viable carcinoma/teratomain 14 residual lesions. CT scans/MRI findings as wellas tumor marker levels showed a complete or partialremission after chemotherapy in all 33 residual lesionsthat contained necrosis or remained in remission dur-ing follow-up. PET also was predictive in 30 of 33lesions yet remained markedly positive in 3 residuallesions at the time of restaging (2 with inflammationand 1 with necrosis; see above). However, in 14 resid-ual masses, the combination of CT scan/MRI and tu-mor marker decline falsely suggested a nonviablemass. In five of these lesions, a positive PET scan afterchemotherapy correctly predicted the presence of vi-able carcinoma/teratoma.

Evaluation of Residual Masses by PET Imaging inPatients with Primary Tumors Without TeratomatousElementsBecause it was shown that two of three histologicallyconfirmed, false negative PET findings were due to thepresence of mature teratoma, a separate evaluationthat excluded patients with primary tumors contain-ing teratomatous elements was performed. The resid-ual lesions from patients with primary tumors contain-ing teratomatous elements had a greater probability ofcontaining mature teratoma in residual lesions afterchemotherapy. Twenty-three patients with a total of48 residual lesions were assessable for this analysis(Table 5). Thirty-three lesions were assessed correctlyby PET: 15 lesions were assessed correctly as positive,and 18 lesions were assessed correctly as negative. Inthree patients, the PET image was considered positive,but the patients had a favorable outcome (see above):CT scans/MRI showed remission of these three le-sions. Moreover, there were still 12 residual masses

without an increased F-18 FDG uptake, yet these pa-tients experienced disease progression within the first6 months after therapy. CT scan/MRI correctly pre-dicted the presence of viable carcinoma/teratoma in 8of these 12 lesions; whereas, in the remaining 4 resid-ual masses, CT scans/MRI also produced false nega-tive results. Thus, with a sensitivity of 56% and aspecificity of 86% and with a negative predictive valueof 60% and a positive predictive value of 83%, thediagnostic efficacy of PET imaging did not improvewhen patients with primary tumors containing terato-matous elements were excluded from the analysis.Similar to the whole group of patients, quantificationof F-18 FDG uptake did not provide additional infor-mation (sensitivity, 59%; specificity, 79%; negativepredictive value, 61%; positive predictive value, 77%).

DISCUSSIONThe accurate assessment of residual masses after che-motherapy for patients with NSGCT remains a diag-nostic problem, because the differentiation betweenviable carcinoma, mature teratoma, or scar tissue/necrosis has been possible only after patients undergosurgical resection and histologic work-up. Patientswith residual masses that contain viable carcinomahave a worse prognosis compared with patients whohave residual necrosis only, even if the residualmasses have been resected completely.20,21 Becauseapproximately 50% of all residual masses contain ne-crosis only, many patients do not benefit frompostchemotherapy surgical resection.4,21–23 Althoughsurgical morbidity and mortality rates have been re-duced substantially during the past decade, the abilityto identify these patients with necrotic residual lesions

TABLE 4Comparison of Positron Emission Tomography with ComputedTomography/Magnetic Resonance Imaging Scan plus SerumTumor Markers (N � 45 patients; 85 residual lesions)

PET (no. oflesions)

CT/MRI and tumor markers (no. of lesions)

Correctpositive

Correctnegative Not correct Total

Correct positive 24 0 5 29Correct negative 0 30 3 33False positive 0 3 0 3False negative 14 0 6 20Total 38 33 14 85

CT/MRI: computed tomography/magnetic resonance imaging; PET: positron emission tomography.

TABLE 5Comparison of the Sensitivity, Specificity, and Negative and PositivePredictive Values of Positron Emission Tomography (Visualassessment) of All 85 Residual Masses with a Cohort of All ResidualMasses and of Patients Without Teratomatous Elementsin their Primary Histology (N � 48 lesions)a

Parameter

All masses

Without primaryteratomatoushistology

% 95% CI % 95% CI

Sensitivity 59 44–73% 56 35–75%Specificity 92 78–98% 86 64–97%Negative predictive value 62 48–75% 60 41–77%Positive predictive value 91 75–98% 83 59–96%

95% CI: 95% confidence interval.a The patients without teratomatous elements in their primary tumors had an extremely low risk of

developing mature teratomas postchemotherapy.

F-18 FDG PET in Patients with NSGCT/Kollmannsberger et al. 2359

by noninvasive methods would allow the avoidance ofunnecessary surgery.24 –26 However, there are no es-tablished criteria to identify patients in whom second-ary surgery can be avoided. Both CT scan as well MRIparameters have not been good predictors of the via-bility of residual masses.27–30 Other studies have in-vestigated clinical prognostic factors for the predictionof residual mass viability after chemotherapy. Clinicalcriteria, such as a residual mass measuring � 3 cm ingreatest dimension and several others, have been in-vestigated that decide in favor of adjunctive surgicalresection rather than observation.20,22,29,31–36 Combin-ing different criteria, Steyerberg et al. developed astatistical model for the prediction of histologic find-ings in residual retroperitoneal masses based on thefollowing factors: presence of teratomatous elementsin the primary tumor, prechemotherapy levels of se-rum tumor markers, size of the residual mass, andpercentage of shrinkage in greatest mass dimension. Agood discriminative ability was achieved for the pre-diction of necrosis, whereas the model only showedpoor discriminative power for the distinction betweencarcinoma and mature teratoma.22,34

It was the objective of the current study to deter-mine whether PET, as a method for the visualizationand quantification of regional glucose metabolismwithin the body, can be used for the evaluation ofresidual masses in patients with NSGCT. In addition,the value of PET imaging was compared with thecommonly used means of assessment, such as CTscans, MRI, and changes in serum tumor markerlevels.

The results of the current study show that a pos-itive PET image after treatment was a strong predictorfor the presence of viable carcinoma/teratoma, be-cause 9 of 10 patients with high F-18 FDG uptakelevels either had a histology of viable carcinoma/ter-atoma or had a recurrence within 6 months after treat-ment (positive predictive value, 91%). However, therewere also three false positive PET results. Histologyrevealed inflammation in two patients and necrosis inthe third patient. F-18 FDG is not a tumor specificagent, and it also may accumulate in tissue macro-phages. This phenomenon has been reported as asource of false positive PET examinations in patientswith malignant disease.14,37– 40 To reduce the rate offalse positive findings due to inflammation, the inter-val between the end of treatment and the PET exam-ination was required to be at least 3 weeks. In addi-tion, it was helpful to correlate the PET results withboth the morphologic information derived from theCT scan and with the decline of previously elevatedtumor markers. Conversely, a negative PET after theend of treatment does not allow one to avoid surgery,

because 37% of all negative PET masses either pro-gressed during 6 months of follow-up, or histologicexamination revealed mature teratoma. Semiquanti-tative calculation of glucose uptake by using SUVsincreased the sensitivity of PET from 59% to 68%,which, nevertheless, remains too low to alter definitivetreatment decisions. Moreover, neither visual judg-ment nor the calculation of SUVs allowed a reliabledifferentiation between mature teratoma and necro-sis, because SUVs for necrosis and teratoma were es-sentially the same. Similar results were obtained byStephens et al. and Cremerius et al., who found thatthe presence of mature teratoma was the most com-mon cause for false negative PET results in residuallesions from patients with NSGCT.14,38 In the currentstudy, the sensitivity and specificity of F-18 FDG PETalone for the prediction of residual mass viability were59% and 92%, respectively, which are similar to theresults obtained by other authors.14,38,39,41,42 Based onthe assessment of sensitivity and specificity of eitherthe radiologic response alone or the changes in tumormarker levels alone, the corresponding values were55% and 42% for sensitivity and 86% and 100% forspecificity, respectively. This indicates that no methodin itself appears sufficiently accurate to predict theviability of residual masses.

To reduce the likelihood of the presence of matureteratoma as the assumed main source of false negativefindings in residual masses, a separate analysis wasperformed that included only patients without terato-matous elements in the primary histology.14,38 How-ever, there were still 12 false negative PET results,suggesting that false negative findings may not onlyoccur due to the presence of mature teratoma (Ta-ble 5).

In contrast to the previous PET investigations inpatients with NSGCT by Stephens et al. and Nuutinenet al., which primarily included patients with retroper-itoneal metastases and a high proportion of postche-motherapy necrosis/teratoma, our study focused on aspecific high-risk population.38,39 Many patients hadadvanced disease with several metastatic sites, and allunderwent high-dose chemotherapy due to their over-all poor prognosis. Thus, viable carcinoma and tumorprogression after high-dose chemotherapy occurredin about 50% of patients.

Furthermore, to our knowledge, this is the firststudy to compare prospectively F-18 FDG PET withestablished criteria for the assessment of response inpatients with NSGCT. In addition, this study was per-formed with blinded readings of CT scans, MRIs, andPET images, with none of the diagnostic physiciansaware of the corresponding test results. Thus, the dif-ferences in sensitivity and specificity of all methods

2360 CANCER May 1, 2002 / Volume 94 / Number 9

can be compared directly. Overall, the diagnostic effi-cacy was similar for all three diagnostic methods.Compared with CT scans/MRI and tumor markers,PET seems to offer additional information. A potentialbenefit of PET was observed in patients with stabledisease or with disease in remission on CT scans/MRIand/or declining or normalized tumor markers, inwhom elevated F-18 FDG uptake correctly predictedthe presence of viable carcinoma/teratoma for fiveresidual lesions. Also, PET may offer valuable informa-tion in addition to CT scans in patients with serumtumor marker negative disease, particularly when tu-mor remission is diagnosed using CT scans. In con-trast, persisting or even increasing tumor markers aswell as progressive disease on CT scans/MRI duringchemotherapy are strong predictors for the presenceof viable carcinoma/teratoma, and PET does not pro-vide additional information in these patients.

PET also appears to yield important informationregarding patients with multiple residual masses. Themanagement of several residual masses of NSGCT atdifferent localizations within a single patient is partic-ularly difficult, because complete resection of all le-sions often is impossible. When PET results were uni-formly positive or negative for all lesions within asingle patient, the clinical behavior of the multipleresidual masses with respect to remission or progres-sion also was uniform in our study. In these patients,a uniformly positive PET appears to be a strong pre-dictor for viable carcinoma, because no false positivePET results were observed. Hartmann et al. founddissimilar histologic results at different anatomic lo-calizations in approximately 30% of patients who un-derwent resection at multiple localizations.4 In pa-tients who underwent resection of a necrotic lesion atone localization and had PET images that were uni-formly negative for all lesions, the probability of ne-crosis at the remaining nonresected lesions clearlywas � 90% (based on the negative predictive value ofa quantitatively assessed PET of 67%). Thus, PET mayadd important information in patients with multiple,incompletely resectable, residual masses.

However, the limitations of the current study alsomust be considered. The limited number of patientsincluded may have given rise to misleading results.Larger studies are necessary to confirm our results. Apossible second limitation may be the method used toconfirm the prediction of PET. Although it would havebeen ideal to have a histologic diagnosis of all tumorsat the time of PET assessment, this was feasible in only33% of all lesions. The results are based largely on thepatient follow-up over a 6-month period after finalassessment. However, because recurrent or progres-sive germ cell carcinoma in high-risk patients, it is

once present, usually occurs rapidly after the end oftreatment, the data obtained in this study still seemvalid. Considering the median follow-up of 27 months,we are confident that the results would not changesignificantly with longer follow-up.

In conclusion, PET performed in conjunction withconventional staging methods appears to offer addi-tional information for the prediction of residual masshistology in patients with NSGCT in several treatmentsituations. However, the limitations of PET imagingmust be considered. A negative PET after chemother-apy may represent either necrosis or mature teratoma,whereas a positive PET image is highly predictive forthe presence of viable carcinoma. Other useful indi-cations for PET imaging in patients with NSGCT in-clude the assessment of patients with multiple resid-ual lesions and patients with tumor marker negativedisease.

REFERENCES1. Einhorn LH. Treatment of testicular cancer: a new and im-

proved model. J Clin Oncol. 1990;8:1777–1781.2. Hartmann JT, Kanz L, Bokemeyer C. Diagnosis and treat-

ment of patients with testicular germ cell cancer. Drugs.1999;58:257–281.

3. Gerl A, Clemm C, Schmeller N, et al. Sequential resection ofresidual abdominal and thoracic masses after chemother-apy for metastatic non-seminomatous germ cell tumours.Br J Cancer. 1994;70:960 –965.

4. Hartmann JT, Schmoll HJ, Kuczyk MA, Candelaria M, Boke-meyer C. Postchemotherapy resections of residual massesfrom metastatic non-seminomatous testicular germ cell tu-mors. Ann Oncol. 1997;8:531–538.

5. Aprikian AG, Herr HW, Bajorin DF, Bosl GJ. Resection ofpostchemotherapy residual masses and limited retroperito-neal lymphadenectomy in patients with metastatic testicu-lar nonseminomatous germ cell tumors. Cancer. 1994;74:1329 –1334.

6. Wahle GR, Foster RS, Bihrle R, Rowland RG, Bennett RM,Donohue JP. Nerve sparing retroperitoneal lymphadenec-tomy after primary chemotherapy for metastatic testicularcarcinoma. J Urol. 1994;152:428 – 430.

7. Zon RT, Nichols C, Einhorn LH. Management strategies andoutcomes of germ cell tumor patients with very high humanchorionic gonadotropin levels. J Clin Oncol. 1998;16:1294 –1297.

8. Strauss LG, Conti PS. The applications of PET in clinicaloncology. J Nucl Med. 1991;32:623– 648.

9. Hawkins RA, Hoh C, Dahlbom M, et al. PET cancer evalua-tions with FDG [editorial; comment]. J Nucl Med. 1991;32:1555–15558.

10. Bangerter M, Moog F, Buchmann I, et al. Whole-body2-[18F]-fluoro-2-deoxy-D-glucose positron emission tomog-raphy (FDG-PET) for accurate staging of Hodgkin’s disease.Ann Oncol. 1998;9:1117–1122.

11. Stumpe KD, Urbinelli M, Steinert HC, Glanzmann C, Buck A,von Schulthess GK. Whole-body positron emission tomog-raphy using fluorodeoxyglucose for staging of lymphoma:effectiveness and comparison with computed tomography.Eur J Nucl Med. 1998;25:721–728.

F-18 FDG PET in Patients with NSGCT/Kollmannsberger et al. 2361

12. Partridge S, Timothy A, O’Doherty MJ, Hain SF, Rankin S,Mikhaeel G. 2-Fluorine-18-fluoro-2-deoxy-D glucose positronemission tomography in the pretreatment staging ofHodgkin’s disease. Influence in patient management in a sin-gle institution. Ann Oncol. 2000;11:1273–1279.

13. Cremerius U, Wildberger JE, Borchers H, et al. Does positronemission tomography using 18-fluoro-2-deoxyglucose im-prove clinical staging of testicular cancer? Results of a studyin 50 patients. Urology. 1999;54:900 –904.

14. Cremerius U, Effert PJ, Adam G, et al. FDG PET for detectionand therapy control of metastatic germ cell tumor. J NuclMed. 1998;39:815– 822.

15. Bokemeyer C, Harstrick A, Beyer J, et al. The use of dose-intensified chemotherapy in the treatment of metastaticnonseminomatous testicular germ cell tumors. German Tes-ticular Cancer Study Group. Semin Oncol. 1998;25:24 –32.

16. Beyer J, Bokemeyer C, Rick O, et al. Salvage treatment ingerm-cell tumors using taxol, ifosfamide, cisplatin (TIP) fol-lowed by high-dose carboplatin, etoposide, and thiotepa(HDCET): first results. Proc Am Soc Clin Oncol. 1998;17:322a.

17. International Germ Cell Cancer Collaborative Group. Inter-national Germ Cell Consensus Classification: a prognosticfactor-based staging system for metastatic germ cell can-cers. J Clin Oncol. 1997;15:594 – 603.

18. Miller AB, Hoodgstraten B, Staquet M, Winkler A. Reportingresults of cancer treatment. Cancer. 1981;47:207–211.

19. Leskinen KS, Nagren K, Lehikoinen P, Ruotsalainen U, Joen-suu H. Uptake of 11C-methionine in breast cancer studiedby PET. An association with the size of S-phase fraction. Br JCancer. 1991;64:1121–1124.

20. Hartmann JT, Candelaria M, Kuczyk MA, Schmoll HJ, Boke-meyer C. Comparison of histological results from the resec-tion of residual masses at different sites after chemotherapyfor metastatic non-seminomatous germ cell tumours. Eur JCancer. 1997;33:843– 847.

21. Stenning SP, Parkinson MC, Fisher C, et al. Postchemo-therapy residual masses in germ cell tumor patients: con-tent, clinical features, and prognosis. Medical ResearchCouncil Testicular Tumour Working Party. Cancer. 1998;83:1409 –1419.

22. Steyerberg EW, Keizer HJ, Fossa SD, et al. Prediction ofresidual retroperitoneal mass histology after chemotherapyfor metastatic nonseminomatous germ cell tumor: multivar-iate analysis of individual patient data from six studygroups. J Clin Oncol. 1995;13:1177–1187.

23. Fossa SD, Aass N, Ous S, et al. Histology of tumor residualsfollowing chemotherapy in patients with advanced non-seminomatous testicular cancer. J Urol. 1989;142:1239 –1242.

24. Donohue JP, Rowland RG. Complications of retroperitoneallymph node dissection. J Urol. 1981;125:338 –340.

25. Donohue JP, Foster RS, Rowland RG, Bihrle R, Jones J, GeierG. Nerve-sparing retroperitoneal lymphadenectomy withpreservation of ejaculation. J Urol. 1990;144:287–291.

26. Doerr A, Skinner EC, Skinner DG. Preservation of ejacula-tion through a modified retroperitoneal lymph node dissec-tion in low stage testis cancer. J Urol. 1993;149:1472–1474.

27. Stomper PC, Kalish LA, Garnick MB, Richie JP, Kantoff PW.CT and pathologic predictive features of residual mass his-

tologic findings after chemotherapy for nonseminomatousgerm cell tumors: can residual malignancy or teratoma beexcluded? Radiology. 1991;180:711–714.

28. Sagalowsky AI, Ewalt DH, Molberg K, Peters PC. Predictorsof residual mass histology after chemotherapy for advancedtestis cancer. Urology. 1990;35:537–542.

29. Donohue JP, Rowland RG, Kopecky K, et al. Correlation ofcomputerized tomographic changes and histological find-ings in 80 patients having radical retroperitoneal lymphnode dissection after chemotherapy for testis cancer. J Urol.1987;137:1176 –1179.

30. Steyerberg EW, Keizer HJ, Sleijfer DT, et al. Retroperitonealmetastases in testicular cancer: role of CT measurements ofresidual masses in decision making for resection after che-motherapy. Radiology. 2000;215:437– 444.

31. Javadpour N, Young JD. Predictors of residual mass requir-ing surgical resection after chemotherapy of Stage III testic-ular cancer. A prospective study. Urology. 1992;40:7– 8.

32. Matsuyama H, Yamamoto N, Sakatoku J, et al. Predictivefactors for the histologic nature of residual tumor mass afterchemotherapy in patients with advanced testicular cancer.Urology. 1994;44:392–398.

33. Steyerberg EW, Keizer HJ, Stoter G, Habbema JD. Predictorsof residual mass histology following chemotherapy for met-astatic non-seminomatous testicular cancer: a quantitativeoverview of 996 resections. Eur J Cancer. 1994;30A:1231–1239.

34. Steyerberg EW, Gerl A, Fossa SD, et al. Validity of predictionsof residual retroperitoneal mass histology in nonseminoma-tous testicular cancer. J Clin Oncol. 1998;16:269 –274.

35. Toner GC, Panicek DM, Heelan RT, et al. Adjunctive surgeryafter chemotherapy for nonseminomatous germ cell tu-mors: recommendations for patient selection. J Clin Oncol.1990;8:1683–1694.

36. Debono DJ, Heilman DK, Einhorn LH, Donohue JP. Deci-sion analysis for avoiding postchemotherapy surgery in pa-tients with disseminated nonseminomatous germ cell tu-mors [see comments]. J Clin Oncol. 1997;15:1455–1464.

37. Strauss LG. Fluorine-18 deoxyglucose and false-positive re-sults: a major problem in the diagnostics of oncologicalpatients. Eur J Nucl Med. 1996;23:1409 –1415.

38. Stephens AW, Gonin R, Hutchins GD, Einhorn LH. Positronemission tomography evaluation of residual radiographicabnormalities in postchemotherapy germ cell tumor pa-tients. J Clin Oncol. 1996;14:1637–1641.

39. Nuutinen JM, Leskinen S, Elomaa I, et al. Detection ofresidual tumours in postchemotherapy testicular cancer byFDG-PET. Eur J Cancer. 1997;33:1234 –1241.

40. Flamen P, Stroobants S, Van Cutsem E, et al. Additionalvalue of whole-body positron emission tomography withfluorine-18-2-fluoro-2-deoxy-D-glucose in recurrent colo-rectal cancer. J Clin Oncol. 1999;17:894 –901.

41. Dohmen BM, Bares R, Gromius R, et al. Improved staging ofgerm cell cancer after chemotherapy by FDG-positron-emission-tomography (PET). Eur J Nucl Med. 1995;22:804.

42. Harms W, Bares R, Kamps H, et al. Therapy control ofmetastatic testicular carcinoma with FDG PET. J Nucl Med.1995;36:198.

2362 CANCER May 1, 2002 / Volume 94 / Number 9