Peek Behind the Paper: statins, Type 2 diabetes and real-world evidence

Written by The Evidence Base

We peek behind the paper with first study author Victoria Zigmont (Southern Connecticut State University, CT, USA) to better understand how real-world evidence was employed to investigate whether an increase in the risk of developing diabetes could be a statin side-effect.

Interview segments:

Please can you introduce yourself, your institution and your research?
What prompted you to study the association between statin use and new-onset Type 2 diabetes?
What was the rationale behind some of the screening criteria you employed to define your study cohort?
Can you explain the importance of using real-world evidence (RWE) to the outcomes of your study; what could this provide that randomized controlled trial (RCT) data could not?
Because of the nature of the study, the conclusions that you can draw from it are correlational; what are you cautious of when assessing correlational data?
What may be the implications of your study be on the use of RWE in decision making and clinical practice?
Do you have further RWE-based studies planned — such as for assessing the links between different classes of statins and diabetes risk?


Please can you introduce yourself, your institution and your research?

I am an Assistant Professor in the Department of Public Health at Southern Connecticut State University (CT, USA). Prior to taking up this position I obtained a Masters in public health from Southern Connecticut State University, and a PhD in public health from The Ohio State University (OH, USA). My PhD concentration was in epidemiology, with a minor in biostatistics.

Back to top of page


What prompted you to study the association between statin use and new-onset Type 2 diabetes?

There has been a significant amount of research concerning the association between patients’ statin use and their risk of developing diabetes. This led to, in 2014, the introduction of a written warning on statin pill bottles indicating a side effect of taking statins could be a higher risk of individuals developing diabetes. I was interested to see, using a large dataset covering a population with insurance, if indeed individuals’ diabetes development was impacted by their use of statins, or, if actions were being implemented within the healthcare system to prevent this from happening. Such actions could have included physicians getting patients to stop taking statins, or recommending weight loss, exercise or engaging in healthy eating behaviors and patterns to prevent patients developing diabetes.

Back to top of page


What was the rationale behind some of the screening criteria you employed to define your study cohort?

I employed a retrospective, cohort study design as I was compiling data for analysis from many several different sources — from pharmaceutical claims, medical claims, medical encounters, biometric screenings and a health survey. My rationale behind specifying criteria for patient inclusion in my study population was to mitigate the effects of any confounding factors that could influence the study findings as much as possible.

I also wanted my study design to ensure that I was comparing alike patients and making fair comparisons between my patient groups. For example, one of the requirements for patient inclusion in the study was for patients to have had indications in their electronic medical records, sometimes also referred to as electronic health records, for use of a statin. That could have taken the form of a patient having already suffered an adverse cardiovascular event — though this was very infrequent in my cohort — or a patient having an indication such as hypertension or hyperlipidemia. Comparing these patients with others who are relatively healthy, with no chance of being prescribed statins, would not have been a fair comparison; I would have been comparing individuals who were at a greater risk of experiencing adverse cardiovascular events — that is why they are being prescribed a statin — with individuals who are completely different.

Another measure I took to ensure that I was comparing ‘apples to apples’ was that I required enrollees to have made more than one medical claim during the study enrollment period. The rationale for this was such that patients who are sicker are likely to go to the doctor more; I did not want anyone in the study who had not made a medical claim because that would mean that they were likely not seeing their doctor during the study period. and I would therefore not have been able to use those patients’ medical claims to measure factors from. This would have introduced a measure of bias and meant that a fair comparison between patient populations would not have been made.

Furthermore, I excluded individuals from study inclusion if they had already been diagnosed with diabetes prior to the start of the study. I was interested in looking at diabetes development; you cannot be at risk of developing diabetes if you already have diabetes. There is existing research concerning the use of statins amongst diabetic patients, however, I was interested in investigating the risks for patients who were taking statins for either primary prevention of cardiovascular disease — or secondary prevention of cardiovascular disease — but were free of diabetes at the start of the study. Cardiovascular diseases and events in the study included hypertension, stroke, ischemic heart disease and acute myocardial infarction, or heart attack.

Back to top of page


Can you explain the importance of using RWE to the outcomes of your study; what could this provide that randomized controlled trial (RCT) data could not?

The benefit of employing RWE is that it allows us to observe real-world patients, real-world behaviors and real-world prescribing patterns of physicians. Furthermore, patients who sign up for a RCT may be different from everyday patients. Oftentimes people are enrolled in RCTs based on the presence of specific clinical characteristics or risk factors. Patients who are interested in enrolling in RCTs are also often different from the general population, because they signed up to be in the study.

One of the benefits of the study design of RCTs is that they allow you to ensure that all study participants are truly comparable with respect to baseline risk factors. There is a lot of control that goes into ensuring this before the trial even starts, such as blinding and randomization; the patients may not know that they are taking active medication. This can mitigate the effects of potential confounding factors, as they can be randomized against.

The problem with RCTs, however, is that they are highly expensive; often they are not financially feasible to conduct. Second, RCTs may be unethical. The benefits of statins for patients’ cardiovascular health have been very well established; ethically, we cannot deny patients the standard of care — prescribing them a statin to prevent a heart attack occurring.

Back to top of page


Because of the nature of the study, the conclusions that you can draw from it are correlational; what are you cautious of when assessing correlational data?

There is always the chance of residual confounding factors affecting the study outcomes; these are things that could have played a role in the study that we didn’t directly examine, or, things that are simply out of our control. The hope is that as more research is conducted using correlational data or cohort studies, we are able to see consistencies across studies. When evidence keeps building — when we keep seeing the same things happening — then we can be more certain that an observed association between two factors may be causal, not just correlational.

Back to top of page


What may the implications of your study be on the use of RWE in decision making and clinical practice?

I think that one of the big strengths of my study is that I used population data from the real-world health care system to assess the risk for developing diabetes amongst statin users compared with non-users. This data can now be used by health care providers and patients to identify proper courses of clinical treatment. I am not a health care provider, so that would be beyond the scope of this research, but my hope would be that health care providers will see these results and will start to consider them in the context of individual patients’ risk factors and use them to make the best decisions for patients.

Back to top of page


Do you have further RWE-based studies planned — such as for assessing the links between different classes of statins and diabetes risk?

I currently do not have any such upcoming studies planned. However, if there were additional data available, I would definitely be interested in assessing the implications of different statin classes —and their doses — on the risk of developing Type 2 diabetes. A benefit of employing electronic medical record data is that it is relatively cheap; I did not have funding to do this study but was able to just pull the data out of the databases and analyze it. However, a challenge with this is that, in order to look at specific differences in patients based upon statin class or dose, you would need to have a much bigger data set. You would need to know that you have enough people who are taking those different statin classes and doses, in order to have enough people in each group to make legitimate comparisons. In my study sample, it seemed that there were not that many differences between what participating patients were being prescribed; I was therefore unable to examine those differences in my study.

Back to top of page


Disclosures:

Zigmont reports no financial or competing interests to disclose.