Skip to main content
Photo of Dr. Dennis Ko

Dr. John O. Parker Lectureship: Using Big Data to Improve the Primary Prevention of Cardiovascular Disease (Presented By: Dr. Dennis Ko)

By Isaac Emon, MSc Candidate and TMED 801 Student

 

On Thursday, November 17th, the Department of Medicine had the privilege of hearing from Dr. Dennis T Ko, MD, MSc, during the Dr. John O. Parker Distinguished Lectureship. Dr. Ko is a Full Professor in the Department of Medicine at the University of Toronto and a senior scientist and interventional cardiologist at Sunnybrook Hospital.

 

Dr. Ko’s presentation discussed the use of big data in improving the primary prevention of atherosclerotic cardiovascular disease (ASCVD), focusing on calibrating traditional cardiovascular risk scores. He began by discussing the heterogeneity of treatment efficacy, explaining how not all patients respond equally to treatment. While some patients may benefit greatly, others may experience a negative effect, emphasizing the importance of developing accurate risk scores to ensure the appropriate use of therapies. He also described the relationship between relative and absolute risk reduction, demonstrating that those with higher risk should receive the greatest benefit from treatment. Dr. Ko discussed the “Treatment Risk Paradox,” describing that treatment should align with the magnitude of benefit patients will receive but that treatment patterns are often the opposite in that lower-risk patients are more likely to receive treatment (1).

 

This is mind, we shifted to discuss the guidelines for primary prevention of ASCVD in Canada. We reviewed the steps for calculating the Framingham Risk Score (FRS) and examined deficiencies in this approach (2). In 2020, Dr. Ko and colleagues assessed the calibration and discrimination of the FRS and the Pooled Cohort Equations (PCE) in Ontario to uncover the validity of these risk scores in different subgroups (3). They found a substantial overestimation of the FRS in Ontario, with the overall predicted rate [of developing ASCVD within 5 years] being 5.8%, while the observed rate was found to be only 2.9% (3). An overestimation was also found using the PCE in Ontario (3). This overestimation of risk has substantial consequences including unnecessary expenditure of resources and harmful side effects and toxicities in patients treated inappropriately, reminding us to be mindful when creating management plans. Dr. Ko has worked to calibrate these risk scores using more up-to-date statistics regarding outcomes with smoking, diabetes, and other risk factors. He demonstrated that a recalibrated FRS was more accurate but still overestimated risk (4).

 

Dr. Ko also reviewed the difficulties of using these risk scores in a clinical setting. To adequately calculate a risk score, the physician needs to understand a detailed history, measure accurate vitals, obtain blood tests, then use a calculator to determine the level of risk. Though these may sound like routine practices, the COVID pandemic and the strict time schedules clinicians follow decrease the simplicity of calculating these scores. In fact, Dr. Ko mentions that only 50% of clinicians in the U.S. use these risk scores. Ultimately, Dr. Ko aims to reduce the physician’s role in obtaining data and is working to find ways to provide already-calculated risk scores to them directly. He describes that by using big data from the Ontario Laboratory Information System (OLIS), relative ASCVD rate can be compared to a variety of clinical and laboratory measures to help predict 5-year risk. This method shows promising accuracy in aligning with observed risk and having clinical utility. Future research should continue using large data sets like OLIS to further our translational research into clinical prevention for ASCVD, but it is important to realize that the use of big data extends far beyond cardiology and its application has value in improving patient care in all disciplines.

 

After the MGR, Dr. Ko offered his time to the TMED 801 class where students were able to ask questions about how his research directly impacts patients and how the topic of ASCVD prevention is often represented in the lay press. We further discussed the use of big data sets in research to improve primary prevention of ASCVD, considered how geographical and socio-economic barriers can impact patients at risk, and contemplated the pros and cons of social media in promoting cardiovascular health. The latter end of the discussion focused on Dr. Ko’s career path and how his medical and educational journey has unfolded thus far. He highlighted how helping patients has driven his passion for furthering the field of ASCVD prevention, emphasized the value in doing what you love, and stressed the importance of always ensuring you find time for yourself and your family.

 

On behalf of the Department of Medicine and the TMED 801 class, I would like to thank Dr. Ko for taking the time to teach and inspire us.

 

 

References:

 

  1. Ko, Dennis T., Muhammad Mamdani, and David A. Alter. "Lipid-lowering therapy with statins in high-risk elderly patients: the treatment-risk paradox." Jama 291.15 (2004): 1864-1870.
  2. Pearson, Glen J., et al. "2021 Canadian Cardiovascular Society Guidelines for the management of dyslipidemia for the prevention of cardiovascular disease in adults." Canadian journal of cardiology 37.8 (2021): 1129-1150.
  3. Ko, D. T., Sivaswamy, A., Sud, M., Kotrri, G., Azizi, P., Koh, M., ... & Anderson, T. J. (2020). Calibration and discrimination of the Framingham Risk Score and the pooled cohort equations. CMAJ192(17), E442-E449.
  4. Sud, Maneesh, et al. "Population-based recalibration of the Framingham risk score and pooled cohort equations." Journal of the American College of Cardiology 80.14 (2022): 1330-1342.

 

 

Comments

Name
Nicole Morris

Wed, 11/23/2022 - 15:55

Hi Isaac,

Thanks for an excellent summary of Dr. Ko’s talk last Thursday. You mention that the COVID-19 pandemic may have impacted the ability of physicians to adequately calculate risk scores. Additionally, Dr. Ko stated that only 50% of clinicians in the U.S. are currently using these scores. Do you know if this number was any higher in the pre-COVID era? Are there any other, perhaps more pressing barriers that are impeding risk score calculation?

Looking forward to your response,

Nicole

Name
Nicole Morris

Hi Nicole,

Thanks for your question. I'm not too sure if this number was higher in the Pre-COVID era. However, I found an article (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4263206/) that looked at factors influencing a physician's use of risk scores and it seems that a lack of trust for their value is a major reason for this. In this study, physician's who were forced to measure cardiac risk scores were often found to not even use them during treatment decisions. Dr. Ko also discussed in his talk that it simply takes time and effort to use these scores and many physicians do not put the time in to calculating them. I think that this belief of 'lack of validity', as well as the time it takes to measure these scores, may both have an impact on usage! Dr. Ko and others are working on ways to improve the simplicity and efficiency of using these scores.

Isaac

Name
Isaac Emon

Name
Tarrah Ethier

Fri, 11/25/2022 - 15:25

Hi Isaac,

You have put together a great blog post, detailing Dr. Ko's lecture and our seminar. Dr. Ko touched on the benefit of using big data from a human research perspective. As you and I both work on pre-clinical animal models, I was wondering your thoughts on the feasibility of introducing big data in this area? Pre-clinical and clinical work are very different so the benefits that Dr. Ko outlined for clinical data may not even be relevant to pre-clinical data. However, I think it is worth exploring!

Looking forward to hearing your thoughts!

Tarrah

Name
Tarrah Ethier

Hi Tarrah,

This is an excellent point. Having human data is very useful to drive the research forward. I know for my research we are combining my animal models with human biobank data to identify prevalence in patients. Maybe there would be a way for you to incorporate "big data" down the road as I think it could be extremely beneficial for comparing sex differences like you are! Feasibility is another question - it is much more difficult to obtain the data than it is to just use data that already exits in patient records. If there was a way you could access large patient databases than maybe sex differences in human models could be investigated alongside your murine models. Happy to chat further!

Isaac

Name
Isaac Emon

Name
Samantha Delios

Sat, 11/26/2022 - 13:47

Hi Isaac!

You did a great job summarizing Dr. Ko's talk last Thursday. As Nicole and you both mentioned, only about 50% of clinicians in the U.S. are currently using these scores. Dr. Ko also mentioned that many aspects to assess these scores have not been updated since the 70s or 80s. Do you believe there is a link between people potentially not using these scores and how they are assessed? I was wondering what your thoughts were, and if you could comment on the benefits vs the cons of using these scores. I am looking forward to your thoughts!
Overall, great job!

Thanks,
Sam

Name
Samantha Delios

Name
Isaac Emon

Thu, 12/01/2022 - 20:47

In reply to by Samantha Delios (not verified)

Hi Sam,

I'm not sure I quite understand the first part of the question but I do believe that the research in which our data is obtained my be skewed. The physicians that fill out these surveys and answer questions regarding their risk score use may be the ones who are more likely to use these scores. Therefore, if anything, I think the 50% may be an over-prediction. The benefits to using these scores are that they offer us information that can be used to make pro-active treatment and prevention plans. They can offer information about whether patients should be started on treatment, and if they should be seeing a dietician, exercising more etc. The cons are that they are categorical and that like anything in medicine, treatment should be focused on treating the patient, not the disease. Therefore, these risk scores are not going to predict each individual perfectly and may lead to inappropriate treatment use. The other con is feasibility - physicians often don't take the time (or may not have the time) to calculate these scores. Studies are working to better implement these scores in a more user-friendly and accurate way.

Isaac

Name
Isaac Emon

Name
Jana Livingston

Mon, 11/28/2022 - 22:33

Hi Isaac,

Excellent summary of Dr. Kos presentation which looked at evaluating current cardiovascular risk assessment scores including the associated criteria, overestimates, and means to improve accuracy. These errors are not only detrimental for patients who may be prescribed medications prematurely but can hinder patient confidence in these risk assessments. Importantly, as you noted, there are various factors that contribute to the overestimation of cardiovascular risk scores, including outdated statistics. In your opinion, what other factors can influence the accuracy of these scores? Likewise, you mentioned socio-economic status, how might this be taken into consideration when formulating risk assessments?

Looking forward to hearing your thoughts!

Jana

Name
Jana Livingston

Hi Jana,

Great questions. I think many factors influence these scores including patient honesty and recall, as well as lifestyle and family history. Some factors require patient history information which can often be incorrectly regurgitated by patients or tracked by physicians, leaving room for human error. Lifestyle also plays a huge factor. Smoking and exercise habits are important factors for cardiovascular risk and are not always consistent from year to year, leading to misinterpretation with regards to health. Family history is also often recorded but many individuals are unsure about family history and may not be able to provide accurate information about previous disease in their family, further leading to missed information. You bring up a good point with regards to socioeconomic status. These patients may not have the same access to resources that others have, or may be exposed to environmental or physical contributors to poor cardiovascular health. It is important to keep these ideas in mind when predicting risk, discussing lifestyle changes with patients, and creating management and prevention guidelines.

Isaac

Name
Isaac Emon

Name
Maria Korovina

Mon, 11/28/2022 - 23:40

Hi Isaac,

Thank you for sharing your blog post! It is clear how much benefit big data may have when predicting 5-year risk. You also mentioned that obtaining adequate risk scores takes a number of time-consuming practices. Furthermore, you highlighted that 50% of physicians do not use these risk-scoring techniques, but many still do. Do you think there might be any hesitancy in implementing a new risk scoring system that eliminates the need for historically used scoring methods? Could this hesitancy come from not only the physicians but maybe from patients that insist that they still want a blood test done "just in case" instead of relying on large data sets?

Best,

Maria

Name
Maria Korovina

Name
Isaac Emon

Thu, 12/01/2022 - 21:14

Hi Maria,

You bring up a good point. I think that there is sometimes hesitancy in makes changes to procedures and guidelines that are often very successful, sort of the "If it ain't broke, don't fix it" idea. However, as we discuss a system that is not well used and does not appear to be the most practical in clinic, I think that there is less hesitancy about a situation like this. I think we can still learn from our previous methods and use the pros and cons when identifying a more usable risk score technique. I do totally agree with the patient idea. I think that people will always want quantitative medical tests and results, and I don't believe that this hinders us from developing accurate cardiovascular risk assessments, but should instead be used in conjunction with novel techniques to optimize our understanding of patient risk.

Isaac

Name
Isaac Emon

Add new comment

Plain text

  • No HTML tags allowed.
  • Lines and paragraphs break automatically.
  • Web page addresses and email addresses turn into links automatically.