Sir,

We read with great interest the recent article by Alvarez and co-workers reporting the effects of nab-paclitaxel on tumour stroma in pancreatic cancer (Alvarez et al, 2013). The authors should be congratulated on their interesting findings from a translational study that investigated the biological effects of neoadjuvant gemcitabine and nab-paclitaxel in patients with resectable disease. Although the primary study endpoint was to determine the effects of this novel chemotherapy regimen on tumour stroma (including an assessment of collagen content and structure of ‘cancer-associated fibroblasts’ (CAF)), some important lessons can be learned from this innovative study regarding the approach of neoadjuvant chemotherapy in localised pancreatic cancer (Heinemann et al, 2013).

The investigators enroled 16 patients with histologically or cytologically confirmed resectable or borderline resectable exocrine pancreatic cancer that had received two cycles of gemcitabine/nab-paclitaxel. After chemotherapy, 3 patients had progressive disease and 12 patients were able to proceed to surgical resection. Of note, in the final pathological assessment of the surgical specimen, two study patients in fact had a neuroendorcine cancer and not pancreatic adenocarcinoma. No objective response by RECIST was observed after neoadjuvant chemotherapy; however, half of patients had a CA 19-9 decrease of >75% and a partial metabolic response by FDG-PET imaging. The authors reported a median CA 19-9 pre-treatment value of 2.588 U/dl in their patient population, ranging from 1 to 36.376 U/dl; this would correspond to a median CA 19-9 level of 26 U ml−1 at baseline with a range from 0.01 U ml−1 to 364 U ml−1, respectively (please note: (U/dl) is a very uncommon unit for reporting CA 19-9; the most commonly used unit in the literature is ( U ml−1)) (Boeck et al, 2006). On the basis of the evidence from previous CA 19-9 biomarker studies, significantly elevated CA 19-9 levels before surgical resection may indicate a potential subradiographic systemic disease rather than a localised disease. In a single-centre trial from the USA, 51 out of 262 patients who underwent staging laparoscopy for radiographically resectable pancreatic adenocarcinoma in fact had unresectable disease. In this study, the median pre-operative CA 19-9 value for patients who underwent resection was 131 U ml−1 vs 379 U ml−1 for those patients with unresectable disease (P=0.003). With the use of a receiver operating characteristics curve for pre-operative CA 19-9 value and tumour resectability, the statistically optimal cut-off value was determined to be 130 U ml−1 (Maithel et al, 2008). In concordance, Katz et al (2010) suggested a pre-treatment CA 19-9 cut-off of 149 U ml−1 as a predictor of completing chemoradiation and subsequent pancreaticoduodenectomy in their neoadjuvant study (Ketz et al, 2010). In the light of the fact that ‘… the content of stroma and CAF may change with different cancer stages’ (as the authors themselves stated correctly in the Discussion section), one should keep in mind that an appropriate selection of patients with a unique stage of disease thus should be an important issue in novel translational studies.

Nevertheless, the data presented are important and highlight an innovative approach to how future translational research should be conducted in pancreatic cancer. However, from a clinical point of view, this small study clearly unbosoms potential pitfalls of neoadjuvant treatment approaches in resectable pancreatic cancer: (1) the difficulty of adequately select patients with a real localised disease, (2) the limitations of obtaining an adequate pre-treatment histological diagnosis and (3) the often missing objective response by RECIST to neoadjuvant chemotherapy. The latter one was already described previously in borderline resectable patients and in fact it may be more appropriate to investigate different clinical methods than conventional CT scan for assessing treatment response to a neoadjuvant treatment regimen (e.g. metabolic response by PET or biochemical response by CA 19-9) (Takahashi et al, 2010; Katz et al, 2012).

We thus would recommend including a rigorous histological (no cytological) confirmation of disease before study entry, an assessment of response by different methods other than CT scan and also a CA 19-9 cut-off for patients entering neoadjuvant trials. Such an approach has recently been included in several large phase III adjuvant trials (e.g. RTOG 0848, Prodige 24/Accord 24), mainly by using a post-resection CA 19-9 level of <180 U ml−1 as an inclusion criterion.