Pancreatic ductal adenocarcinoma (PDAC) is one the most lethal cancers, with an incidence rate of about 15 per 100 000 per year in Norway (2023) (1). About 85% of patients with PDAC are not eligible for upfront surgery with curative intent. Many of these receive neoadjuvant or palliative therapy, but unfortunately, the response rates are low (2). Biomarkers of chemotherapy response are crucial for selecting patients who are likely to benefit from treatment in a personalized way (3).

In PDAC, carbohydrate antigen 19-9 (CA19-9) is the most commonly used biomarker for therapy response, but has limited predictive power (4). Considering new methods in oncology in general, tumor characteristics like expression of specific receptors or the presence of gene rearrangements have so far been the typical candidates as biomarkers (3). The gut microbiota composition has been increasingly associated with multiple health characteristics, including the response to some cancer therapies (5, 6). Recently, Tintelnot et al. identified circulating levels of the gut microbial metabolite indole 3-acetate (3-IAA) as positively associated with response to chemotherapy in metastatic pancreatic ductal adenocarcinoma (PDAC) (7). In an elegant experimental and molecular characterization study, the mechanism of increased chemotherapeutic effect was found to depend on oxidation of 3-IAA by myeloperoxidase from neutrophils, which subsequently increased reactive oxygen species in tumor cells and hence tumor cell death. The study introduced a novel concept, whereby a metabolite produced by gut microbes is absorbed systemically and modifies chemotherapy response in the host (8). In the study, chemotherapy response in mice could even be improved by supplementing the diet with the 3-IAA precursor tryptophan, leading to increased 3-IAA levels, suggesting that a similar adjuvant approach may demonstrate clinical efficacy in humans.

Chemotherapy in metastatic PDAC is given with palliative intent, but it is the primary treatment for patients with borderline resectable or locally advanced PDAC in order to improve resection rates with curative intent (2). We therefore measured 3-IAA in a cohort of patients with borderline resectable or locally advanced PDAC before starting chemotherapy (9), aiming to investigate if elevated serum 3-IAA associates with therapeutic response in a PDAC population without distant metastases.

In total, 124 patients (median 68 years, 47% male) with borderline resectable (n=68) or locally advanced PDAC (n=56) were included ( Table 1 ). Seventy-six (61%) patients were treated with fluorouracil, leucovorin, irinotecan, and oxaliplatin (FOLFIRINOX), 26 (21%) with gemcitabine plus nab-paclitaxel, and 22 (18%) with gemcitabine monotherapy or other chemotherapy regimens.

The median 3-IAA concentration before chemotherapy was 290 (IQR 203–417) ng/mL. This was within the range of concentrations observed in n=100 healthy individuals analyzed in the context of another study (median 377 (IQR 303–464) ng/mL). Serum 3-IAA was positively correlated with age (Spearman’s ρ=0.33, p<0.001; Supplementary Figure S2A ) but not with tumor diameter or serum CA19-9 ( Supplementary Figures S2B, C ). Pre-treatment 3-IAA was similar in the different chemotherapy groups ( Supplementary Figure S2D ).

We found no statistically significant differences in pre-treatment 3-IAA concentrations according to response as defined by CT scan or CA19-9 decline ( Figures 1A, B ). The 3-IAA concentration was numerically higher in patients who later had surgical resection and in those alive after one year, but the difference was not statistically significant ( Figures 1C, D ). In participants with a second sample taken, 3-IAA concentrations were on average increased after chemotherapy (median 265 nmol/L before and 328 nmol/L after), but the increase was not statistically significant (p=0.24, Supplementary Figure S2E ).

The median survival time was 499 days (95% confidence interval (CI) 437–597). Twenty-six (21%) patients were event-free at their last follow-up date. We stratified participants into quartiles by pre-treatment 3-IAA concentrations but found no significant differences in survival between the groups ( Figure 1E ), a finding supported by a Cox model where 3-IAA (log₂) had a HR=0.93, 95% CI [0.74–1.16], p=0.52. The model estimates were similar in the subgroups of borderline resectable or locally advanced cancer. Since 41 (33%) of the participants eventually had surgical resection of their primary tumor, which could influence survival, we evaluated the prognostic value of 3-IAA when censoring these cases at their time of surgery. Also here, we found no evidence of an association between 3-IAA and overall survival ( Supplementary Figure S3 ). Finally, using the full cohort, we adjusted 3-IAA for age, ECOG, CA19-9 and tumor classification, where 3-IAA (log₂) had an adjusted HR=0.87, 95% CI [0.68–1.12], p=0.28.

In the present study, serum 3-IAA concentration was not associated with response to chemotherapy or overall survival in a population-based cohort of borderline resectable or locally advanced PDAC starting chemotherapy. The results suggest that high 3-IAA levels do not predict chemotherapy response in borderline resectable or locally advanced PDAC.

The only data available for comparison are from the seminal paper by Tintelnot and co-workers (7) where an association between 3-IAA and chemotherapy response in PDAC was found. There are several differences between these studies that could potentially explain these conflicting observations. The association between chemotherapy response and 3-IAA in the Tintelnot et al. study was investigated in a total of 47 patients divided on two cohorts of metastatic PDAC (9). In contrast, while we studied more than the double number of patients (n=124), our population had no signs of distant metastasis. Patients with non-metastatic cancers often have a smaller total tumor load, better functional status (12), and differences in tumor biology compared with metastatic cancers (3, 13), all of which may influence the effect of chemotherapy and its modifiers.

Another major difference is the analytical methods. The median 3-IAA concentration in the present study was about 10-fold higher than what was found in Tintelnot et al. In a healthy control population analyzed in the context of other ongoing studies, the median 3-IAA was within the same order of magnitude. We therefore speculate that the discrepancy in 3-IAA concentrations arose mainly from assay differences, and it is difficult to exclude that this may in part contribute to the diverging results. Notably, the present data were generated with high quality LC-MS/MS-based methodology with a radioactively labelled internal standard. Nonetheless, if inter-individual variation in 3-IAA measurements was consistent with the two methods, we would expect to find an association with survival in our cohort if it truly existed. In our univariable model, 3-IAA had a wide confidence interval (0.74–1.16). While this makes it challenging to confidently assert the presence or absence of an effect, any potential clinical significance appears most likely to be small.

A third point is that apparent predictive performances in discovery cohorts are often optimistic and models commonly perform worse in new populations for a variety of reasons (14). This is illustrated in the study from Tintelnot et al. where the effect size observed in the second human PDAC cohort was lower than in the first cohort (explained variance in the correlation between 3-IAA and survival time was R² = 0.51 in a cohort from Hamburg and R² = 0.24 in a cohort from Munich).

One of the imitations of our study is that 54% of the patients received a biliary stent due to jaundice, and it remains unclear whether hyperbilirubinemia influences the gut metabolism of tryptophan, the precursor of 3-IAA (15, 16). Furthermore, only 28% of the cohort had a second blood sample withdrawn for serum 3-IAA measurement and only 61% were treated with FOLFIRINOX.

In conclusion, our results do not align with the optimism generated by Tintelnot et al.’s study. Additional validation of 3-IAA as a biomarker of chemotherapy response in PDAC is necessary before initiating human clinical trials aiming to modify 3-IAA levels in this setting. Furthermore, subsequent research in this field should probably focus on metastatic PDAC.