How should predicted TLC and RV be derived?

The ATS-ERS standards on lung volume measurements says that measured TLC and RV can be calculated either by

RV = FRC – ERV then TLC = RV + SVC

or by

TLC = FRC + IC then RV = TLC – SVC

with the preference going to the first method. Strictly speaking, given the same FRC and SVC measurements either method is going to end up with exactly the same calculated TLC and RV values. Conceptually speaking I believe that TLC = FRC + IC is a more relevant way to think about TLC but this is only because I think that patients find it easier to perform a quality IC maneuver than a quality ERV maneuver.

A while back I found out that the predicted TLC in my lab’s test systems was being derived from the predicted RV from one set of equations and the predicted FVC from another set of equations (i.e. predicted TLC = predicted RV + predicted FVC). This is likely done so that there will be no discrepancy between the predicted FVC and predicted SVC on reports. I am not sure this is the correct decision since SVC does tend to be slightly larger that FVC but the difference is admittedly small (<1%) in healthy subjects so it is not likely to be significant.

Does it matter, however, for predicted TLC and RV which value’s reference equation you start with and which FVC reference equation you use with them? 

There are, of course, many different reference equations for lung volumes and spirometry, but to keep this simple I will choose the ones that I think are the most common and most relevant. For a 50 year old, 175 cm Caucasian male therefore, the predicted lung volumes look like this:

Equation: TLC FRC RV SVC
Quanjer 6.90 3.42 2.16 4.74
Crapo 6.74 3.60 1.98 4.76

And the predicted FVCs look like this:

Equation: FVC:
NHANESIII 4.88
GLFI 4.80
Morris 1988 4.71
Knudsen 1983 4.50

For a 50 year-old, 165 cm Caucasian female the predicted lung volumes look like this: 

Equation: TLC FRC RV SVC
Quanjer 5.76 2.97 1.97 3.76
Crapo 5.79 3.27 2.03 3.79

And the predicted FVCs look like this: 

Equation: FVC:
NHANESIII 3.66
GLFI 3.57
Morris 1988 3.41
Knudson 1983 3.28

If you start with the predicted RV, then the range of predicted TLCs (and the percent difference from the original predicted TLC) derived using a predicted FVC look like this: 

Male: Original: NHANESIII GLFI Morris Knudson
Quanjer 6.90 7.04 (+2.0%) 6.96 (+0.8%) 6.87 (-0.4%) 6.66 (-3.6%)
Crapo 6.74 6.86 (+1.8%) 6.78 (+0.6%) 6.69 (-0.7% 6.48 (-4.0%)
Female: Original: NHANESIII GLFI Morris Knudson
Quanjer 5.76 5.63 (-2.3%) 5.54 (-3.9%) 5.38 (-7.1%) 5.25 (-8.8%)
Crapo 5.79 5.69 (-1.7%) 5.60 (-3.3%) 5.44 (-6.0%) 5.31 (-8.2%)

If you start with predicted TLC, then the range of predicted RVs (and the percent difference from the original predicted RV) derived using a predicted FVC look like this 

Male: Original: NHANESIII GLFI Morris Knudson
Quanjer 2.16 2.02 (-6.5%) 2.10 (-2.8%) 2.19 (+1.4%) 2.40 (+11.1%)
Crapo 1.98 1.86 (-6.1%) 1.94 (-2.0%) 2.03 (+2.5%) 2.24 (+13.1%)
Female: Original: NHANESIII GLFI Morris Knudson
Quanjer 1.97 2.10 (+6.6%) 2.19 (+11.2%) 2.35 (+19.3%) 2.48 (+25.9%)
Crapo 2.03 2.13 (+4.9%) 2.22 (+19.0%) 2.38 (+17.2%) 2.51 (+23.6%)

So for average males, when comparing the original predicted TLC to the predicted TLCs derived using a predicted RV and FVC the differences range from +0.14 L (+2.0%) to -0.26 L (-4.0%). For average females, the differences range from -0.10 L (-1.7%) to -0.51 L (-8.8%).

For average males, when comparing the original predicted RV to the predicted RVs derived using a predicted TLC and FVC the differences range from -0.08 L (-6.5%) to +0.26 L (+13.1%). For average females, the differences range from +0.10 L (+4.9%) to +0.51 L (+25.9%).

Note: For these examples I used average heights from the more common reference equations. I have noted before that when a subject is at or outside the normal range of heights (63” to 76” for Caucasian males, 57” to 70” for Caucasian females) the differences between reference equations become more pronounced so the results presented are probably at the lower end of what is routinely possible.

In both absolute and a percent difference values, there was less difference in the derived TLC and RV predicted values for males than for females. It does depend however, on which reference equations are paired together and which direction the calculations are going in. The NHANESIII and GLFI reference equations come from much larger populations and underwent a far more significant statistical analysis than the Morris or Knudson reference equation and for these reasons I think that they should the ones used by all PFT labs. They are also the equations that showed the least difference between the original predicted TLC and RV, and the derived TLCs and RVs. There are, of course, valid reasons having to do with continuity and ethnicity why a PFT Lab would choose a different reference equations than these.

Does deriving TLC and RV from separate equations matter? And does it matter which direction the calculations move towards?

For the majority of patients the source of the predicted TLC will likely not matter all that much. It will matter, of course, for patients that are at the lower limit of normal because there even a 1% difference in predicted TLC can change how the results are interpreted. From my own experience I’d say the number of patients who will be affected will only be a few percent of all those who have their lung volume measurements (although that will depend on the selected reference equations) and that the “error” in using a derived TLC is probably smaller than the potential errors within the lung volume test itself.  

Interestingly, although the magnitude of the differences between the original and derived TLCs and the original and derived RVs were similar in liters, they were significantly different when looked at as a percent. For this reason I suppose it actually does make sense to start with the predicted RV and to derive TLC rather than the other way around. Having said that, when lung volumes are measured TLC is far and away the most important and critical result. For this reason alone I think that the measured TLC should be compared to a “real” predicted TLC and not one derived from different reference equations.

I think we all need to be more aware where our predicted values come from. It took some significant digging into my lab’s software to find out how the predicted TLC was being derived. I don’t think this was due to any particular subterfuge on the part of our equipment vendor, just that there are so many reference equations and so many ways to use them that it was never explicitly documented.

Reference equations have always seemed to be an arcane part of our field. Their uses and limitations are often marginally understood and for this reason I think that any inconsistencies in their use are often overlooked. I will be the first to admit that my understanding of statistics is only basic at best. Even so, I believe that any individual responsible for interpreting PFT results should take the time to know which reference equations are being used and how the predicted values are being derived simply because of the implications this has on the interpretation. 

References:

Crapo RO, Morris AH, Clayton PD, Nixon CR. Lung volume in healthy nonsmoking adults. Bull Eur Physiopathol Respir 1983; 18:

Hankinson, JL, Odencrantz JR, Fedan KB. Spiromtric reference values from a sample of the general U.S. Population. Amer J Resp Crit Care, 1999; 159: 179-187.

Knudson RJ, Lebowitz MD, Holberg CJ, Burrows B. Am Rev Resp Dis 1983; 127: 725-734

Morris JF, Koski A, Temple WP, Claremont A, Thomas DR. Fifteen-year interval spirometric evaluation of the Oregon Predictive

Quanjer PH, Stanojevic S, Cole TJ, Baur X, Hall GL, Enright PL. Hankinson JL, Ip MS, Zheng J, Stocks J. Multi-ethnic reference values for spirometry for the 3-95 year age range: the global lung function 2012 equations. Eur Respir J 2012; 40: 1324–1343.

Stocks J, Quanjer PH. Reference values for residual volume, function residual capacity and total lung capacity. Eur Respir J 1995; 8: 492-506.

Creative Commons License
PFT Blog by Richard Johnston is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.