A pathologic complete response (pCR) and minimal residual disease (pMRD) after preoperative chemotherapy (PCT) for early stage or locally advanced breast cancer (BC) correlates with a good prognosis.
A pathologic complete response (pCR) and minimal residual disease (pMRD) after preoperative chemotherapy (PCT) for early stage or locally advanced breast cancer (BC) correlates with a good prognosis.
Patients who received from 6 to 8 cycles of PCT for BC were monitored by 18F-2-fluoro-2-deoxy-D-glucose positron emission tomography (18F-FDG-PET), and the maximal standardized uptake value (SUVmax) was calculated at baseline, after 2 cycles, after 4 cycles, and at the end of PCT. SUVmax percentage changes (Δ-SUV) were compared with the pathologic response rate. Patients who had a pCR or pMRD in the tumor and an absence of cancer cells in ipsilateral axillary lymph nodes were defined as having obtained an optimal pathologic response (pR), whereas all the other conditions were classified as a pathologic nonresponse (pNR).
Of 34 patients, 7 (21%) achieved a pR (3 patients had a pCR, and 4 patients had pMRD). After the second cycle, the Δ-SUV threshold with optimal negative predictive value to predict a pR was 50%. Twenty-six patients (76%) had a Δ-SUV >50%, including all 7 patients who had a pR and 19 patients who had a pNR. Conversely, all 8 patients who had a Δ-SUV ≤50% had a pNR. All 8 of those patients had estrogen recepetor-positive tumors.
Early evaluation of metabolic response by 18F-FDG-PET during PCT was able to identify 30% of patients, all with estrogen receptor-positive tumors, who would not obtain pR after completion of chemotherapy program. Cancer 2010. © 2010 American Cancer Society.
Preoperative chemotherapy (PCT) is the standard treatment for locally advanced breast cancer and is an optional treatment modality for women who would prefer conservative surgery when it is not immediately achievable. Indeed, in the majority of patients, PCT can induce a reduction in the tumor volume that makes conservative surgery feasible. Currently, the achievement of a pathologic complete response (pCR) has emerged as the primary endpoint of PCT in patients who have operable breast cancer. In fact, a pCR is associated with a favorable prognosis, because patients who achieve a pCR have a far lower risk of subsequent recurrence and a better overall survival than patients who have residual invasive tumor at the time of surgery.1-3 These results support pCR after PCT as a prognostic factor and as a potentially useful short-term endpoint for testing new therapies.4 However, application of the pCR definition across clinical trials of PCT is not uniform. The current consensus is that the preferred definition of pCR is the absence of both residual invasive cancer within the breast and in the ipsilateral axillary lymph nodes.5 Because the presence of residual disease after neoadjuvant treatment includes a broad range of actual responses, from a near pCR to frank disease resistance, a composite index of the residual cancer burden after PCT recently was proposed.6 Patients with minimal residual disease had the same 5-year prognosis as those with pCR irrespective of the type of neoadjuvant chemotherapy administered, whether they received adjuvant hormone therapy, or the pathologic residual disease stage. Therefore, the combination of pCR and minimal residual disease expands the subset of patients who can benefit substantially from PCT.
The availability of an early predictive test of the pathologic response to PCT could allow for the better tailoring of individual patient treatment, including treatment interruption if there is limited activity. In this instance, stopping treatment also could avoid unwarranted side-effects. Because the standard duration of PCT varies between 4 and 8 cycles, the optimal timing to obtain a response prediction appears to be after the early cycles, at a stage when changing the treatment strategy is still a viable and reasonable option. Until now, early changes at physical examination and breast imaging7, 8 have exhibited a low correlation with the ultimate pathologic assessment of residual disease and are not considered useful in predicting pathologic response in clinical practice.
Over the last few years, is has been reported that 18F-2-fluoro-2-deoxy-D-glucose (FDG)-positron emission tomography (FDG-PET) is capable of predicting the pathologic response at an early stage during PCT for early stage and locally advanced breast cancer9-15 by virtue of the precocious changes in tumor metabolism that can be demonstrated by FDG-PET. However, the heterogeneity of methodologies and, to some extent, the findings across the studies currently make the role of early FDG-PET evaluation in clinical practice unclear.
In the current study, we investigated the value of FDG-PET scan monitoring to predict the pathologic response after PCT, and we paid particular attention to the optimal timing of early evaluation and its correlation with the standard biopathologic tumor profile. This study is a part of a broader study project on PCT for early or locally advanced breast cancer that started at our institution in 2004 (the Arianna 1 Project).
Consecutive patients with newly diagnosed early and locally advanced breast cancer (tumor [T] classification T2-T4, lymph node [N] classification N0-N3, and metastasis [M] classification M0 according to the International Union Against Cancer TNM classification16) and who had adequate medical status for chemotherapy and breast surgery were eligible for the study. Patients who had M1 oligometastatic breast cancer also were eligible provided that the therapeutic program included breast surgery postchemotherapy. The histologic diagnosis of breast cancer was confirmed by core-needle biopsy. Physical examination, mammography, breast ultrasound, breast magnetic resonance imaging, chest and abdomen computed tomography (CT) scans, bone scans, and FDG-PET/CT studies were obtained at baseline in all patients. Physical examination, mammography, breast ultrasound, breast magnetic resonance imaging, and 18F-FDG-PET/CT studies were repeated every other cycle and before surgery. Written informed consent was provided by all patients, and the study was approved by the local ethics committee.
Patients received from 6 to 8 cycles of anthracycline-based and taxane-based PCT every 21 days. Surgery (including axillary lymph node dissection) was undergone after patients completed PCT and consisted of a conservative procedure whenever possible. Patients who underwent breast-conserving surgery subsequently received radiotherapy.
Fresh surgical specimens (breast and axillary lymph nodes) obtained postchemotherapy were evaluated by an experienced breast pathologist (D.S.). The surgically removed specimen was thinly sliced and fixed immediately in neutral buffered formalin. If the tumor was detected macroscopically, then its size was measured; if no macroscopic lesion was identified, t hen wide slides of all breast tissue were sampled guided by radiographic images of the specimen, according to Proceedings of the 2003 Consensus Conference on Neoadjuvant Chemotherapy in Carcinoma of the Breast held in Philadelphia.17 The whole tumor area or “tumor bed” was taken for histologic examination. Handling sections were fixed according to standard procedures in 10% neutral buffered formaldehyde and embedded in paraffin. Then, 5-μm-thick sections were prepared and stained with hematoxylin and eosin. The histologic response to chemotherapy was assessed by the same pathologist using a grading system according to the Miller-Payne classification described by Smith et al. and Ogston et al.3, 18 This grading system assesses the degree of reduction in tumor cellularity obtained by comparing the tumor cellularity a pretreatment core biopsy with the tumor cellularity observed in the residual breast tumor tissue removed at surgery. The tumors were graded on a scale from 1 to 5 as follows: tumor regression grade (TRG) 1, no response to treatment; TRG 2, <30% reduction in cellularity; TRG 3, from 30% to 90% reduction in cellularity; TRG 4, >90% and <100% reduction in cellularity; and TRG 5, a complete response with no residual tumor. Histologic grading of the pathologic response also was evaluated concomitantly on the axillary lymph nodes as the lymph node regression grade (NRG). In particular, axillary lymph node status (ALN) was classified according to a 4-grade scale as follows: A, true ALN negative; B, ALN positive, no therapeutic response; C, ALN positive, evidence of a partial pathologic response; and D, ALN previously positive but converted to lymph node negative after chemotherapy. TRG and NRG served as the reference standards for the evaluation of treatment response with FDG-PET. Two regression groups based on pathologic response were defined: optimal responders (TRG 4 and 5 with NRG A and D) and nonresponders (TRG 1, 2, and 3 with no NRG).
18F-FDG-PET/CT studies were repeated after each PCT cycle in the first 10 patients. From those preliminary data, we observed that the main metabolic changes in tumor localization occurred after the second course of PCT19; therefore, subsequent patients underwent PET/CT scans at baseline, after the second and fourth cycles of PCT, and at the end of all PCT treatment, before surgery.
18F-FDG was produced in the local radiopharmacy by using standardized synthesis techniques. All imaging studies were obtained with a hybrid PET/CT scanner (Discovery LS; GE Medical System, Waukesha, Wis). 18F-FDG-PET/CT studies were obtained according to standard procedure.20 The patients fasted for at least 6 hours, and glycemia levels were monitored in patients with diabetes. Patients received an intravenous injection of 18F-FDG 5.3 megabecquerels per kilogram (MBq/Kg), and imaging acquisition started from 60 to 70 minutes after the radiotracer injection. Patients were asked to urinate immediately before image acquisition to minimize bladder activity.
18F-FDG-PET/CT emission data were acquired for 4 minutes per bed position; then, 35 images were reconstructed after a correction for CT data nonuniform attenuation. The CT parameters were 120 kV, 60 mA, 0.8 seconds per tube rotation, and 30-mm bed speed per gantry rotation (multislice technology enabled the acquisition of slices from 4 mm to 5 mm thick per tube rotation). Thus, we demonstrated that CT does not release a high radiation dose to the patient but still is efficient for distinguishing different tissues with good spatial resolution. CT images subsequently were merged with PET/CT images to obtain an accurate localization of 18F-FDG-PET/CT findings.
The PET/CT scan was read by 2 nuclear medicine physicians, and the report was written by consensus: Any area of increased 18F-FDG uptake was taken into account for evaluation. The metabolic activity of each pathologic lesion was measured using the maximum standardized uptake value (SUVmax) method. The SUVmax was calculated on the basis of the following formula: tissue concentration (MBq/g)/injected dose (MBq)/body weight (g). The SUVmax was measured at the breast tumor and at the ipsilateral axilla level before and during treatment.
The SUVmax from PET/CT scans obtained after the second, fourth, sixth, or eighth courses of CT were compared with the SUVmax from PET/CT scan obtained at baseline, which was considered as the reference value, to obtain the percentage reduction in SUVmax (Δ-SUV). A PET/CT metabolic response was considered complete when tumor SUVmax during treatment became equal to the normal surrounding tissue uptake (usually values <2).
To identify an optimal threshold value of Δ-SUV capable of separating patients who had an optimal pathologic response (pR) and patients who had a pathologic nonresponse (pNR), a receiver operating characteristic analysis was performed after 2 cycles of PTC using STATA software (version 9; StataCorp, Chicago, Ill). The area under the curve with 95% confidence interval, standard error, sensitivities, and specificities were assessed. The test was considered positive when the Δ-SUV value was greater than the cutoff value, and vice versa. Statistical analyses were performed by running SPSS version 13.0 for Windows (SPSS Inc., Cary, NC) on a PC. Two-tailed P values <.05 were considered statistically significant.
Thirty-four patients with newly diagnosed breast cancer, whose principal characteristics are listed in Table 1, were enrolled. Twenty-eight patients had localized breast cancer according to standard staging evaluation. An additional 6 patients had oligometastatic disease (2 patients had a single bone lesion, 1 patient had 3 liver lesions, 1 patient had a single liver lesion, 1 patient had mediastinal lymph node metastases, and another patient had contralateral axillary lymph node lesions).
|Characteristic||No. of Patients (%)|
|Lymph node classification|
|M1, oligometastases||6 (18)|
|UICC TNM stage|
|Not evaluable||12 (35)|
|Hormone receptor status|
|Not evaluable||1 (3)|
|Positive (+3 IHC and/or amplified FISH)||7 (21)|
|Negative (0, +1 IHC, or not amplified FISH)||25 (73)|
|Not evaluable||2 (6)|
|Anthracycline/taxane regimens, 6 cycles||8 (23)|
|Anthracycline/taxane-based sequential regimens, 8 cycles||26 (76)|
|Breast-conserving surgery||19 (56)|
PCT consisted of anthracycline-based and taxane-based regimens for 6 cycles in 8 patients and anthracycline-based and taxane-based, sequential 4-cycle regimens for a total of 8 cycles in 26 patients. All but 2 patients underwent surgery at a median interval of 24 days (range, 11-52 days) between the last cycle of chemotherapy and surgery. Intervention consisted of conservative surgery in 19 patients and mastectomy in 13 patients. One patient underwent simultaneous mastectomy and wedge resection of 3 liver metastases.
Pathologic responses after the completion of chemotherapy are reported in Table 2. Five patients were not evaluable: Three 3 of those 5 patients chose to undergo surgery at another hospital; however, from a review of their surgical specimens by our pathologist, they were classified as nonresponders. Two of those 5 patients did not undergo surgery because of the persistence of over-large local disease extension; however, they did undergo a new breast core-needle biopsy and were classified as nonresponders. Overall, 3 patients (9%) had a pCR and 4 patients (12%) had pathologic minimal residual disease; thus. an overall pR was observed in 7 patients (21%), and a pNR was observed in 27 patients (79%). It should be noted that the patient who had liver metastases had TRG 5 both in the breast lesion and in the liver lesions.
Table 3 provides a detailed description of major clinical and biologic tumor characteristics, pathologic responses, and FDG-PET results for each patient. The baseline FDG-PET scan was obtained before breast biopsy at a median interval of 11 days (range, 3-34 days) from the scan to the first cycle of PCT. In all patients, the baseline SUVmax value in primary tumor was abnormal (median SUVmax, 9.6; (range, 2.5-23); whereas the baseline SUVmax value in axillary lymph nodes was abnormal in 20 patients (59%; median SUVmax, 6.2; range, 2.4-23.4).
|Patient No.||TNM Stage||ER Status||HER2 Status||TRG||NRG||Pathologic Response||Basal SUVmax||Δ-SUV, % of Patients|
|After 2 Cycles||After 4 Cycles||Before Surgery|
The baseline FDG-PET study changed the clinical stage in 6 patients: One patient with stage IIB disease was restaged with stage IV disease because of a single bone lesion, 3 patients with stage IIA disease were restaged with stage IIB disease, 1 patient with stage IIB disease was restaged with stage IIIA disease, and 1 patient with stage IIIA disease was restaged with stage IIIC disease. The median baseline SUVmax was 11.3 (range, 4.5-23) in pathologic responders and 9.3 (range, 2.5-19.9) in nonresponders (P = .41).
The Δ-SUV after 2 cycles, after 4 cycles, and before surgery is detailed for individual patients in Table 3. All but 1 patient presented with a reduction in SUVmax after 2 cycles. The median Δ-SUV after 2 cycles, 4 cycles, and 6 to 8 cycles was 71% (range, ≥14%-100%), 96% (range, 4%-100%), and 100% (range, ≥28%-100%), respectively. The mean Δ-SUV value was greater in patients who had a pR than in patients who had a pNR throughout the period of FDG-PET monitoring (Fig. 1); however, the difference did not reach statistical significance.
We chose to define the cutoff value of Δ-SUV for differentiation between pathologic responders and nonresponders after 2 cycles of PCT that had optimal sensitivity and negative predictive value (NPV). The highest NPV was obtained with a 50% Δ-SUV decrease threshold. All 7 patients who had a pR had Δ-SUV values >50%, and all 8 patients who had Δ-SUV values ≤50% had a pNR, as illustrated in Figure 2. At this cutoff level, the sensitivity, specificity, and NPV of the Δ-SUV after 2 cycles were 100%, 30%, and 100%, respectively. Of 27 patients who had a pNR, 8 patients (30%) were identified correctly by a 50% Δ-SUV threshold. In 6 of those 8 patients, the pNR was represented by poor reduction in tumor cellularity (TRG 1 or 2). Although no changes occurred at a 45% threshold, lower values of sensitivity, specificity, and NPV were obtained at a 55% threshold (Table 4). By increasing the number of cycles, sensitivity and the NPV with a 50% threshold remained the same, whereas specificity decreased (Table 5). FDG-PET accuracy turned out to be scarce at any time during monitoring because of the very low specificity as a result of the high number of false-positive findings.
|Variability||% (No. of Patients/Total No.)|
|Δ-SUV Threshold = 45%||Δ-SUV Threshold = 50%||Δ-SUV Threshold = 55%|
|Sensitivity||100 (7/7)||100 (7/7)||71 (5/7)|
|Specificity||26 (7/27)||30 (8/27)||33 (9/27)|
|PPV||26 (7/27)||27 (7/26)||22 (5/23)|
|NPV||100 (7/7)||100 (8/8)||82 (9/11)|
|Accuracy||41 (14/34)||44 (15/34)||41 (14/34)|
|Variable||No. of Patients/Total No. (%)|
|After 2 Cycles||After 4 Cycles||Before Surgery|
|Sensitivity||7/7 (100)||7/7 (100)||7/7 (100)|
|Specificity||8/27 (30)||5/27 (19)||6/27 (22)|
|PPV||7/26 (27)||7/29 (24)||7/28 (25)|
|NPV||8/8 (100)||5/5 (100)||6/6 (100)|
|Accuracy||15/34 (44)||12/34 (35)||13/34 (38)|
The analysis of the relation between the pathologic and metabolic response after 2 cycles of PCT according to estrogen receptor (ER) status and human epidermal growth factor receptor 2 (HER2) status is reported in Table 6. All 8 patients who had Δ-SUV values ≤50% (true nonresponders) had ER-positive tumors, representing the 40% of this patient class (8 of 20 patients). None of the 13 patients who had ER-negative tumors had Δ-SUV values ≤50%. At the same time, 6 of 7 true nonresponders had HER2-negative tumors. Overall, FDG-PET monitoring predicted that 6 of 16 individual patients (37.5%) who had ER-positive/HER2-negative tumors would obtain a pNR.
|Patient Group||No. of Optimal Pathologic Responses/Total No. of Patients (%)|
|ER Status||HER2 Status|
|PET responders: Δ-SUV >50%||3/12||4/13||7/25||1/6||6/19||7/25|
|PET nonresponders: Δ-SUV ≤50%||0/8||0/0||0/8||0/1||0/6||0/7|
|Total||3/20 (15)||4/13 (31)||7/33 (21)a||1/7 (14)||6/25 (24)||7/32 (22)b|
To our knowledge, all studies regarding the role of FDG-PET in the monitoring of patients with localized breast cancer during PCT have evidenced a correlation between early changes in the SUVmax and the pathologic response; however, the role of this kind of monitoring in clinical practice remains unclear because of the substantial heterogeneity of the results. Some important issues underlying the heterogeneity across studies are the definition of pathologic response, the timing of early FDG-PET evaluation, and the choice of the Δ-SUV threshold to separate responsive and nonresponsive patients.
The definition of a histopathologic response differs across studies9, 10, 12, 13 or is lacking completely14: either information on the concomitant pathologic status of axillary lymph nodes is not reported10, 12, 15 or, ultimately, patients who obtain a suboptimal pathologic response are included in the group of responsive patients.12 Consequently, the incidence of patients who are considered optimal pathologic responders has wide variability (range, 17%15 to 47%).12 We used the Miller-Payne grading system because it is considered the most complete classification: It grades pathologic remission in both the primary tumor and the axillary lymph nodes by comparing negative lymph nodes with those that already have responded to chemotherapy,19 and it restricts the definition of an optimal pathologic response (pR) only to those patients whose tumors had a very high response (ie, a reduction in cellularity >90%) with the concomitant absence of tumor cells in the axillary lymph nodes. Patients with all other conditions were considered to have obtained a pNR. By using this grading system, our pR rate was 21%.
Various investigators have reported the results of PET monitoring after the first cycle,13 after the first and second cycles,10, 15 or at various subsequent time points during PCT in addition to the first and second cycles.10, 12, 13 There does not appear to be a consensus regarding the optimal timing of early PET evaluation. The changes in SUV occur rapidly during the first part of treatment; and, even if the decrease in SUVmax continues up to the end of chemotherapy, the curve tends to flatten out (Fig. 1). In our study, the SUVmax curves declined in responsive and nonresponsive patients and tended to diverge over the whole PTC treatment: The mean differences were 7% after 2 cycles and 20% by the end of treatment. From Figure 1, we can observe that the difference between the 2 curves after the first cycle was lower than after 2 cycles. McDermott et al.13 have suggested that the best time to monitor chemotherapy response is most likely between the end of the first cycle and the chemotherapy midpoint. Bearing in mind the need to look for indications that may be helpful in the decision-making process, based on our results, we suggest that the best timing to perform the early evaluation of response by FDG-PET is immediately before the third cycle.
A critical point concerns the choice of the Δ-SUV cutoff value to identify patients who will attain a pR and those who will not reach this result. In the clinical setting of primary systemic treatment for surgically operable early breast cancer, a reliable test to identify the individual patients who will not respond optimally after 2 cycles would allow us to bring forward the surgical procedure or to change the medical treatment schedule, ie, change over to a different chemotherapy regimen or move on to the other chemotherapy regimen directly when the sequential 2-regimen PCT program had been planned. Conversely, when the clinical setting is characterized by locally advanced breast cancer, the same early information could make a change in the strategy a feasible option, even if it most likely has a more limited clinical bearing and prognostic significance. Hence, the capacity of early FDG-PET to identify individual patients who will not respond optimally is more interesting for clinical practice. Consequently, the NPV, rather than the overall accuracy of the test, has major relevance in the analysis of the results. The Δ-SUV cutoff value after 2 cycles should have a very high NPV, possibly 100%. In the study by Rousseau et al.,12 the highest NPV (85%) was obtained with a 40% Δ-SUV threshold after 2 cycles. This result appears to have limited clinical application because of the relatively high proportion of false-negative findings, most likely because those investigators included patients who had a suboptimal pathologic response in the group of responders. Schwarz-Dose et al.,15 using a definition of optimal pathologic response according to Honkoop et al.,21 obtained a 96% NPV with a 45% cutoff value in a multicenter study that included 64 patients who underwent FDG-PET monitoring after 2 cycles. This corresponds to 1 false-negative patient in 23 negative patients. These results are very close to those observed in our study, in which an NPV of 100% was obtained with a Δ-SUV threshold of 50%.
In the current study, PET early identified 8 of 27 individual patients (30%) who would not achieve a pR. This corresponds to 23.5% of the whole series of enrolled patients. This percentage appears to be too low to recommend the introduction of FDG-PET in the standard baseline workup and monitoring of PTC for patients with early or locally advanced breast cancer. However, in our analysis according to biologic tumor characterization, the early identification of nonresponders was set aside exclusively for patients who had ER-positive tumors and for those who had ER-positive/HER2 negative tumors when both parameters were available. In particular, early PET evaluation predicts pNR in 40% ER-positive tumors. In previous reports on the role of early PET evaluation during PCT, an analysis of the correlation between biologic characteristics of individual tumors and the Δ-SUV was lacking. Our results are in keeping with the well known lower chemosensitivity of ER-positive tumors compared with ER-negative tumors in the preoperative setting.22, 23 In addition, the subgroup of ER-positive/HER2-negative tumors is biologically heterogeneous, because these tumors generally are both responsive to hormone manipulations to various degrees and variously responsive to chemotherapy, including poor sensitivity and resistance to cytotoxic agents.
Potential drawbacks for the generalization of our observations include the variability of PET/CT technology in current international clinical practice and the consequent possible differences in SUV values from center to center. It is well known that many factors may influence the SUV calculation. Several reports indicate that injection accuracy,24 uptake time, fasting period, and glucose levels25 affect SUV values. However, the influence of these factors can be minimized by standardizing the methodological approach as much as possible.26 SUV calculations also may depend on other factors, such as the type of tomography, the acquisition, and the reconstruction parameters.27 This finding implies that there is significant variability in SUV values across centers; and, finally, it prevents the possibility of deriving an absolute SUV cutoff value from 1 institution to another. Therefore, only SUV changes expressed as percentage modifications (Δ-SUV) can be considered, because the exact same protocol is used to calculate those values.
The chief weakness of the current study is the small number of patients studied. For this reason, the results (and, in particular, the identification by FDG-PET monitoring of individual patients with ER-positive tumors and possible ER-positive/HER2-negative tumors who will not obtain a pR with PCT) must be confirmed. A much larger study will be necessary to determine whether PET/CT has any role to play in the early detection of nonresponders. For PET/CT to have value, it must reliably identify these patients in a prospective manner so that therapeutic intervention may be redirected thoughtfully. Consequently, our findings are to be considered only as a proposal that will require confirmation.
In conclusion, the current study confirmed that the early evaluation of metabolic response by FDG-PET monitoring during PCT for breast cancer correlates with the pathologic response documented at the time of surgery. However, such evaluation does not appear to be useful in selectively identifying those patients who will achieve an optimal pathologic response (pCR or pMRD) because of the high number of false-positive results. Conversely, FDG-PET individually was able to identify 30% of patients who would not have an optimal pathologic response. It is noteworthy that this predictive power is limited only to ER-positive tumors, an observation that warrants further study in a larger patient series. If these findings are confirmed, then FDG-PET performed at baseline and after 2 cycles of PCT could supply important information to be used in the decision-making process.
The authors made no disclosures.