# DLCO, de-constructed

My wife watches the Food Network a lot and I occasionally watch it with her but I can only take so much of it before I go off and read or work on one of my projects. I’ve noticed however in the various cooking contests that sometimes a chef will deconstruct a familiar recipe. This more or less means they break the recipe down into its components and present them as separate pieces or perhaps by putting what goes inside on the outside instead.

I’ve discussed the DLCO test with numerous people and have found that many know and understand (or at least remember) the ATS/ERS criteria for test quality. At the same time however, there seems to be very few people that understand the formula used to calculate the single-breath DLCO and I suspect this is probably because most of us didn’t like the mathematics classes we had to attend in high school or college (and tried to forget what we learned as quickly as we could afterwards).

The DLCO formula isn’t that complicated however, and more importantly all the components of the DLCO test and the reasons for the ATS/ERS quality criteria are embedded within it. All this seems to be a good reason to de-construct the DLCO “recipe” and try to explain it’s various pieces.

As a reminder the single-breath DLCO formula is:

Where:

VA = alveolar volume in ml

BHT = breath holding time in seconds

Pb = barometric pressure

PH2O = partial pressure of water vapor in the lung

FITrace = fractional concentration of tracer gas in the inspired DLCO mixture

FATrace = fractional concentration of tracer gas in the alveolar sample

FICO = fractional concentration of CO in the inspired DLCO mixture

FACO = fractional concentration of CO in the alveolar sample

I think the part that bothers everybody the most is:

and that’s because there’s two different things going on here. First, the part within the brackets:

is intended to correct the initial CO concentration for the dilution that occurs when the DLCO test gas mixture is inhaled and mixes with the gas that was within the lung at the start of the inhalation. The whole point of the DLCO test is to measure CO uptake but the initial concentration for this measurement is not what’s in the tank, it’s what’s in the lungs after it has been diluted by the lung’s residual volume and deadspace gas.

# Why the FEV1/FVC ratio LLN as a percent of the predicted FEV1/FVC ratio is important

My medical director and I had a discussion today about where the cutoff for a normal FEV1/FVC ratio would be for a 93 year old patient of his. Part of the problem is that there are almost no reference equations for patients this age and the best you can usually do is to extrapolate. Another part is that anybody in their 90’s is a survivor and must have had good lung function throughout their life to reach that age, which means that they aren’t average so it’s not clear how well extrapolation actually works in this population. The final part is that the guidelines for PFT interpretation that are used by my lab were put into place about 40 years ago and reflect the thoughts at that time. I updated part of the guidelines with the 2005 ATS/ERS interpretation algorithm about 10 years ago, but the thresholds for normalcy (as well as the reference equations we use) still haven’t changed all that much. I’ve brought this issue up a number of times over the years (usually every time I get a new medical director) but haven’t gotten a consensus from the pulmonary physicians on either the need for change or for what threshold values should be used.

Anyway, both my medical director and I felt felt that the LLN for the FEV1/FVC ratio (when viewed as a percent of the predicted FEV1/FVC ratio) is probably lower for a 75 year old (and certainly for a 93 year old) than it is for a 25 year old, and that the current lab guidelines for interpretation were probably diagnosing airway obstruction in the elderly more often than they should. My lab currently uses the NHANESIII reference equations for spirometry however, and I wasn’t sure they showed this particularly well since the equations for the FEV1/FVC ratio and its LLN are quite simplistic compared to those for FVC and FEV1.

The NHANESIII reference equations were published in 1999 and at that time they were derived from the largest population that had ever been studied (7428 subjects, 40.9% male, 59.1% female) and with the most sophisticated statistical analysis that had been used up until that time. In 2012 however, the Global Lung Function Initiative (GLI) released a set of reference equations using data obtained from 73 centers world-wide on 97,759 subjects (44.7% male, 55.3% female). Statistical analysis of the GLI data was performed using the Lambda, Mu, Sigma (LMS) approach and a set of equations were derived that covered ages 3 to 95.

I have some reservations about how well the GLI equations match the population served by my lab but it’s a moot point whether I like them or not since even now, 5 years after the GLI equations were published, my lab’s software has not been updated to include them. The reason for this is that the GLI spirometry equations use what are called “splines” to generate the spirometry reference values and these are taken from a look-up table. My lab’s software does have an equation editor but it will not accommodate lookup tables so the GLI equations can’t be added. I’m sure our equipment manufacturer could get around this if they really wanted to, but so far it hasn’t happened.

I do have a lot of respect for the GLI equations however, and think that the overall view they give of the normal distribution of FVC, FEV1 and the FEV1/FVC ratio is far more correct than those of any prior studies. Using a spreadsheet tool downloaded from the GLI that lets me generate the GLI spirometry predicted values and the NHANESIII reference equations I decided to take a closer look at their predicted FEV1/FVC ratios and their LLNs.

# What’s wrong with an elevated DLCO?

Well, not necessarily anything, although as usual that depends on the circumstances. Recently I was contacted by an individual who was concerned that their DLCO had decreased from 120% of predicted to 99% of predicted. They also mentioned that their DLCO results have normally ranged from 117% to 140% of predicted over the last 9 months.

More interestingly however, they said that

“the technician told me before I even took the test that anything over 100% for DLCO is essentially a testing error.”

Wow. That statement is wrong on so many levels it’s hard to know where to start but I’ll give it a shot anyway.

First, there are a variety of DLCO reference equations. The ATS/ERS guidelines recommends that PFT Labs pick the reference values that most closely matches their patient population but how this is done is left to individual labs. There are at least a couple dozen DLCO reference equations to choose from and probably about a half dozen of these are in common use in PFT labs around the world.

Because no patient population is ever going to precisely match those of a study this means that DLCO results are going to tend to be above or below 100% of predicted depending on which reference equation the lab is actually using. This also means that if results from otherwise normal subjects are mostly above or mostly below 100% of predicted then the wrong reference equations are being used.

# Z Score to remember is -1.645

The use of Z scores to report PFT results, both clinically and for research is occurring more and more frequently. Both the Z score and the Lower Limit of Normal (LLN) come from the same roots and in that sense can be said to be saying much the same thing. The difference between the two however, is in the emphasis each places on how results are analyzed. The LLN primarily emphasizes only whether a result is normal or abnormal. The Z score is instead a description of how far a result is from the mean value and therefore emphasizes the probability that a result is normal or abnormal.

Reference equations are developed from population studies and the measurements that come from these studies almost always fall into what’s called a normal distribution (also known as a bell-shaped curve).

A normal distribution has two important properties: the mean value and the standard deviation. The mean value is essentially the average of the results while the standard deviation describes whether the distribution of results around the mean is narrow or broad.

The simple definition of the Z score for a particular result is that it is the number of standard deviations that a result is away from the mean. It is calculated as:

# Assessing post-BD improvement in FEV1 and FVC as a percent of the predicted

The 2005 ATS/ERS standards for assessing post-bronchodilator changes in FVC and FEV1 have been criticized numerous times. A recent article in the May issue of Chest (Quanjer et al) has taken it to task on two specific points:

• the change in FVC and FEV1 has to be at least 200 ml
• the change is assessed based on the percent change (≥12%) from the baseline value

The article points out that the 200 ml minimum change requires a proportionally larger change for a positive bronchodilator response in the short and the elderly. Additionally, by basing the post-BD change on the baseline value it lowers the threshold (in terms of an absolute change) for a positive bronchodilator response as airway obstruction become more severe. As a way of mitigating these problems the article recommends looking at the post-bronchodilator change as a percent of predicted rather than as a percent of baseline.

The article is notable (and its authors are to be commended) because it studied 31,528 pre- and post-spirometry records from both clinical and epidemiological sources from around the world. For the post-bronchodilator FEV1 and FVC:

• the actual change in L
• the percent change from baseline
• the change in percentage of predicted
• the Z-score

were determined.

# Is there such a thing as a normal decrease when the FEV1 isn’t normal?

I’ve mentioned before that my lab’s database goes back to 1990, so we now have 27 years of test results available for trending. At least a couple times a week we have a patient who was last seen 10 or even 20 years ago. When I review their results I try to see if there has been any significant change from their last tests. Since the last tests are often quite some time in the past the changes in an absolute sense are often noticeably large. The question then becomes whether or not these changes are normal.

Although the ATS/ERS, NIOSH and ACOEM standards for spirometry address changes over time they don’t specifically discuss changes over a decade or longer. Instead, without indicating a time period (other than saying a year or more), the concensus is that a change greater than 15% in age-adjusted FVC or FEV1 is likely to be significant. A change in absolute values greater than:

Or if the current FEV1 is less than:

Then the change is likely significant.

This sounds fairly reasonable and although we could quibble about the importance of how quickly or slowly this age-adjusted 15% change occurs and how well it applies when the patient’s latest age is beyond the reference equation’s study population (we have a fair number of 90+ year old patients nowadays) or when it’s across a developmental threshold (adolescent to adult), it’s still a good starting point.

I’ve been more or less following these rules for the last several years, when the results for a patient whose last test was 18 years ago came across my desk. The FEV1 from the current spirometry was 71% of predicted and the FEV1 from 18 years ago was 70% of predicted. Strictly speaking the absolute change was about -15% (the FEV1 was 2.06 L in 1999 and 1.76 L in 2017, a 0.30 L change) but when adjusted for the change in age, that’s only 40% what a significant change would need to be:

Given that the FEV1 percent predicted from both the older and newer test were essentially identical I automatically started to type “The change in FEV1 is normal for the change in age” when it suddenly occurred to me that neither FEV1 was normal in the first place so how could I be sure the change be normal?

# Another post-BD FVC conundrum

Okay, this may be wrong but at the moment I’m can’t seem to find a reason why it should be. A report like this came across my desk a couple of days ago.

 Observed: %Predicted: Post-BD: %Predicted: %Change: FVC: 4.59 94% 4.87 100% +6% FEV1: 3.38 89% 3.58 94% +6% FEV1/FVC: 73.6 95% 73.5 95% 0

Not particularly unusual and it would usually be interpreted as being within normal limits without a significant post-BD change. If you calculate the FEV1/VC ratio using the pre-BD FEV1 and the post-BD FVC however, it’s 89% of predicted and this indicates mild airway obstruction. But you’re not supposed to use the post-BD FVC this way, are you?

Well, why not?

# Is there airway obstruction when the FEV1 is normal?

I’ve been reviewing the literature on PFT interpretation lately and in doing so I ran across one of the issues that’s bothered me for a while. Specifically, my lab has been tasked with following the 2005 ATS/ERS guidelines for interpretation and using this algorithm these results:

 Observed: %Predicted: LLN: Predicted: FVC: 2.83 120% 1.76 2.36 FEV1: 1.77 100% 1.26 1.76 FEV1/FVC: 63 84% 65 75

would be read as mild airway obstruction.

Although it’s seems odd to have to call a normal FEV1 as obstruction I’ve been mostly okay with this since my lab has a number of patients with asthma whose best FVC and FEV1 obtained at some point in the past were 120% of predicted or greater but whose FEV1 frequently declines to 90% or 100% of predicted. In these cases since prior studies showed a normal FEV1/FVC ratio then an interpretation of a mild OVD is probably correct even though the FEV1 itself is well above the LLN, and this is actually the situation for this example.

# The FVC/DLCO ratio. Will the real percent predicted please stand up?

Recently a reader asked me a question about the FVC/DLCO ratio. To be honest I’d never heard of this ratio before which got me intrigued so I spent some time researching it. What I found leaves me concerned that a lack of understanding about reference equations may invalidate several dozen interrelated studies published over the last two decades.

Strictly speaking the FVC/DLCO ratio is the %predicted FVC/%predicted DLCO ratio (which is usually abbreviated FVC%/DLCO%) and it appears to be used exclusively by specialists involved in the treatment of systemic sclerosis and related disorders. Specifically, the ratio is used to determine whether or not a patient has pulmonary hypertension. The basic idea is that (quoting from a letter to the editor):

“As we know, in ILD both FVC and DLCO fall and their fall is proportionate, whereas in pulmonary arterial hypertension DLCO falls significantly and disproportionately to FVC.”

A variety of algorithms using the FVC%/DLCO% have been devised. Inclusion factors are usually an FVC < 70% of predicted and a DLCO (corrected for hemoglobin) < 60% of predicted. A number of different threshold values for FVC%/DLCO% have been published ranging from 1.4 to 2.2 with the differences appearing to be dependent on study population characteristics and the type of statistical analysis performed. It is thought that individuals meeting the inclusion factors with an FVC%/DLCO% ratio above the threshold most probably have pulmonary hypertension.

# When no change is a change, or is it?

I was reviewing a spirometry report last week and when I went to compare the results with the patient’s last visit I noticed that the FVC and FEV1 hadn’t changed significantly. However, the previous results were from 2009 and when the percent predicted is considered there may have been a significant improvement.

 2009 Observed: %Predicted: FVC: 2.58 87% FEV1: 1.60 72% FEV1/FVC: 62 82%
 2016 Observed: %Predicted: FVC: 2.82 104% FEV1: 1.65 82% FEV1/FVC: 59 79%

The answer to whether or not there was an improvement would appear to depend on what changes you’d normally expect to see in the FVC and FEV1 over a time span of 7 years. The FVC and FEV1 normally peaks around age 20 to 25 and then declines thereafter.