The Utility of ¹⁸F-FDG PET for Suspected Recurrent Non–Small Cell Lung Cancer After Potentially Curative Therapy: Impact on Management and Prognostic Stratification

Patient Eligibility

Consecutive patients referred for clinical evaluation of suspected relapse of NSCLC between November 1996 and December 1998 were prospectively entered into a database. Patients were eligible for this study if they had a confirmed histologic diagnosis and had been treated with curative intent >6 mo before referral for a follow-up PET scan. Patients were excluded from analysis if they already had confirmed systemic metastatic disease based on either biopsy or unequivocal conventional imaging appearances, unless that disease was amenable to aggressive local therapy. FDG PET was performed for clinical indications in all patients. Follow-up of clinical outcome was approved by our institutional review board.

PET Scan Acquisition and Processing

PET scans were acquired on a PENN-PET 300-H scanner (UGM Medical Systems, Inc., Philadelphia, PA). Patients fasted for a minimum of 4 h before the scan. Emission data were acquired 1 h after intravenous administration of 74–120 MBq FDG. The scan routinely incorporated the lower neck, thorax, and upper abdomen, consistent with the usual extent of staging CT scans. If additional but unconfirmed sites of metastatic disease were suggested on conventional staging, the scan was extended to include the appropriate regions. PET was not used to evaluate the brain. The performance characteristics of this scanner and processing methods have been described previously (9–11).

The emission scan was reconstructed using the iterative technique of ordered-subset expectation maximization. Image sets, both with and without measured attenuation correction, were reviewed on a computer console using a display program that allows interactive review of orthogonal slices. An experienced nuclear medicine physician generated a clinical report after reviewing previous imaging results and clinical information. This included, in almost all cases, direct correlation with the recent CT scan. When the CT scan was unavailable, the report of the CT investigation was obtained. Standard uptake values were not routinely calculated. Once issued, the PET report was entered into the database and was not reinterpreted in the light of subsequent clinical information.

Determination of Disease Extent

The medical record was reviewed to determine the disease stage at initial diagnosis according to the 1997 update of the international staging system for lung cancer (12), and the primary treatment delivered subsequent to this diagnosis was recorded.

The staging schema used for classification of newly diagnosed NSCLC was believed to be inappropriate for the restaging setting. Rather, the patients were classified according to the extent of apparent relapse based on information available on conventional staging investigations performed before the PET scan (pre-PET extent category) and based on incorporation of the PET information with the previous information (post-PET extent category). The pre-PET extent category included biopsy results when they were available before the PET scan. The PET results were not interpreted in a masked manner; thus, this study evaluated the incremental, not the independent, information provided by PET.

Patients were classified as negative if imaging showed no abnormality suggestive of recurrence or, in the case of pre-PET evaluation, if biopsy of a residual structural abnormality had negative findings. Patients with abnormality confined to the previous primary tumor site and potentially suitable for salvage surgery or radical radiotherapy were classified as having local relapse. Patients with abnormalities involving intrathoracic lymph nodes and potentially amenable to radical radiotherapy were classified as having limited nodal relapse. Patients with intrathoracic lymph node abnormalities too extensive for radical radiotherapy or with lymphadenopathy beyond the thorax were classified as having extensive nodal relapse. Finally, patients with abnormalities involving major organs or the skeleton were classified as having systemic relapse.

Conventional restaging routinely involved CT with dynamic contrast medium, unless contraindicated because of contrast allergy. Bone scanning was performed only for patients with musculoskeletal symptoms or elevated alkaline phosphatase levels. CT or MRI of the brain was performed only for patients with clinically suspected cerebral metastasis. On CT scans, symptomatic or progressively enlarging lobular masses were considered suggestive of residual disease unless found to be negative for disease on histologic examination, whereas bandlike abnormalities were considered likely to reflect scarring. Mediastinal and hilar lymph nodes were regarded as positive for tumor on CT if they were >1 cm in maximum transverse diameter and larger than on the baseline evaluation. Histologically unconfirmed systemic metastases found on conventional staging were considered positive for disease if active treatment was planned on their basis but equivocal if, despite their presence, the referring clinician was still considering curative salvage treatment of suspected more localized relapse.

The post-PET extent of disease relied fully on the results of PET when these were discordant with other imaging findings, even when the results of prior biopsy had been negative. Focal areas of increased activity of greater intensity than mediastinal soft tissues on attenuation-corrected images were considered positive for active disease. However, diffuse changes in the lung or pleural reflections conforming to a prior radiotherapy treatment volume were generally reported as negative for active disease. These appearances are common after high-dose radiotherapy and are believed likely to be inflammatory. When a baseline PET scan was available, it was used for comparisons of the extent and intensity of ongoing abnormalities at previously documented sites of PET abnormality.

In addition to extent, the specific sites of abnormality indicating relapsed disease were recorded for both conventional imaging and PET. These sites were used for validation. Both the status of relapse and the distribution of disease were confirmed whenever possible by histologic examination. When histologic confirmation was unavailable, evidence of progression on serial imaging within 6 mo was considered to indicate a site of disease relapse, whereas no evidence of progression after at least 12 mo of follow-up was considered to confirm absence of active disease at that site. Patients who died without clear evidence of disease progression on conventional imaging were considered not assessable. Lack of imaging evidence of progressive disease after delivery of radical or high-dose palliative radiotherapy to a suspected localized recurrence was not considered to confirm this as a site of relapse unless relapse had been histologically confirmed before treatment began.

Assessment of Impact

Referring clinicians are asked to prospectively record a pre-PET management plan on our routine clinical request form. This plan was entered in a database. If not explicitly recorded on the referral request, treatment intent was categorized by an oncologist experienced in lung cancer management, who synthesized the available pre-PET clinical and imaging information. For 20 patients, a pre-PET management plan had not been recorded on the request form, and we relied on the patient’s pre-PET medical record. The post-PET plan and intent were determined from the medical record or contact with the referring clinician and in all but 1 patient reflected the treatment actually delivered.

The impact on management was considered high when the treatment intent or modality was changed (e.g., from palliative to curative treatment or from surgery to radiotherapy). Although not immediately altering treatment delivery, the change of a patient from expectant palliative treatment for suspected relapse to observation on the basis of negative PET findings was classified as of high impact because of the potential for psychologic reassurance of the patient from such a change. The impact was recorded as medium when the method of treatment delivery was changed (e.g., a change in radiation treatment volume). When the PET results did not indicate a need for change, the impact was considered to be low. PET was considered to have had no impact when the management chosen conflicted with post-PET disease extent and was believed to be inappropriate on the basis of a synthesis of all available information.

Patients within any given disease category could have a range of impact outcomes. For example, patients whose CT findings showed extensive local recurrence amenable only to palliative treatment could be converted to curative salvage surgery if PET showed a small recurrence, even though both the pre-PET extent and the post-PET extent would be categorized as localized recurrence. Alternatively, discordance between extent categories did not necessarily affect management. For example, the palliative radiotherapy planned for some patients with symptomatic local recurrence was still considered appropriate even though additional but asymptomatic sites of disease were detected.

Follow-Up

After treatment, progress updates were obtained from the medical record, family physician, or treating oncologist. When relevant, details of the date and cause of death were obtained. The disease status at the time of death was recorded. For patients treated with curative intent, the site or sites of first progression were recorded.

Statistical Methods

Survival was measured from the date of the restaging PET scan to the date of death from any cause. Survival estimates were obtained using the Kaplan–Meier product-limit method and compared using the Cox proportional hazards regression model. The prognostic significance of individual factors has been summarized using hazard ratios representing the relative death rate for a given group relative to a baseline group. The 95% confidence interval (CI) is given for the main results. Unless otherwise indicated, 2-sided probability values are reported with no adjustment for multiple comparisons.

Patient Population

In all, 63 patients met the eligibility criteria (41 men, 22 women; mean age ± SD, 66 ± 9 y). The histologic cancer types were squamous cell carcinoma in 36 patients; adenocarcinoma in 19 patients; large cell carcinoma in 7 patients; and non–small cell carcinoma, not otherwise specified, in 1 patient. The primary curative treatment had been surgery in 18 patients (29%), radical radiotherapy with or without concurrent chemotherapy in 33 patients (52%), and chemoradiotherapy plus surgery in 12 patients (19%).

Detection of Relapse

Relapse involving at least 1 site was confirmed in 42 (67%) of 63 patients by death from progressive disease (n = 30), by pathologic examination (n = 5), or by serial imaging (n = 7). PET was positive for disease in 41 (98%) of these patients. The 1 false-negative PET result was for a study that was limited because of the patient’s claustrophobia and that did not include the site of confirmed recurrence. Conventional staging suggested relapse in all 42 of these patients, yielding a sensitivity of 100%. In 1 patient whose CT scan showed a suspected isolated systemic metastasis but whose PET findings were negative, a small metachronous primary bronchial malignancy subsequently developed and was detected on follow-up bronchoscopy. The site of structural abnormality did not progress, and the PET result was considered true-negative for systemic relapse at this site but incorrect for extent because the developing metachronous primary tumor was missed.

At the study censor date and with a minimum follow-up of 12 mo, 17 patients (27%) had no evidence of relapse. The PET findings were negative in 14 of these patients, giving a specificity of 82%. All but 2 had abnormal CT findings suggestive of thoracic relapse, giving a specificity of only 12% for conventional imaging. Both of the patients without abnormal imaging findings had other clinical evidence of relapse (progressive loss of weight in one and a recurrent pleural effusion in the other). Three PET studies had false-positive findings, all in patients also incorrectly thought to have intrathoracic relapse on the basis of CT findings. One of these scans, seen early in our experience, clearly showed postirradiation changes and was incorrectly interpreted. Because of the discordance in the distribution of abnormality seen on CT compared with that seen on PET, ongoing observation rather than planned palliative radiotherapy was instituted for this patient when the hemoptysis that had led to her referral settled. The other 2 false-positive PET scans showed abnormalities confined to the neck that, in retrospect, may have reflected asymmetric uptake in strap muscles. Neither patient had active treatment, and they remained free from progression of disease in the neck after 707 and 840 d of follow-up. PET correctly excluded suspected intrathoracic relapse in both these patients. In 4 patients, the disease status could not be evaluated because of death from other illnesses (2 patients) or because of treatment introduced on the basis of imaging findings that were not pathologically confirmed (2 patients).

Comparison of Disease Extent as Evaluated Conventionally and by PET

The relapse extent determined by PET was different from the conventionally determined extent in 44 (70%) of 63 patients (95% CI, 57%–81%). Overall, 33% of patients had less extensive disease and 37% had more extensive disease than conventional evaluation suggested (Table 1). Within each category were also some differences in the distribution of active disease that had an impact on patient management. Overall, conventional staging was correct in its assignment of extent category in 12 (24%) of 49 evaluable patients, whereas PET was correct in 42 (86%) of 49 patients.

The concordance between the pre- and post-PET evaluations of the extent of disease relapse is presented in Table 2. Validation of the extent of disease as assessed by conventional evaluation techniques and PET is also displayed. For 44 patients, the pre- and post-PET categories of the extent of disease relapse differed. For 36 (82%) of these patients, the accuracy of the extent of disease could be validated. In 31 (86%) of 36 patients, follow-up has shown that PET was correct, whereas for only 1 patient (3%) was the pre-PET classification correct (P < 0.0001). This patient was the one with claustrophobia. For 4 other patients, the extent of disease as assessed by both PET and CT appears to have been incorrect. For 12 patients, the extent category was concordant on conventional and PET imaging but either the distribution of disease or the level of confidence in the imaging result was sufficient to change management after PET. Similarly, there were 5 patients for whom the extent category differed but for whom management was not altered because the treatment planned was still deemed to be appropriate even after PET.

Impact of PET Results on Patient Management

The PET results had a high impact on management in 40 patients (63%). The pre- and post-PET treatment plans for these patients are summarized in Table 3. Of note, 6 patients were changed from curative to palliative care after PET showed more extensive disease than had been suspected on conventional evaluation. Additionally, 8 patients for whom active palliative therapy was being considered and 1 patient for whom an invasive diagnostic procedure was planned received no active therapy after negative results were found on PET. Thus, 9 patients were spared the morbidity of these interventions. An example is shown in Figure 1. Another 6 patients whose disease was considered to have relapsed but for whom active palliative treatment was not planned unless their symptoms increased also had negative PET findings. Because of the potential psychologic impact on patients of being cleared of active disease, the impact of PET in these instances was also classified as high.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Hemoptysis and residual mass 9 mo after radical chemoradiotherapy for left upper lobe NSCLC suggested local relapse despite partial CT response between baseline (A) and posttreatment (B) scans. Representative transaxial (C), sagittal (D), and coronal (E) planes of PET scan showed low-grade pleural changes consistent with postradiotherapy effects but no evidence of recurrent disease. Palliative radiotherapy was cancelled, and patient remained well >12 mo after PET without evidence of active disease.

PET had a medium impact on management in 8 patients (13%). Management changes in these patients primarily included changes in radiation treatment volume. PET had a low impact (i.e., did not change the planned management) on 13 patients (21%), for whom either the relapse extent was concordant with that found on conventional restaging (7 patients) or the documentation of a different distribution of disease did not alter the appropriateness of the planned treatment (6 patients). Although PET had shown more extensive disease in 5 of 6 of these patients, the planned palliative treatment was considered appropriate because the additional sites of disease were asymptomatic at that time. The remaining patient had PET findings showing local recurrence after previous resection of a T2 N2 tumor. Although the mediastinum was negative for tumor on PET, we believed that the residual enlarged nodes found in the mediastinum on CT scanning needed to be included in the radical radiotherapy volume.

PET had no impact on management in 2 other patients (3%) who had discordant PET and CT findings. These 2 patients received active treatment despite negative PET findings and remain free from progression at >19 mo. One underwent resection of a lung mass and associated nodes with pathologic confirmation of a complete response to previous chemoradiation, whereas the other received radical radiotherapy of a chest wall mass that was negative for tumor on PET.

The post-PET management plan differed from the treatment actually delivered in 1 patient who died before commencing palliative radiotherapy. This patient was found to have local relapse on both pre-PET and post-PET assessment, and the treatment decision was therefore unaffected by the PET result. Another patient thought to have only local relapse on CT was found to have possible right peritracheal nodal disease also on PET and was to undergo palliative radiotherapy, but because her hemoptysis settled, this therapy was not subsequently delivered. Definite recurrent disease had not been evident during a follow-up of 722 d, and the findings of both studies are therefore considered to have been false-positive. This patient was the one mentioned earlier with the false-positive result related to suspected postradiotherapy inflammatory change.

Confirmation of PET Results in Patients with High- and Medium-Impact Studies

Of the 40 patients in the high-impact group, confirmation of the true disease status was assessable by histology or by serial imaging in 37 (93%). Of these 37 patients, PET was shown to be correct in 33 (89%). The 4 instances in which PET was incorrect included the study that was false-negative but technically limited by claustrophobia, the 2 studies that showed false-positive lower neck uptake, and a study in which PET correctly excluded systemic metastasis but failed to detect a metachronous lung primary. All 4 assessable cases in which PET modified delivery of a planned therapy (medium impact) were confirmed.

Prediction of Survival by PET

Survival data were analyzed with a closeout (study censor) date of July 1, 1999. Thus, the potential follow-up ranged from 6 to 31 mo (0.5–2.6 y), with a median of 19 mo (1.5 y). All 63 patients entered into the study had a known status at the closeout date. All patients with negative PET findings were followed for a minimum of 12 mo after the scan (range, 371–849 d).

Estimated survival at 1 y was 62% (95% CI, 50%–75%), and estimated survival at 2 y was 40% (95% CI, 22%–59%). Neither the stage at initial diagnosis nor the primary treatment delivered was associated with survival in this selected patient cohort. Positive PET findings were, however, associated with a poorer prognosis than were negative findings (Fig. 2). Further, the extent of relapse assessed after the PET scan was a highly significant prognostic factor (P < 0.0001) (Table 4; Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Graph shows strong prognostic stratification by whether PET was positive or negative for relapse in 63 consecutive patients in whom relapse was suspected on basis of conventional imaging findings (n = 61) or clinical features (n = 2).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Graph shows strong relationship between apparent extent of disease on PET scanning and survival in 63 consecutive patients in whom relapse was suspected on basis of conventional evaluation. This relationship remained significant after adjusting for pre-PET evaluation of extent of disease relapse or for treatment subsequently delivered. ENR = extensive nodal relapse; LNR = limited nodal relapse; LR = local relapse; SR = systemic relapse.

There was also a significant difference in the prognosis of patients given no treatment, curative treatment, palliative treatment, or supportive treatment on the basis of PET (P = 0.011) (Table 5). Of note, the group receiving curative treatment had similar survival to the group not treated after having negative scan findings.