Online Tool PREDICT+ Underestimates Survival in HER2+ Breast Cancer

In an interview with Targeted Oncology™, lead study author Elisa Agostinetto, MD, of the Istituto Clinico Humanitas in Rozzano, Milan, Italy, discussed the accuracy of PREDICT+ in HER2-positive breast cancer in greater detail and its clinical implications.

PREDICT+ is a free, widely used, online tool that is used to make prognostic predictions for patients with breast cancer by predicting how different therapies after surgery can improve survival. However, it’s place in HER2-positive breast cancer remains unclear.

In order to test the reliability of PREDCIT+ in this patient population, investigators used data from the ALTTO trial (NCT00490139), which compared the safety and efficacy adjuvant lapatinib (Tykerb) and/or trastuzumab (Herceptin) in patients with HER2-positive breast cancer. In total, 2794 patients were included in the analysis with a median follow-up of 6 years. Overall, 182 deaths were observed.

Investigators used PREDICT+ to predict 5-year survival for each patient and then compared it to actual survival. PREDICT+ was found to drastically underestimate the 5-year overall survival for the overall population, which could impact patient designs about reproductive decisions.

In an interview with Targeted Oncology™, lead study author Elisa Agostinetto, MD, of the Istituto Clinico Humanitas in Rozzano, Milan, Italy, discussed the accuracy of PREDICT+ in HER2-positive breast cancer in greater detail and its clinical implications.

 

What was your American Society of Clinical Oncology’s Annual Meeting presentation about this year?

AGOSTINETTO: What we presented at ASCO 2021 are the results of our study, evaluating the prognostic role of PREDICT+ in the patients with HER2-positive early breast cancer. PREDICT+ is a widely used online tool that helps to show how the treatments administered after a breast cancer surgery can really improve the survival rate.

However, its prognostic role in HER2-positive early breast cancer patients was quite unclear. So, we aimed to really evaluate the prognostic role of PREDICT+ in patients with HER2+ positive early breast cancer. And to do so we use the data from the ALTTO trial. The ALTTO trial is probably the largest trial ever conducted in the adjuvant setting for patients with a HER2-positive early breast cancer.

So, basically what we did, we took the data from the patients enrolled in the ALTTO trial who received trastuzumab-based therapy, started concomitantly with chemotherapy. And we calculated for each patient the estimates of survival by PREDICT+, and then we compare the estimated survival by PREDICT+ with the observed survival of these patients. Overall, our analysis included 2794 patients. And the what we observed is that PREDICT+ highly underestimates the survival of patients enrolled in the ALTTO trial. And this underestimation was quite relevant because it was 6.7 percentage points in the 5-year overall survival in the overall population. Moreover, this finding was consistent across all the subgroups we analyzed. And the largest difference between predicted and observed the survival was observed for patients with the hormone receptor negative disease, patients with node negative disease, and patients with the larger tumor size.

So, why there was these underestimations of course, we can make some hypotheses. And one possible explanation can be related to the fact that we know that the patients with HER2-positive breast cancer are experiencing a consistent shift towards better survival across the years. And this is mainly thanks to effective therapies that are now available for these patients. So, it might be that the tool that was developed and validated 10 years ago, does not reflect these changes, for instance.

Very briefly, it was our study, I think that the main messages that we wanted to provide with our worker are first, that PREDICT+ should be used with caution to give a prognostic estimation in patients with the HER2-positive early breast cancer. Second, is that we really need reliable tools to provide the prognostic estimations in patients with HER2-positive early breast cancer. In HER2-positive disease, prognostication is really of paramount importance, because it has several implications. For instance, one of these is for young women in pre-menopausal status, the planning of a reproductive life, meaning that prognostication can really affect the choice of having or not have a pregnancy for instance.

I'm very happy and honored to say that our work was awarded with an ASCO merit award. So therefore, I really want to thank all the authors that participated in the study, starting with Matteo Lambertini, MD, PhD and Evandro de Azambuja, MD, PhD who mentored this project. But of course, the main thanks goes to the patients enrolled in the study and to their families.

What was the data and reasoning behind the study?

The main rationale that led us to perform this study is really related to the fact that although one may think that prognostication in HER2-positive disease is not as important as in other breast cancer subtypes. For instance, in hormone receptor positive/HER2-negative breast cancer, prognostication is particularly important because you can really choose whether the patient would receive an adjuvant chemotherapy, or the adjuvant endocrine therapy alone. On the contrary, in HER2-positive breast cancer, almost the totality of patients in the early setting receive chemotherapy. The way of thinking is that prognostication can be less important in HER2-positive disease. On the contrary, as I said, I think that HER2-positive disease prognostication is very important. And one of the reasons is the one that I mentioned earlier, and it is really the planning of the reproductive life. Meaning that prognostication can really affect the choice for a young woman to have or not to have a pregnancy, for instance. But also, in terms of the choice of adopting, escalation and de-escalation treatment strategies, for instance. So, the idea was really to investigate the prognostic performance of PREDICT+, and the ALTTO trial was really the best study to test this because it has a very large population. And also, the follow up is very long. So, it was really the best study to test this prognostic performance tool.

What were the results? Did you find anything shocking or significant?

In our study we assessed the calibration and discriminatory accuracy of PREDICT+. And our main results are related really to the fact that PREDICT+ underestimate the survival of the patients enrolled in the ALTTO trial. And this underestimation in the overall population was 6.7 percentage points in the 5-year over survival, this is in the overall population. But then the finding was consistent in all the subgroups that we analyzed. The other important finding is that the largest of differences between the observed survival and the predicted survival was observed for patients with the hormone receptor negative disease, for patients with node positive disease, and patients with the larger tumor size. In terms of the discriminatory accuracy, the discriminatory accuracy of PREDICT+ was overall low. And again, this finding was consistent across all the subgroups that we analyzed. So, these are the main results of the study.

What are the next steps for this research?

Well, of course, it would be interesting to understand why there was this underestimation. One hypothesis can be related to the fact that the patients included in clinical trials are overall more selective, meaning that they can be healthier, with less comorbidities. This can create a discrepancy between the real-world data and the data collected from the clinical trials. So, I think that as future perspective, the most important thing, and what we really need from the future study is having data that can help us to provide a very precise and reliable prognostication for patients with HER2- positive early breast cancer. In this sense, probably the integration of genomic data, like genomic signature research, together with clinical pathological data can help us in this sense.