Trial Limitations

to etoposide and etopophos is an important confirmatory finding indicating that indeed CAP7.1 is a safe option, which might deliver a local cytotoxic tumor effect compared to etoposide or etopophos under a comparable safety profile. Therefore, the clinical use of CAP7.1 is considered to be at least as safe regarding the hematotoxic effects as and superior regarding organ toxicitiy compared to etoposide or etopophos as drugs with existing market approval.

of the so far observed data rather than to decide once and for all whether CAP7.1 therapy is actually effective and safe or not as would normally be the goal in a final analysis of a traditionally designed study. Thus, the latter question of CAP7.1 efficacy and safety in this study will also only be definitely answered when the full recruitment target of in this case 50 patients (chapter 2.3) is finally reached unless, however, the observed effects at the respective interim analyses are so extreme, that results become statistically significant already earlier and the study is prematurely halted on this basis. Therefore, however, having identified strong and encouraging trends in this first interim analysis is actually a very encouraging, positive outcome even if statistical significance was not reached because the study was deliberately designed in a way that statistical significance at first interim analysis is extremely difficult and rather unusual and unlikely to be reached.

Another relative limitation is the time interval between RECIST assessments used. Even though an 8 week (every 2^nd cycle for CAP7.1 patients) interval is already short compared to other studies in the same indication, for the severely ill patient population of this study with an extremely short life expectancy of sometimes even only weeks rather than months, a six week or even four week interval would probably have been more meaningful from a pure research perspective allowing for more precise measurements in a research environment that is also looking for effects lasting rather for weeks than months. However, best research practices always need to be weighed up against practical concerns and quality of life considerations for this severely ill end stage patient population. Therefore, what may be desirable in terms of research practice may sometimes not be practically or ethically viable and in this study 8 week intervals for different reasons seemed the overall best compromise. Furthermore, however, potential problems with RECIST measurements and their timing were already discussed in the previous section (4.1) using examples from this study as was the need for the conduct of an additional external, independent radiological review according to EMA and FDA guidelines.

Furthermore, missing values constitute a limitation for this study. However, these missing values lead to an underestimation rather than an overestimation of the CAP7.1 effect so that the results presented in this thesis can be considered very conservative and robust. Further RECIST measurements were missing mainly due to adverse events or death or other circumstances not allowing CT scans to be performed.

Another limitation was the CAP7.1 starting dose. According to protocol, the CAP7.1 starting dose was 200 mg/m². However, for different reasons only eight patients of the safety population

(SP) were started on 200 mg/m² while 12 patients were started on 150 mg/m² and one patient was even started on 110 mg/m². Therefore, the majority of patients were actually started at a lower dose than 200 mg/m², which may have affected efficacy and safety results.

Two other limitations follow from ethical considerations. First of all, even though the study had a randomized design it was open label since it did not seem feasible to conduct a double-blind trial for a cytotoxic drug which potentially therefore may have introduced some bias into the study. Second of all, for ethical reasons as well, patients in the BSC group were allowed to cross-over to CAP7.1 therapy once they progressed under BSC only therapy. This practice, however, on one hand limited the value of the OS analysis since in the end basically also all BSC group patients except for one patient received CAP7.1 after progression. However, on the other hand this unique design feature also made an intra-patient comparison of the two treatments possible which added significant value to the study.

Finally it needs to be mentioned that this phase II clinical trial was designed at a time when GEM/CIS had just been established as the new 1^st-line therapy standard for advanced BTC while most of the other more recent findings from the literature as summarized in the introduction of this thesis were still not known/available. The study design therefore also did not consider/address some of the main important literature findings such as the fact that resected patients generally have a better therapy outcome than unresected patients under any additional subsequent therapy. However, as shown in chapter 3.2 (baseline characteristics) about one third of patients were surgically pretreated, 75% of which were in the BSC group. Therefore, this fact can potentially only lead to a further underestimation of the CAP7.1 effect since the majority of surgically pre-treated patients was actually in the BSC group. In addition, there is also a strong general individual prognostic factor dependence of therapy outcome emerging from the current literature so that outcome may in fact be very different for different patient sub-groups. Also this finding can potentially introduce bias to the study if not appropriately considered. However, because both these findings had not yet emerged so clearly from the literature at the time this study was designed, the study was also not designed to enable meaningful relevant subgroup analyses which in turn would have surely also required a larger sample size. Nevertheless, however, because besides the typical between treatment group comparisons this study also provides a within patient comparison of the two treatments for 9 of the BSC group patients, it still provides invaluable information also in the light of these recent literature findings in a very elegant way because these individual within patient comparisons actually are completely independent of prognostic factor profile differences between the patients including also

differences in surgical pre-treatment since their reference point for comparison is the same patient with the same prognostic factor profile no matter how complex this profile otherwise may be. For this reason, the cross-over design of this study is a blessing which further increases the overall validity and quality of the study especially also in the light of all recent literature findings and even if it limits the value of the OS analysis in turn. Furthermore, since Moriwaki (2016)⁹¹ in addition were able to show that mPFS is an appropriate surrogate endpoint for mOS in phase II trials in this indication (see above), also the latter limitation seems not really relevant anymore.