One of the most physiologically complicated yet overly simplified “set it and forget it” protocols in emergency medicine just might be the acute treatment of hyperkalemia. We all have the protocol in our head and can write it up in our sleep (hopefully) but how effective is it? How safe is it? How evidenced based is it? So many questions so little time. To think that we can still improve on this time-tradition sequence tells me that we still don’t, “K”-now it all.. Mmm .. “K”…. (see what I did there? Lol)
Today’s Quick hit article is on this topic of how much insulin is needed in the treatment of hyperkalemia in dialysis patients. I think this may be the only method lacking, retrospective, observational, chart review that has or ever will change my practice (with all due respect to the authors). This is a before and after study of the amount of insulin given to patients who are hyperkalemic AND have ESRD (end stage renal disease).
McNicholas. Treatment of Hyperkalemia With a Low-Dose Insulin Protocol Is Effective and Results in Reduced Hypoglycemia. Kidney Int Rep. 2017 Oct 24;3(2):328-336.
The Bottom Line:
This was a single center before and after trial of an intervention to decrease the amount of hypoglycemia by decreasing the amount of regular insulin given IV to treat hyperkalemia. The big flaw here? This study lacks methods necessary for statistical rigor (see Table 1). The treatment of hyperkalemia with the use of insulin in ESRD patients may be decreased from 10 units of regular insulin IV down to 5 units of regular insulin IV. This decreased the amount of hypoglycemia (12% vs 6%) induced while still appropriately lowering the potassium by 0.5 to 1 meq/L. Despite the high risk of bias and very small sample size used, it seems to make sense to lower the insulin dose in ESRD patients to 5 with minimal(?) harm. I will continue to order frequent point of care glucose checks since hypoglycemia was not abolished. I will also repeat the K at appropriate intervals in the case that a repeat dose of insulin or other treatment is needed. Therefore, although I don’t think this study is evidence for using lower insulin in ESRD patients I do think they came up with the right answer anyway.
The Shout out:
I want to give a big shout out to an amazing mentor Dr. Paul Blackburn. We discussed this article and he told me this has always been his practice. Dr. Blackburn never stop teaching, you have and always will have something to teach us young (and not so young) docs! Thank you!
This was a single center, before and after trial, of an intervention to decrease the amount of hypoglycemia by decreasing the amount of regular insulin given to treat hyperkalemia. The study was performed in an urban academic ED in Seattle. They noticed a significant amount of hypoglycemia, including hypoglycemia to <40 in the ESRD patients. Therefore, they undertook an intervention using education to recommend lowering the dose of IV insulin given form 10 units down to 5 units. They called the pre-intervention “audit 1” and the post intervention “audit 2”. We will get into the details but first let’s go over the methods using my chart review checklist (See my post on Chart reviews). By looking at Table 1 you can see that very few of the methods necessary to validate a chart review were followed. Therefore, this chart review is at extremely high risk of bias. Fortunately, the authors do mention a number of significant limitations. One noteworthy limitation is the very small numbers. What is strange is the number of cases of hyperkalemia decreased in the after period for an unknown reason. Another issue is that some of the patients weren’t treated with insulin at all! They also noted that although compliance with their new protocol of decreased amount of insulin improved (25% getting 10 units vs 2% getting 10 units) they still did not have complete compliance. The way they organized their data actually makes it quite difficult to discern differences. In fact you can’t really tell how many of the ESRD patients were hypoglycemic in audit 1 vs audit 2. Nor can you tell in audit 2 if the K dropped to an appropriate level. A sample size calculation was not given so we don’t even know if these numbers were either clinically or statistically significant. The most glaring example is when they say that they noticed in the pre-audit a “large” amount of hypoglycemia. However, they note that “There was a trend for ESRD patients who were treated with 10 units versus 5 units of insulin to develop hypoglycemia (9 of 32 patients [28%], 5 units, vs. 6 of 11 patients (54%), 10 units i.v. insulin, P = 0.1).” So is this a even a real difference? Clinically it sounds like it however they state there were “fewer patients being treated for hyperkalemia between the 2 audit periods”. There are certainly a lot of holes in this swiss cheese. On the other side however, this seems like it is an intervention that makes sense and in their small sample size they did have few cases of hypoglycemia and still a reduction (I think?) in the K. Therefore, even though this study isn’t evidence for using lower insulin in ESRD patients they came up with the right answer anyway.
|1. Investigator bias|
|a. Question appropriate for chart review?||Yes|
|b. Financial/Intellectual disclosures supplied?||Yes|
|2. Charts bias|
|a. Methods of Chart Identification (CC vs ICD-10)||No|
|b. Sufficiently sampled?||Unknown|
|c. Was a power calculation supplied?||No|
|d. A priori inclusion criteria?||No|
|e. A priori exclusion criteria?||No|
|f. Table of clinical characteristics?||Yes|
|g. Flow diagram delineating how the study population was derived?||Abbreviated|
|3. Data bias|
|a. Defined A priori?||No|
|b. Coding guide for abstractors?||No|
|c. Coding guide provided?||No|
|d. Standardized Data Collection tool (DAT)?||No|
|e. Was the DAT pilot tested?||No|
|f. Was the DAT provided?||No|
|g. Is there missing or conflicting data?||Unknown|
|h. How is missing data handled (sensitivity analysis)?||unknown|
|4. Abstractor bias|
|a. Blinded to study hypothesis?||Unknown|
|b. Trained appropriately?||Unknown|
|d. Inter-rater Reliability?||Unknown|
|e. Intra-rater reliability?||Unknown|
|5. Reliability bias|
|a. Kappa and percent agreement calculated for the data?||No|
|b. What level of reliability and why was that level chosen?||Unknown|
|c. Which of the collected variables were checked for reliability?||Unknown|
|d. What percent of the data was checked for reliability?||Unknown|