Kidney Beans: Renal Function and Drug Dosing (Part 1)
Editor’s Note: J Nile Barnes, PharmD, BCPS is an experienced clinical pharmacist and paramedic. He has over 30 years of experience in healthcare, including more than 20 years teaching paramedic and pharmacy school.
This particular blog set comes from years of teaching pharmacist interns about drug dosing and renal function testing. This topic is a standard topic on every APPE rotation he teaches. He has also presented this topic at a national clinical nursing convention.
Basically, we’re all lucky to be able to learn from his formidable brain!
The following is the first in a three-part series on renal function and drug dosing. This first part will discuss the techniques for assessing renal function, Part 2 will discuss the mechanisms for acute and chronic kidney dysfunction, and finally Part 3 will give some practical pointers for drug dosing with decreased renal clearance.
So let the learning begin!
Preceptor: What’s his renal function like?
Student: 47.72 mL/min
Preceptor: Really…47.72, not 47.73 mL/min?
What just happened here? The student dutifully calculated a creatinine clearance (CrCl), and her calculator gave her 8 digits. So she rounded it to two decimal places, like she always does. Why is the preceptor sarcastic?
Frankly, it’s because the estimate is not that precise or accurate. Let’s talk about why that is.
CrCl has been used as a surrogate marker for glomerular filtration rate (GFR) for decades. In the early 1970s, two physicians, Drs. Donald Cockcroft and Henry M. Gault in Montreal, Quebec, devised a formula that predicts CrCl using serum creatinine. In their paper, Cockcroft and Gault looked at 249 male patients with reproducible paired serum creatinine levels. These 249 pairs were reduced to seven points which could be described by one equation:
The results of this equation correlated well with measured urinary creatinine clearance. And thus, the Cockcroft-Gault equation became gospel.
(No disrespect to the authors…using it improperly is our problem, not theirs.)
The problem is…. Well actually there are several problems. But they are primarily based on the assumptions that go along with the use of the equation and the differences between those assumptions and our current patients.
Caveat #1: The calculated CrCl from Cockcroft-Gault is an estimate. It is not the patient’s clearance, merely our estimation of what it should be based on patient parameters. Along with that, clearance varies during the day, such that a 24-hour measured CrCl is actually an average. So we have an estimate of an average of a varying CrCl.
Caveat #2: The data from the study is a bit cherry picked (albeit reasonably so). Still, it affects our confidence in the calculation. If we’re being honest (and of course we are because we’re tl;dr), there were actually 505 patients in the study group, but only 249 had reproducible serum creatinine levels. The other half did not. So the lesson here is that the equation is only applicable when the serum creatinine is stable.
(Needlepoint this on a pillow, engrave it in stone, whatever it takes to log this MAJOR point away. It’s even so important that it bears repeating on a blog that’s all about being concise and to the point.)
Cockcroft-Gault is only remotely applicable in stable renal function situations.
There have been equations developed for unstable renal function, but they are difficult to use and have limited validation.
Caveat #3: Cockcroft and Gault were at a veterans’ hospital. In the mid-1970s, the vast majority of Canadian veterans were male. The 0.85 factor for female patients was not part of the original study. “Typically,” women have 85% of the muscle mass of a male patient of an equal weight, and since creatinine is a breakdown product of muscle, it was reasoned that the clearance of a female’s creatinine should only be 85% of a male’s. Of course, this is not always true!
Caveat #4: Another issue to think about is the laboratory method of measuring creatinine. When Cockcroft and Gault collected their data, creatinine concentration was determined by a “wet chemistry” method known as the alkaline picric acid method. In the 21st century, the method used by the vast majority of commercial clinical laboratories is the IDMS method (Isotope Dilution Mass Spectroscopy method).
The older picric acid method uses a colorimetric technique to determine the concentration of creatinine in a sample. Other non-creatinine chromogens (e.g., acetone, acetoacetate, pyruvate, ascorbic acid, glucose, some cephalosporins, barbiturates, and some proteins) can interfere and falsely elevate the measured creatinine concentration.
The IDMS method avoids this and typically has a 0.23 mg/dL lower measurement than the picric acid method on an identical sample. This means that comparisons between current values and older cut-points may be skewed.
Now, with these pieces of information in mind, let’s go back to the precision problem to which our (sarcastic) preceptor alluded. Since the original equation reduced 249 pairs of data to seven, the precision of the equation is called into question. Looking at other measures of GFR and comparing, some sources estimate that Cockcroft-Gault’s CrCl is off from measured by as much as +/- 30%!
This means that an estimated CrCl of 100 mL/min might be as low as 70 mL/min…or as high as 130 mL/min! Little different, eh?
Of course, the same rationale applied to a calculated CrCl of 10mL/min means it may be as low as 7 mL/min or as high as 13 mL/min. Not quite as impressive numerically, but honestly, that discrepancy is likely less of a problem since there is very little renal clearance happening with either value.
The 30% discrepancy seems most worrisome in the middle, where a CrCl estimate of 40mL/min might be as low as 28 mL/min or as high as 52 mL/min. Smidge harder to interpret across that range…
If you go back to high school chemistry or physics, you will also note the significant digits issue raises its ugly head in the student’s answer. Rarely are there more than two significant digits going into a CrCl estimate calculation, and yet she came up with a 4 digit answer.
Let’s just mention a few other issues.
What weight do we use in the calculation? Actual? Adjusted? Ideal? Lean? Cockcroft and Gault used actual body weight. The Canadian men of the mid-1970s were appreciably thinner than most North Americans are now. The use of an ideal or adjusted body weight seems reasonable these days, but neither has clear evidence.
Also where does ideal body weight come from? The formula most often seen is now called the Devine equation. Dr. Broca, a famous French surgeon, wrote it in 1871 as the following:
Weight (in kg) should equal Height (in cm) – 100, plus or minus 15% for women or 10% for men.
Of course, Broca wrote it in French.
Then, an unknown person translated this into pounds and inches and reported this rule:
For women, allow 100 lbs for the first 5 feet and 5 lbs for each additional inch. For men, allow 110 lbs for the first 5 feet and 5 lbs for each additional inch.
Finally, in 1974, Devine converted this rule to the following two formulas:
Men: Ideal Body Weight (in kilograms) = 50 + 2.3 kg per inch over 5 feet.
Women: Ideal Body Weight (in kilograms) = 45.5 + 2.3 kg per inch over 5 feet.
Absolutely none of this ideal body weight is based on a study. No science here. (Btw, some people subtract 2.3 kg for every inch under 5 feet; again, no science, move along, nothing to see here.)
So, what about Adjusted Body Weight (AdjBW) or Lean Body Weight (LBW)? These have been validated in specific populations - but for other uses.
AdjBW has been validated for aminoglycoside dosing where AdjBW = Ideal BW + 0.4*(Actual BW - Ideal BW). The thought here is that the added weight a person has above their ideal is not all fat but contains some muscle mass. Arbitrarily 40% was chosen as the factor to represent this, and it was validated for that use.
LBW excludes all fat, even the acceptable fat allowed in the IBW. LBW has been used for dosing very water-soluble drugs that have little, if any, fat solubility (notably, digoxin). Neither of these weights has been validated in CrCl, but all four weights (Actual, Adjusted, Ideal, and Lean) have been reported and are frequently found on online java-scripted web pages.
When talking about women a moment ago, we mentioned muscle mass. How do we handle calculations for patients with significant deviations from normal muscle mass leading to decreased creatinine production? Spinal cord injury patients? Muscular dystrophies? Amputations?
Patients who produce less than 900 mg of creatinine per day do not make a sufficient amount for the Cockcroft-Gault equation to accurately estimate clearance, and urine measurements of CrCl should be used.
While we are talking about muscle mass, many folks routinely round an elderly patient’s SCr up to 0.8 or 1 mg/dL, if it’s a small value. Their thought process is that older patients have less muscle mass and therefore less creatinine production - and we must correct for it when using Cockcroft-Gault. Hopping strictly onto the evidence train, there is no data that this correction is accurate.
The production of creatinine leads to the question about elimination of creatinine. It is a function of GFR, and in stable normal patients, creatinine’s t½ is about 4 hours in young healthy males. But in someone with severe renal dysfunction, it may be greater than 3 days! Generally speaking, today’s serum creatinine level reflects yesterday’s renal function.
Also remembering that calculated CrCl is an estimate of an average in patient with varying renal function may make it easier to realize it is not the actual GFR - rather just a representation.
Okay, so now we have utterly beat up the Cockcroft-Gault Creatinine Clearance estimate of GFR. I mean, we literally took the gloves off against the poor thing, UFC style.
But don’t think we’re being biased. Other estimates are going to have similar problems with assumptions (we’ll take on MDRD below). But before we get to that, let’s chat about a couple other important markers of GFR: urine output and serum urea nitrogen.
Urine output (UO) is not a direct measure of GFR. But clearly, if there is no UO, there is no glomerular filtration. Patients who become oliguric (i.e., <0.3 mL/kg/hour or <500 mL/day of UO) or anuric (<50 mL/day), have reduced GFR - even if the CrCl estimate appears to be good. This goes back to the earlier crucial points that today’s SCr and calculated CrCl reflect yesterday’s renal function and that the estimating equations are for stable (unchanging) renal function.
BUN, by the way, is no longer BUN. It is technically SUN or serum urea nitrogen. But by convention in the United States, we still use the abbreviation of BUN even though the sample used is from serum.
Urea is primarily produced in the liver (>99%) from dietary protein. A typical diet produces about 12 grams of urea per day, and the kidneys excrete roughly 10 grams daily. Without renal failure, BUN is primarily a reflection of urea production and will likely remain normal until at least 50% of renal function is lost. Neither SCr nor BUN have a linear relationship to renal function. Instead, both have a parabolic rise as renal function drops:
Since we mentioned alternative estimations of GFR, let’s talk about the Modification of Diet in Renal Disease (MDRD) estimations. MDRD is actually a group of studies for which the first paper was released in the late 1990s, and the resultant equations were purported to have a better estimation of GFR than CrCl.
While MDRD has turned out to be a useful tool for estimating GFR in regard to staging CKD (we will get to that later), it has some limitations when it comes to drug dosing. Regardless, it was quickly adopted by many clinical laboratories because it did not include weight and they could report it right next to the SCr and BUN on the lab report.
They could have done the same with CrCl, but that would mean trusting someone else to properly weigh the patient and report height (if we are going to use IBW or AdjBW), and most laboratory directors were not willing to do so. So they jumped to MDRD instead to bypass the anthropometrics confusion.
The MDRD group reported a couple of different equations (6, 5, and 4 variables), but all report the units as mL/min/1.73m2 (not mL/min as CrCl is usually reported). That last term (/1.73m2) effectively normalized the results for a “typical”-sized person. Which is fine for staging CKD, but it’s not so good for reporting the patient’s GFR. Of note, most labs report the MDRD result as eGFR (estimated GFR).
There were also several limitations in the MDRD papers:
- The original data with 1628 participants had a mean eGFR ~40mL/min/1.73m2, and no patients had an eGFR above 60mL/min/1.73m2.
- Much like the Cockcroft-Gault study, the population was primarily white men.
- Only 12% of subjects were African-American, and none were of Asian descent.
- A mere 6% had diabetes, which is a very common cause of intrinsic renal disease.
- None of the participants in the study were hospitalized.
- None of the subjects were children, and few were elderly.
As such, results greater than 60 from the MDRD equations should only be reported as >60mL/min/1.73m2. An actual value should not be reported.
Interestingly, the same group of researchers has published a new algorithm called CKD-EPI that did have participants with eGFR > 60. But this is definitely not yet mainstream.
Student: So what am I to do? Everything I know about renal function has just been thrown out the window!
Well, not exactly thrown out the window…but everything you know about everything comes with basic assumptions. These assumptions for renal function estimates have just been pointed out to you.
Also, we have been using these equations and assumptions for over 40 years, and if considerate, you can use them too.
Let’s look at a couple of scenarios.
You are working the community pharmacy counter at the giant HMO-owned and operated mega hospital and clinic. You have access to the patient’s medical record and can see that the last SCr she had (last month) was 1.4 mg/dL, her BUN was 16 mg/dL, and her albumin was 3.3 mg/dL. She is picking up an antibiotic for pneumonia, and you want to see if it is dosed properly.
She is 5 feet tall, 120 pounds, and 65 years of age. You log onto your computer and pump out the results from one of the many web pages, and you get something like this:
After getting over the fact that there are decimals in the results, you see that the eCrCl results are all well within 30% of each other. This should give us some confidence that this patient’s actual CrCl (as an estimate of GFR) is near 40 mL/min.
We can also look at ABW, IBW, and % overweight and see that her ABW is only 10 lbs over her IBW and that she is only 9% overweight. We know she is 65 years old and might have decreased muscle mass, but we can figure that she is height and weight proportionate. We can also see her in the waiting area, and she has defined musculature. Therefore we rule out decreased muscle mass.
The MDRD equations and the CKD-EPI all put her normalized eGFR between 50 and 55 mL/min/1.73m2, which would be classified as CKD stage 3 (stay tuned for Kidney Beans: Part 3 for further discussion). This is pretty consistent with decreased renal function associated with age.
This particular java script also “un-normalizes” the eGFR by multiplying her eGFR by her Body Surface Area (about 1.54 m2) and dividing by the normal 1.73m2 (which is how the bottom 2 results end up with units of mL/min). These estimates of her GFR are near 45-50.
Also helping me believe this value is a normal BUN. Additionally, since she is an outpatient, we can reasonably suspect she is not oliguric or anuric. To be sure, we could check the notes in the chart for any mention of either and/or ask her if she is urinating normally.
Taken together, we could be quite confident she has a CrCl in the range of 40-50 mL/min. If she was prescribed an antibiotic with a CrCl cut-off of 30 mL/min for adjustment, we could feel safe that it was the correct dose.
You are working inpatient and have an antibiotic order for a patient with an old C4 fracture who has been a tetraplegic for decades. He is usually wheelchair-bound, and if you look at him, you see no muscle mass anywhere but his neck and face.
He is 50 years old, and the lab data shows a SCr of 0.3 mg/dL, a BUN of 8 mg/dL, and an albumin of 3 mg/dL. He is 110 lbs and 65 inches tall (long, he has to lie down to measure his height).
STOP. Resist the urge to calculate his CrCl. Don’t do it. No estimate is going to get it right.
Instead whip out the computer and find a java-script that will calculate a timed-measured CrCl. Contact your friendly neighborhood admitting physician and ask if he will order a timed-measured CrCl.
The evidence seems clear that if you have a good (that is, a normal amount of urine for the time frame) urine sample, an 8-hour or 12-hour timed-measured CrCl is as good as a 24-hour test. The timed test literally measures volume cleared over time, hence, mL/min. You simply need the urine creatinine concentration, urine volume, the time that it took to collect the sample, and a simultaneous serum creatinine. This will yield a much more accurate measured CrCl:
CrCl = [urine creatinine (mg/dL) x urine volume (mL)] / [serum creatinine (mg/dL) x time (min)]
Of course, this method has its caveats as well. It assumes the following:
liver creatine production is constant,
creatine conversion to creatinine is constant,
no effect from oral intake,
no effect from GI excretion,
creatinine is freely filtered (not secreted or reabsorbed),
assays are accurate,
urine collection is complete, and
renal function is stable.
None of these is completely true at all times! But sometimes we have to work with assumptions.
So, in the patient described above, if the 1200mL of urine is collected over 24 hours (1440 minutes) and the urine creatinine concentration is 40 mg/dL with a serum creatinine of 0.4 mg/dL, you should find a CrCl of about 83 mL/min.
Alright, this is the end of Kidney Beans: Part 1! Hopefully, now you have a better idea of the assumptions and problems with our estimates of CrCl as a marker for GFR. In Part 2, we will talk about acute kidney injury (AKI) and chronic kidney disease (CKD).