The role of chemotherapeutic drugs in the evaluation of breast tumour response to chemotherapy using serial FDG-PET

Positron emission tomography (PET) imaging is increasingly incorporated into the routine management of cancer therapy. The predominant tracer in use is [¹⁸F]flourodeoxyglucose (FDG), based on the well-known increased glucose requirement of cancer cells [1]. FDG-PET over the past few years has demonstrated promise in the assessment of breast cancer response during chemotherapy, thus facilitating clinical decisions regarding patient management and providing valuable information about ways to tailor therapy to individual patient requirement [2‐6].

It has been established for over a decade now that patients who achieve a complete pathological response (cPR) in the primary breast tumour and the axillary lymph nodes have improved rates of disease-free survival [7‐10]. Only 20 to 30% of patients, however, respond to treatment with cPR [11]. The majority of patients achieve a partial pathological response (pPR) to chemotherapy, and these patients are at higher risk of locoregional recurrence than those patients with a cPR [8]. It is therefore imperative to assess tumour response to treatment as early as possible in order to identify those tumours that demonstrate sensitivity to the chemotherapy regimen, to direct appropriate treatment and to spare patients unnecessary chemotherapy. Conventional imaging, such as mammography and ultrasound, provide only information on anatomical changes in tumour morphology and size. Any residual masses present after completion of chemotherapy may be composed of viable tumour cells and/or fibrotic tissue, but conventional imaging modalities are largely unable to differentiate between the different cell types. Histopathology is still required in order to accurately characterise the remaining masses.

Since tumours undergo changes in metabolism before apparent changes in size, it is important to evaluate functional changes as early as possible during treatment. Several previous studies have shown that high baseline uptake of FDG by the tumour is associated with poorer survival compared with tumours with lower uptake [12‐14]. It has also been demonstrated that a significant reduction in glucose uptake on FDG-PET during or after chemotherapy can predict cPR [4, 5, 15, 16]. The information obtained from FDG-PET studies to date further suggests that changes in glucose uptake by the tumour are independent of other, traditional, predictive breast cancer markers [17]. The impetus now is to identify chemosensitive tumours as early as possible during treatment. One study has recently demonstrated that a reduction in glucose uptake after just one cycle of neoadjuvant chemotherapy is predictive of cPR [4].

Many of the reported studies, however, investigated small samples of patients undergoing various different chemotherapeutic regimens. We still know little about the functional changes of tumours in response to specific drug regimens. Hence, we designed the present prospective study to analyse serial changes in tumours of women with locally advanced breast cancer undergoing neoadjuvant chemotherapy with docetaxel followed by anthracycline-based chemotherapy (FEC100), or of women undergoing FEC100 treatment followed by docetaxel using FDG-PET, mammography, ultrasound and clinical assessment - and to compare these changes with histological changes in sequential biopsies.

Sixty-eight patients were recruited into the study. Of these patients, one withdrew from the study and seven had not completed treatment and evaluation at the time of analysis. These patients were not included in the final analysis. The final number of participants therefore included 31 patients who were randomised to receive docetaxel followed by FEC100 and 29 patients who received the same drugs in reverse order.

Tables 1 and 2 demonstrate the demographic characteristics of patients and tumour characteristics according to drug regimen used. There was no significant difference in any of the patient or tumour parameters studied between the two groups of patients. The majority of patients (50/60, 83%) presented with ductal carcinoma. Seven patients (11.6%) had lobular carcinoma and three patients (5%) had mixed ductal/lobular carcinoma. Most patients (33/60, 55%) had high-grade tumours. Thirty-seven out of 60 patients (61.6%) were oestrogen receptor-positive. Although more of these patients were randomised to receive docetaxel initially rather than FEC100 (71% compared with 52%, respectively), the difference in proportion was not significant. Similarly, the proportions of women who were progesterone receptor-positive and HER2-positive were not significantly different between the two groups. The proportion of women who underwent mastectomy or a wide local excision upon completion of chemotherapy was not statistically different between the two groups. The median (ranges) resting time prior to PET imaging was 60 minutes (45 to 75 minutes).

Tumour size on mammography and ultrasound and the SUV uptake on FDG-PET were not significantly different between the two groups of patients at baseline (Table 2). The median (range) SUV_max uptake for all women was 6.5 (1.58 to 26.6). Higher baseline SUV_max uptake was associated with higher rates of final pathological response, although statistical significance was not reached (Figure 1).

The numbers of tumours that were classified as cPR, as pPR and as pathological response not included in cPR or pPR are presented in Table 3. All but one tumour were downstaged during neoadjuvant chemotherapy. Overall, 23% of tumours (14/60) were classified as having undergone a complete pathological response. Thirty per cent of tumours (18/60) underwent a partial response, and the other tumours (28/60 (47%) recorded only minor or no changes during chemotherapy. There was no significant difference in tumour response between the two drug regimens (P > 0.05).

Table 4 demonstrates the reduction in tumour volumes after four and eight cycles of chemotherapy as observed on mammography, ultrasound and FDG-PET according to the drug given. There was no statistically significant difference between the two groups of patients when comparing tumour volume reduction or SUV_max uptake on either imaging modality. The only significant difference observed was associated with SUV_max uptake after four cycles of chemotherapy with anthracyclines. In this group of women, minor or no final pathological response was associated with minimal reduction of SUV_max uptake compared with women receiving docetaxel (P = 0.01). These outcomes are also highlighted in Figure 2. Women receiving FEC100 first who achieved a cPR had a median reduction of SUV_max of 87.7% from baseline (Figure 2a). In contrast, those on the same drug whose tumour recorded no or only minor pathological response had a median reduction of only 27% (Figure 2b). These data are significantly different from those by women who received docetaxel initially (P = 0.01). This group had no significant difference in the median SUV_max reduction according to pathological response. The percentage reduction in SUV_max uptake after all eight cycles is shown in Figure 3. There were no significant differences between the two drug sequences in terms of pathological response (P > 0.05).

The sensitivity, specificity, positive predictive value, negative predictive value and accuracy rates for both treatment arms are presented in Table 5. Highest sensitivity (0.87) was obtained for SUV_max reduction after chemotherapy with FEC100 (Arm B). In contrast, sensitivity was 0.83 for women in Arm A. Specificity and overall accuracy was best among women who received Arm B (0.76 and 0.79, respectively). Similarly, both positive and negative predictive values scored highest in this group of women (0.58 and 0.94, respectively).

The proportion of women who achieved SUV_max reduction ≥ 75% after four cycles of chemotherapy in either treatment arm was also compared according to nodal status at surgery (Table 6). There was no difference between the proportion of women with ≥ 75% reduction of SUV_max in either Arm A or Arm B (P = 0.63). Overall, women with ≥ 75% reduction of SUV_max after four cycles were more likely to have negative nodes at surgery (P = 0.13). There was no statistical difference between the proportion of women with negative nodes who were in either Arm A or Arm B (P = 0.9). Among those with positive nodes, however, the majority were women who had received docetaxel rather than FEC100 initially (90% versus 10%, respectively; P = 0.001).

The histological cancer category (ductal, n = 50 vs. lobular or lobular/ductal, n = 10) did not influence the baseline SUV_max uptake (P = 0.48) or that recorded at the halfway point or at the end of treatment (P = 0.96 and P = 0.51, respectively; Table 7). Of the 10 women with lobular or lobular/ductal cancer, seven (70%) had residual nodal disease upon completion. This observation compares with 26/50 (52%) of women with ductal cancer (P = 0.4).

Women who present with locally advanced breast cancer are today routinely managed using neoadjuvant chemotherapy, with the aim of downstaging the primary tumour before surgery. The most commonly used chemotherapeutic drug in this setting is docetaxel. Sequential anthracycline-based chemotherapeutic drugs are increasingly used but little information is so far available on the sensitivity of the tumours to these drug regimens during treatment. The standard sequence of drugs for adjuvant therapy has to date involved anthracyclines followed by taxanes, but phase II trials have recently suggested that the reverse sequence may improve efficacy and lower toxicity [19]. In our neoadjuvant setting, there was no significant difference in the efficacy between the two drug regimens. In fact, slightly more women reported a complete pathological response when treated first with FEC100 compared with docetaxel (27% versus 19%, P = 0.46).

As expected, mammographic and ultrasound measurements of the breast tumours did not prove useful predictors of tumour response. These traditional imaging modalities utilise tumour size as a measure of tumour cell kill. Such criteria have been used for over 50 years but their clinical usefulness remains limited. Functional imaging has more recently moved to the forefront of these investigations. PET with or without co-registration of CT using the radiolabelled glucose analogue FDG shows promise for diagnosis, for staging and as a predictive marker of tumour response for a range of tumours, because changes in glucose metabolism occur prior to any significant changes in tumour size [20].

Results in breast cancer patients to date have shown that SUV on FDG-PET after only one or two cycles of chemotherapy can be used to measure early response, with recorded sensitivities of between 61 and 100% and specificities of between 75 and 100% [15, 21‐24]. Analysis of these outcomes is difficult, however, since the disease stage, the definition of response, the drugs and sequence of drugs used, the calculation of FDG uptake and the cutoff points used to evaluate sensitivities and specificities varied among the patient cohorts. Further, most of these studies evaluated small populations (< 25 participants) - with the exception of the study by Rousseau and colleagues that evaluated 64 patients with stage II and III breast cancer [22], and the study by Smith and colleagues who investigated 31 women with large (> 3 cm) breast tumours [25].

One aspect of these studies that has not received a lot of attention is the SUV uptake in response to the drug regimen used. We hypothesised that the drug regimen may significantly affect the degree of reduction observed on serial PET scans because different drugs target different biological actions in the cancer cells. Most studies to date have all evaluated FDG-PET in cohorts of women who received various drugs with a range of dosage and sequences. It is thus difficult to assess the true usefulness of PET as a predictor of tumour response to treatment. In general, however, the literature has demonstrated that tumours with high levels of FDG uptake at baseline are more likely to achieve a complete pathological response [14, 22, 24, 25]. These findings were confirmed in our study.

Women in our cohort who recorded SUV_max reductions > 75% after four cycles of chemotherapy were more likely to achieve a cPR of the tumour. Further, tumours with SUV_max reductions < 30% after four cycles of FEC100 were more likely to remain insensitive to the treatment regardless of the subsequent use of docetaxel. Sensitivity, specificity, positive and negative predictive values, as well as overall accuracy were highest (when the 75% cutoff value was used) among women receiving FEC100 initially. Berriolo-Riedinger and colleagues achieved sensitivities and specificities of 0.91 and 0.86, respectively, after one treatment cycle [4]. In their study, however, women received four different treatment protocols.

Sensitivity of 0.77 and specificity of 0.80 were achieved in a study of 50 women with very large breast tumours (mean size 4.3 cm) who had FDG-PET before and after two cycles of chemotherapy with epirubicin and cyclophosphamide and, if axillary nodes were positive, additional taxanes [24]. Rousseau and colleagues achieved sensitivities and specificities of 0.61 and 0.96, respectively, in their cohort of women using the 60% cutoff value after two cycles [22]. Our comparative values (when calculating a 60% cutoff value) after four cycles are 0.87 and 0.56 in women receiving FEC100 and are 0.83 and 0.32 in women receiving docetaxel. These values are similar to those recently reported by Jung and colleagues [26]. In their study, 66 women were divided into two groups, with 31 women receiving docetaxel plus capecitabine and the other 31 women receiving doxorubicin plus cyclophosphamide. FDG-PET was carried out before and after four cycles of chemotherapy. The authors used an 84.8% cutoff value to achieve sensitivity and specificity values of 70% and 69.6%, respectively [26]. The patients were not analysed according to drugs used, but were evaluated as a whole group.

In an earlier study, patients receiving cyclophosphamide, doxorubicin, vincristine and prednisolone with or without docetaxel were scanned using FGD-PET after one cycle and four cycles of treatment [25]. There was no difference in the SUV_max reduction between the two treatment regimens used at either time point. Using a 20% reduction in SUV as the cutoff value, this study recorded sensitivity and specificity values of 0.90 and 0.74, respectively [25]. The patients included in that study had locally advanced cancer > 3 cm, which may have contributed to the relatively high sensitivity achieved. In our study, patients were included if their cancer was > 2 cm in size. Our less restrictive inclusion criteria may hence have resulted in the lower sensitivity and specificity values since the sensitivity of FDG-PET is known to be lower when the tumours are smaller [17].

It is interesting to compare the different drug-induced tumour responses observed on PET. The overall reduction in tumour volume was similar in all imaging modalities investigated regardless of the drugs used (Table 3). The bulk of tumour reduction occurred during the first four cycles. Women who received FEC100 and who had a median SUV_max reduction of only 27% or less, however, were significantly more likely to have treatment-resistant tumours (P = 0.01). In contrast, tumours of women who had received docetaxel could not be easily stratified. Our results further show that women who reported reductions of SUV_max uptake ≥ 75% halfway through treatment were significantly more likely to have positive nodes upon completion of chemotherapy if they had received docetaxel initially (90% vs. 10%) (Table 5). This association has not previously been reported. Since nodal status is an important prognostic indicator of overall survival and recurrence, further investigations in larger cohorts of women with adequate follow up of at least 5 years are required.

Our results demonstrate that the usefulness of PET in predicting tumour response can be influenced by the type of drugs used. This is an important consideration in the evaluation of PET during chemotherapy. All previous studies, with the exception of that by Smith and colleagues [25], evaluated groups of patients with mixed drug protocols. Given the various modes of actions of these drugs on tumour cells, it is not surprising that statistically significant differences in metabolic activity of the tumours may be observed on PET once therapy is initiated. The mechanisms of drug responsiveness/resistance are not fully understood. Docetaxel and FEC100 affect multiple pathways in tumour signalling and growth. Docetaxel disrupts microtubules, enhances apoptosis and affects angiogenesis [27, 28], while FEC100 interferes with the synthesis of DNA and reduces mitotic and metabolic activity. Today, most combination neoadjuvant chemotherapies involve docetaxel with anthracyclines since this combination improves the pathological response rates in primary breast cancers and results in improved disease-free survival and overall survival, at least in the adjuvant setting [28‐30]. Nevertheless, randomised prospective studies are needed in order to clarify how each drug protocol affects tumour glucose metabolism and the appearance/activity on serial FDG-PET studies.

SUV uptake values can be influenced by a range of factors. Core or fine-needle aspiration biopsies can potentially cause local inflammatory reaction, which could affect SUV uptake and mask true tumour uptake. This can impact on subsequent results and interpretation of the SUV values reported. In our study, the initial (baseline) PET studies were carried out at least 10 to 14 days after the diagnostic core or fine-needle aspiration took place, and are therefore unlikely to significantly affect the SUV uptake recorded for the baseline imaging. Once recruited, all women were biopsied only after the PET imaging took place in order to avoid influencing SUV measurements at the mid-treatment assessment. The final PET imaging took place before surgical intervention and hence did not have any impact on the third PET imaging report.

Another potential influence on SUV measurements is related to the variations of resting times that were permitted during the study. The protocol in our centre stipulates a 60 minute resting time prior to PET imaging. This was adhered to in the great majority of patients. On occasion, however, technical issues or delays in workflow resulted in fluctuations of the resting time. This was kept to a minimum whenever possible (variations of ± 15 minutes were permitted). Other factors potentially resulting in SUV variation or inaccuracy include partial volume effects (in which there is underestimation in small tumours), patient movement between PET and CT, including respiratory motion, and variation in FDG uptake time between initial and restaging PET scans. Whilst variation in technique may impact on the SUV measurements, it is very unlikely that a marked reduction in SUV (75%) will occur purely due to technical factors. Finally, we chose to report SUV_max values rather than SUV mean values because these are the values that are routinely reported to the treating oncologists and also to facilitate comparison of our results with those reported in the literature [22‐24, 26].

Our study assessed tumour response after four cycles of chemotherapy. Recent studies have shown that optimal response evaluation exists as early as after two cycles [22, 23]. It would have been useful to have carried out this earlier scan in our cohort of patients and to compare the sensitivity and specificity with other published studies. This should be investigated in future studies in order to obtain drug-specific response rates at a stage during treatment when patients are not exposed to several cycles of toxic and potentially ineffective therapy. Our study was also limited by the heterogeneity of the breast tumours in our cohort of women. This heterogeneity possibly contributed to the large variation of SUV_max values observed at baseline among our patient cohort and may be reflective of the variable biological factors inherent to breast cancer.

We have previously observed a correlation between PET SUV and a 12-gene signature [31]. Despite the biological diversity of breast cancers in our cohort of women, tumour type (ductal versus lobular) did not appear to influence SUV_max uptake at any stage of treatment (Table 6). There was no association between residual nodal disease and PET response according to histological types of cancer. Nevertheless, given the small number of women with lobular or lobular/ductal breast cancer in our study (n = 10), the next series of studies should investigate the relationship between various histological breast cancer types, drugs and SUV uptake on FDG-PET in a larger cohort of women. Such information will provide clinicians with better tools to direct individualised patient management and predict treatment responses early on during therapy.

Our results have shown that SUV_max uptake by breast tumours on FDG-PET during neoadjuvant chemotherapy may be dependent on the drug regimen used. The predictive value of FDG-PET appears to be more useful in patients treated initially with anthracycline-based chemotherapy rather than taxane-based chemotherapy. Care must be taken when evaluating FDG-PET in settings where patients receive varied drug protocols. More studies are needed to clarify and standardise the relationships between tumour type, chemotherapeutic drugs and SUV uptake on serial FDG-PET imaging.