Sodium-Glucose Cotransporter-2 Inhibitors and the Risk for Diabetic Ketoacidosis.
J. Falatko D.O.
Mathematics and statistics are tools that use to help us understand the world around us. Statistics can help medical researchers identify differences, trends, and truths that direct patient care. With each passing generation they become more and more sophisticated. With this sophistication leads to expertise, which can lead to a very specific lens to view datasets through.
A savvy statistician and an even savvier investigator can analyze a dataset and spin it to find what they are looking for, without really considering the significance of their findings.
I am not some scientific nihilist. There is good research being done and there are some truths that are being discovered. That doesn’t mean there’s no gamesmanship happening out there. Physicians and researchers in academic medicine get paid to publish. There is a common catch phrase in academic medicine “publish or perish.” And you can perish while publishing, if you publish studies no one cares about. In fact, most studies in which the result is not significant are unlikely to be read, so they aren’t published. The goal is to achieve significance. Investigators need to publish and journals want significant results to attract readers. Everyone’s incentives are aligned.
The study presented here is a good example of the importance of keeping statistics in perspective. The authors analyzed a huge database, had a decent hypothesis and found a significant association with harm. It was published in a prestigious journal. But some studies because of scope and size have statistical significance, but lose clinical significance. Which can be lost if interpreting the study in a vacuum.
A quick review of Sodium Glucose Co-transporter-2 Inhibitors (SGLT-2). These medicines treat Type 2 Diabetes. They bind to a channel in the kidney that allows blood sugar to pass into the urine and excreted. The result is a lower blood sugar level. They have been found to have several other benefits besides lowering blood sugar, which I will discuss below. SGLT-2s have some downside. Urinary tract infections, frequent urination, kidney injuries, and although rare, there is a higher risk of limb amputation with these medications.
Let's dig into the study.
This was a retrospective matched cohort. It was completed using the database from the Canadian nationalized healthcare system. In this type of study, the database is searched for patients taking an SGLT-2. Once they are identified they are then matched with other similar diabetics that are not treated with an SGLT-2. The endpoint in this study was the occurrence of a condition called diabetic ketoacidosis or DKA.
DKA occurs when the body can no longer utilize insulin, and therefore cannot metabolize sugar. To get energy it burns fat, muscle, and lactic acid which leads to ketosis, which leads to high blood acidity, which can be life-threatening (although death in adults due to DKA is rare).
I don’t care much about the validity of the study in this case because I found the results section far more interesting.
Results:
The study size was large. There were 215,000 new SGLT-2 users eligible for inclusion. They matched them to 208,000 new DPP-4 users (you don’t need to remember DPP-4, these are a different class of diabetes medication chosen arbitrarily by the investigators as a control group). So, there were 208,000 patients in each arm. Their matching protocol was successful since they did not identify any significant differences between the two groups. Patients were followed for a total of 370,500 person years (mean 0.9 years/patient)
If these seem like huge numbers, that’s because they are. They also represent the massive scale required to obtain the result. It kind of reminds me of using the Hubble telescope to look at the surface of the moon because you want to see the individual grains of sand.
There were 372 episodes of DKA in the SGLT-2 group vs. 133 in the control group. The incidence rate was 2.03 per 1000 person years vs. 0.75 per 1000 person years for SGLT-2 group and control group respectively. This is close to a 3-fold increase in the incidence of DKA. The confidence interval for the hazard ratio was (1.99-4.08).
To put 1000 person years in perspective, this would be the equivalent of observing 10 people for 100 years or 100 people for 10 years. If you were to give 100 patients an SGLT-2 for 10 years 2 of them would have 1 episode of DKA.
What does all of this actually mean? How significant is this risk? One standardized statistic used to help summarize the significance of the findings is number needed to treat (NNT), or in this case the number needed to harm (NNH). This number tells a physician the number of patients that need to be on this medicine to cause one event. The equation is below.
NNH = 1 / absolute risk reduction
NNH = 1 / 0.00116 = 862
To interpret this number, you have to treat 862 people with an SGLT-2 to cause one episode of diabetic ketoacidosis. Honestly, that’s a lot.
Besides lowering blood sugar SGLT-2 inhibitors have some other benefits. They can help diabetics lose weight, they help prevent heart attacks and strokes. In patients with heart failure they can keep them out of the hospital even if they don’t have diabetes. Let’s look at some NNT’s for these endpoints.
For weight loss, 33% of patients in one study were found to have lost > 5% of their body weight while taking an SGLT-2 compared to 2.5% for the control arm.
NNT = 3.3
For cardiovascular events, two studies have shown benefit. The first showed an absolute risk reduction of 1.6%, the second showed an absolute risk reduction of 2.6%.
1) NNT = 62.5
2) NNT = 38.5
The truth is probably somewhere close to 50.
For patients with heart failure, the DECLARE-TIMI 58 trial showed an absolute risk reduction of 0.8% for hospitalizations due to heart failure.
NNT = 125
This result was confirmed by a meta-analysis which showed a relative risk reduction of 31% for heart failure hospitalizations. I couldn’t calculate an NNT because the raw numbers were difficult to find, but the confidence interval was narrow (0.61-0.79) suggesting there is decent benefit.
To summarize, for every episode of DKA that is caused, 261 patients become less obese, 17 cardiovascular events are prevented, and there are 7 less admissions to the hospital for heart failure.
Now, I know it is important to discuss risks of medications, but that should be placed in perspective with the benefits.
The authors got a good publication out of this. It was published in Annals of Internal Medicine, which is a highly reputable journal. It will likely be featured in ACP journal club. Many physicians will read this and I imagine it will get some press. Certainly, a new warning will be added to the package insert. There is already a commercial on daytime television from a law firm warning patients about the newly discovered harm. But did they find anything of real clinical significance? Or just fodder for medical malpractice?
They described the result as robust proof for the risk of DKA. I would argue that it is robust… statistically, but irrelevant clinically. For adults, the in-hospital fatality rate for DKA is 0.4%. That would be 3 patients in this study. Based on my experience these deaths are due to electrolyte imbalances that could be corrected with proper DKA protocols. The cardiac events and heart failure hospitalizations prevented significantly outweigh this potential harm.
So, if a patient of mine ever sues me for failing to warn them about the risk of DKA, I hope they will at least thank me for the lower blood sugar levels, the 10 lbs they lost, the heart attack I prevented, and the improvement in their heart failure. To think, they wouldn’t even need the medicine in the first place if they just changed from 6 cans of regular mountain dew a day, to diet mountain dew.
Link: https://www.acpjournals.org/doi/10.7326/M20-0289