Evaluating the Effectiveness of PREDICT Breast v3 in Breast Cancer Prognosis

Breast cancer is the most common cancer diagnosed in women worldwide. In 2022, about 2.3 million new cases came to light. In the US, breast cancer is the top type of cancer among women, with high rates of both diagnosis and deaths. It’s estimated that in 2024, there will be over 310,000 new cases and around 42,000 deaths due to breast cancer. When women receive a breast cancer diagnosis, one of the big decisions they face is whether to undergo systemic treatment after surgery.

#What Is Systemic Treatment?

Systemic treatment refers to therapies that work throughout the body. For early-stage breast cancer, this kind of treatment aims to lower the chances of the cancer coming back and to improve survival rates. Having accurate predictions about survival rates and the benefits of treatment is crucial. It helps doctors recommend the right approach to treatment, reducing side effects and keeping the patient's quality of life intact.

#Prediction Models

Several prediction models were created to help doctors choose the best systemic treatments based on patient and tumor details, like tumor size and hormone receptor status. Some of these models, like PREDICT Breast, have been updated regularly. The most recent version, PREDICT Breast v3, was released in May 2024.

However, not all models are available anymore. For instance, Adjuvant! Online is no longer accessible, and CancerMath hasn’t been updated for a while. PREDICT Breast has shown its mettle through updates since its launch in 2011.

#Importance of Validation

PREDICT Breast v1 and v2 have been validated in many countries, including the UK, Canada, Japan, and others. However, PREDICT Breast v3 had only been validated in the UK, the country where it was developed. This is a bit like baking a cake but only tasting it in one kitchen. To check if it’s good everywhere, we need to try it in different places, like the diverse communities in the US.

This study aims to fill that gap by validating PREDICT Breast v3 using the latest data from the Surveillance, Epidemiology, and End Results (SEER) program. This will help assess how well the model predicts patient outcomes across various American groups.

#Study Population

The study looked at a vast dataset from SEER, covering breast cancer cases from 2000 to 2018. SEER collects extensive information about cancer patients across the country. In total, they recorded over 1.2 million cases. The study focused on women aged 25 to 84 who were newly diagnosed with breast cancer within that time frame. The researchers excluded anyone who already had advanced cancer at diagnosis or those with insufficient data. This left a solid group of around 628,000 women, representing different ethnic backgrounds.

#Key Factors in the Model

To make accurate predictions with PREDICT Breast v3, the model uses a set of input variables. These include details like the patient’s age, the size of the tumor, and whether certain hormone receptors are present. The model also considers different treatment types, such as chemotherapy and radiation. Some treatment data isn't available, leading to assumptions about certain patient groups. For instance, the model assumes that all younger patients who received chemotherapy got a specific type.

#How Predictions Are Made

For predicting survival rates, the researchers used a special model that takes various risks into account. This includes both breast cancer-related mortality and other causes of death. The process essentially combines different factors to generate a survival estimate for each patient.

#Evaluating Model Performance

To see how well the model performed, researchers looked at three key areas: calibration, goodness-of-fit, and discrimination.

• Calibration checks if the predicted number of deaths matches the actual deaths.
• Goodness-of-fit is like comparing how the predicted deaths stack up against actual deaths across different risk categories.
• Discrimination assesses how well the model can distinguish between patients who will survive and those who will not.

A good performance in these areas means the model is reliable for patient decisions.

#Characteristics of Patients in the Study

Among the 628,753 women, most had hormone receptor-positive types of breast cancer, which generally has a better outlook. The researchers gathered information on their demographics, tumor characteristics, and treatment types.

The predictions made by PREDICT Breast v3 showed a good match with the actual observed deaths after 10 and 15 years. However, there were some issues with certain ethnic groups. For example, the model tended to overestimate deaths in non-Hispanic Asian women with a specific type of breast cancer, while underestimating deaths in non-Hispanic Black women. It’s a bit like predicting that everyone would love the same flavor of ice cream when everyone has different tastes.

#Highlighting Model Strengths and Weaknesses

Overall, PREDICT Breast v3 fared well, especially for the majority of breast cancer patients in the US. It provided good predictions for both ER-positive and ER-negative patients. In general, the model performed similarly to previous Validations done in the UK.

However, the hiccups in accuracy for specific groups were notable. The model overestimated survival in some groups and underestimated it in others. This kind of discrepancy can lead to major miscalculations when doctors make treatment decisions. For instance, it might lead to prescribing a treatment that isn’t as effective for a certain patient group or overlooking an effective one.

#Future Enhancements

While PREDICT Breast v3 is impressive, there are ways to make it even better. It looks like adjusting the model to account for population specifics in terms of baseline hazards could improve its accuracy. Also, including more prognostic markers, like various gene expressions or genomic risk scores, could lead to richer predictions.

The objective is to ensure that the model works well for all groups, reflecting the diversity of the population accurately, rather than a one-size-fits-all solution.

#Conclusion

In summary, breast cancer is a major health concern for women worldwide, with significant focus required on the right treatment options. While prediction models like PREDICT Breast v3 play a vital role in guiding treatments, it’s essential to ensure their recommendations are well-matched with diverse patient backgrounds.

The research confirms that PREDICT Breast v3 has solid principles and could really help. However, continuously improving it to acknowledge the unique characteristics of different ethnic groups will ensure that all women fighting breast cancer get the best possible care. After all, we all want to win this tough battle, and every detail counts when it comes to saving lives.