As a continuation of this post, I am taking a closer look at two feeding studies. Both studies report data on dietary potassium intake and urine potassium measures.
The articles today are:
Why these studies are helpful?
Both of these studies help us understand the relationship between potassium intake and urine potassium measures by:
- They measured the amount of potassium in the diets in the lab. In both of these studies, we know EXACTLY how much potassium people were supposed to be eating.
- These studies were done in adults with CKD. This is important. In my last post, I highlighted that there are several reasons to question the accuracy of urine potassium measures in CKD.
A closer look at Study 1
In the Lobene et al study, dietary potassium intake was compared with 24 hour urine and spot urine potassium. The Tanaka equation was used on the spot urine. The Tanaka equation attempts to estimate a 24 hour urine result from a spot urine sample.
In this study, the authors reported that 24 hour urine was poorly correlated with dietary potassium intake (r=0.37). Spot urine using the Tanaka equation was poorly correlated with both dietary intake (r=0.55) and 24 hour urine collection (r=0.32)
Quick tip for interpreting r-values:
Great correlation = 0.9-1.0
Good correlation = 0.7-0.9
Moderate correlation = 0.5-0.7
Low correlation = 0.3-0.5
No correlation = 0.3-0
If the value is negative, it means as one goes up, the other goes down. If the value is positive, it means the both go in the same direction.
In addition to correlation, the authors also use a Bland-Altman Plot to explore how potassium intake compared with 24 hour urine. The authors explicitly state that in the CKD population, 24 hour urine overestimated actual intake. In the healthy controls, 24 hour urine potassium was lower than actual intake.
How to interpret this Bland-Altman plot
Ok to interpret this graph, let’s ask ourselves the following questions (courtesy of this website):
How big is the average discrepancy between methods (the bias)? You must interpret this clinically. Is the discrepancy large enough to be important? This is a clinical question, not a statistical one.
- The mean difference is 54mg. Clinically this is insignificant. I don’t care if someone is eating 2054mg or 2000mg of potassium.
How wide are the limits of agreement? If it is wide (as defined clinically), the results are ambiguous. If the limits are narrow (and the bias is tiny), then the two methods are essentially equivalent.
- The limits of agree (marked as Mean-2SD) are clinically significant at 1051 and 1159mg. For example, if the 24 hour urine said someone was eating 3051mg when they were actually only eating 2000mg, I would consider that really wrong. On the flip side, if the urine said someone was only eating 841mg when in fact they were eating 2000mg, I would also consider that really wrong.
Is there a trend? Does the difference between methods tend to get larger (or smaller) as the average increases?
- Lucky us! In this case the authors help us answer this question. The authors specifically state that there is a tendency to over-estimate potassium intake at higher levels of intake. Here’s what you see in the graph – notice that when values on the X-axis (named: Average of intake and 24 hour UK) are under about 2700mg, all the dots are above the mean line. When the values are over 2700mg (with the exception of 1 dot), all the dots are below the mean line.
- Why does this matter? It tells us that the error in the tool may not be the same based on the intake. Ok, lets use an example. Lets say the tool is known to underestimate potassium by 500mg for everyone then I could say “Ok, I am going to add 500mg to each value”. BUT, in this case the error isn’t the same. For those eating more the error OVERESTIMATES (example – if you are eating 4000mg the tool might say you are eating 5000mg). While for those eating less, the error underestimates (example if you are eating 1500mg, the tool says you are eating 500mg).
Is the variability consistent across the graph? Does the scatter around the bias line get larger as the average gets higher?
- In this case, only 3 values are outside of the first mark (the red box). And again the authors make interpretation easy by explicitly stating that there is a similar amount of bias at either end of intake. If the variability of bias was higher, there may have been more dots outside the red box. And there may have been more dots above the box or below the box.
Author Conclusions
The authors conclude: The estimated 24 hour urine (using the Tanaka equation) was a poor indicator of true potassium intake in adults living with CKD. Though provide that caveat that this requires a larger sample to confirm it. In the discussion, the authors state that there was poor agreement between measured 24 hour urine and intake for potassium.
Take Aways for Study 1
This study suggests that when we know how much potassium someone is actually eating and compare it to the a 24 hour urine potassium or a spot urine that the results won’t match. Even worse that depending on how much someone is eating the results will provide a variety of wrong answers – the error isn’t the same at all levels of potassium intake.
Or said another way, here is what it means for me:
- If I see a paper that has used the Tanaka equation to estimate potassium intake in CKD – I will not trust the results of that paper
- 24 hour urine potassium is also poorly correlated with potassium intake in CKD and I will be cautious when interpreting the results of these studies
Wow, that post got really long in a hurry! I am going to go into Study 2 in the next post!
One thought on “A deeper dive into urine K in CKD”