Blog

Late in 2012 the Global Lung Function Initiative (GLFI) released their reference equations for spirometry. Although not without some criticism this was an important step towards the development of a single spirometry reference equation that can be used by PFT Labs worldwide. The GLFI recently announced that they were accepting data for their TLCO Task Force. This is project similar to what was done for spirometry, and is intended to create a set of reference equations for TLCO (DLCO) that is applicable to all ages and ethnicities.

The GLFI Task Force is actively seeking test results on representative and “healthy” populations from labs anywhere in the world and in particular results from young and elderly individuals. They are looking for data sets from a minimum of 100 (or 150, there is some confusion about the minimum number in the Task Force’s own documentation) individuals. PFT Labs that wish to participate will need to provide information about each subject’s age, height, weight, gender, ethnicity and health status. Patient results will be de-identified before submission to protect patient privacy. Labs will also need to provide information about the test systems (brand and model number) and software (version number) used to perform the tests.

Please visit the GLFI TLCO Task Force web page if you have any interest in this project.

I applaud this project and look forward to seeing their results. The current DLCO reference equations have severe limitations both in the number of subjects and in the range of ethnicities included in the studies and the GLFI TLCO project should go a long way towards clearing up the many of the known inconsistencies.

The lungs are the gas exchange organ of the body. The mechanical aspects of the lung, which do of course have a bearing on gas exchange, can be assessed by spirometry and lung volume tests but for a complete assessment of an individual’s lung function a diffusing capacity test (DLCO) must be performed as well. The actual gas exchange rate at any moment can be highly variable and depends on a number of factors such as cardiac output, pulmonary capillary blood volume and ventilation-perfusion matching that are difficult to measure. For this reason the diffusing capacity test, more so perhaps than any other pulmonary function test, must be performed in a highly standardized way in order to produce results that can be meaningfully trended over time and meaningfully compared to other individuals.

Accurate diffusing capacity results therefore depend on attention to details such as inspiratory time, inspiratory capacity, breath-holding time, washout volume and alveolar sample size. Even when these values are essentially identical results can still vary dramatically from test to test. For this reason my PFT lab’s policy is to perform a minimum of two DLCO tests and if the results aren’t reproducible then a third test and possibly a fourth test. Whenever possible the closest (not the highest) results from two good quality tests are averaged and reported. We think that this approach gives is the best way to get accurate and reproducible DLCO test results from our patients.

This may give my lab DLCO results that are adequate for trending but how do we know when they are normal? Like everybody else we have to rely on a reference equation generated from a study of presumably normal individuals. Selecting the proper reference equation continues to be an ongoing problem. I have been able to find fourteen different DLCO reference equations that appear to be in more-or-less common use.  Even after comparing them however, I am not sure the selection process is any clearer or easier.

The ATS and ERS do not specifically recommend any particular reference equation. They have recommended using reference equations that have a lower limit of normal (LLN) but at this time I do not believe that any study meets this requirement. There are DLCO studies that publish a LLN but it is a single value that is applied to all ages and heights and for this reason I do not think that this is particularly useful.

I believes the ATS-ERS recognizes the limitations of LLN in reference equations because the ATS-ERS statement on PFT interpretation advocates the use of a DLCO percent predicted of 80% as the normal cutoff. It also indicates that 79% to 60% of predicted is a mild reduction, 59% to 40% is a moderate reduction and that values less than 40 of predicted are severely reduced.

The majority of reference equations I was able to find are for Caucasian populations (9), with the next largest group being Chinese (2), followed by Hispanic (1), Indian (1) and Brazilian (1). I am concerned that I was not able to find any reference equations for Blacks, which is a glaring omission. I was able to find two studies that compare results from a Black study population to Caucasian values but neither study generated reference equations for Blacks. The Chinese and Indian reference equations produce the lowest values which is consistent with the spirometry and lung volume reference values for the same populations. The the reference values for Hispanic and Brazilian populations were not significantly different from those for Caucasians.

Other than ethnicity, there are differences in methodology that can affect a reference equation. One of these is how breath-holding time (BHT) is measured. Calculated DLCO is inversely related to the measured BHT. This means that for the same DLCO test effort, the shortest measured BHT produces the highest calculared DLCO. The current ATS-ERS recommendation is to use the Jones-Meade method because it is least affected by airway obstruction. The Ogilvie method produces similar results to Jones-Meade in normal patients, but tends to underestimate BHT when airway obstruction is present. The ESP method always produces the smallest BHT which leads to the largest calculated DLCO results.

Another reason to select one reference equation over another is that the FIO2 used to measure DLCO will also affect results. Because carbon monoxide competes with oxygen for uptake by hemoglobin, lower FIO2’s lead to higher measured DLCO’s and vice versa. European Pulmonary Function labs tend to use an FIO2 that matches that of expired alveolar air (~18%) whereas American labs use an FIO2 that matches room air (~21%).

There are two reference equations that seem to be in error. I suspect there is a transcription error in the female Indian reference equation [M] since the values it produces do not match those discussed in the text of the article they came from. One of the Chinese female reference equations [E] may be accurate for the average height range of the study’s population, but when at higher heights, even those at the upper limit of the published range (174 cm) the results show a negative slope for age (i.e. results get higher with increasing age). Although this may be due to a transcription error it may instead be due to the unusual form of the equation since it includes a weighting factor for [height x age] as well as for height and age.

The differences between reference equation values can be dramatic and this has implications about whether an individual’s measured DLCO will be considered normal or not. For Caucasian males of average height and weight (175 cm, 80 kg) the range of predicted normal values at age 50 is 29.85 to 36.09 (80% cutoff 23.88 to 28.87) which is 19% of the average value. For Caucasian females of average height and weight (160 cm, 60 kg) the range of predicted normal values at age 50 is 17.44 to 27.47 (80% cutoff 13.95 to 21.98) which is 42% of the average value.

Height and age are the primary factors for all DLCO reference equations. The apparent decline in DLCO with age (which is based solely on the equations and not on any longitudinal studies and is independent of height) ranges from 0.117 to 0.246 ml/min/mmHg per year for Caucasian males and 0.068 to 0.179 ml/min/mmHg per year for Caucasian females.

The effect of height is a bit more difficult to assess because it depends on what age you are comparing them at. Using the same height limits I’ve used in the past (160 to 194 cm for males and 140 to 180 cm for females) the differences graph like this:

So how do you decide which reference equations to use? I wish I could be definitive about this issue but the fact is that the selection of reference equations for DLCO remains a subjective process. The first place to start should be (with limitations) the ethnicity of the population your lab services. The problem with this is first whether there is a reference equation that meets this criteria and second is in defining ethnicity. Even for Caucasians, a supposedly well-defined group, there is a very broad range of predicted normal values and I suspect that with sufficient additional studies this would also prove true for other, currently underrepresented, ethnicities.

Note: This is one reason why some researchers (not reported here) have advocated the use of the VA measured during the DLCO test as a factor in reference equations. VA, however, is usually underestimated (less than measured TLC) when airway obstruction is present, and I think this fact alone places severe limitations on the validity of VA as a factor in reference equations.

Factors like the choice of BHT methodology and the FIO2 used should also be a consideration. Studies that don’t don’t match a lab’s procedure or equipment requirements should probably not be selected.

Next, simply because a given reference equation produces results that are near the average value for all equations does not make it more accurate if for no other reason than the average value is affected by the equations chosen to calculate it. Having said that, there are worse reasons to select a reference equation and I think that you need some particularly good justification to use reference equations that are at the top or the bottom the range for age or height. When I look at graphed results, for Caucasian males equation D seems to be nearest the median for all of the heights and age slopes. For Caucasian females, equation K seems to be closest to the median. That’s a subjective assessment however, and you may well see something different than I do.

When we went through our last hardware and software upgrade I’m afraid that we took the easy course and continued to use the same reference equations the lab has used for the prior 30 years. During the initial software setup we thought we had selected this equation but in fact it defaulted to a different one. We started getting complaints from physicians almost immediately because patients appeared to have large changes in their DLCO, which was true when looking at percent predicted but not true when looking at the actual values. For this reason we switched back to the old reference equation. In retrospect I think this was a mistake because the reference equation we use is somewhat of an outlier, particularly for females but at the time we had limited resources to devote to this issue and there were other problems that looked much bigger.

I would recommend exploring these equations and for this purpose I’d like to share a simple spreadsheet I developed to calculate predicted DLCO based on age, height and weight.  

Updated to Adult Reference Equation Explorer:

Adult Reference Equation Explorer – MS Excel Version

Adult Reference Equation Explorer – LibreOffice Version

The Global Lung Function Initiative (GLFI) has indicated that they are starting to develop their own reference equations for DLCO. I welcome this because they have support from researchers around the world and a very sophisticated approach towards statistics. Almost all of the current DLCO reference equations are simple linear regressions. Linearity was shown to be incorrect in spirometry first by NHANESIII and then more definitively by the GLFI. I suspect that DLCO is also less linear than the equations would lead us to suspect particularly for patients that are towards extremes of height and age.

Finally, I am concerned that although the current equipment for DLCO testing meets the letter of the ATS-ERS standards, there are a number of ATS-ERS suggestions for improved testing (PAO2, CO back-pressure) that are not implemented by any manufacturer I am aware of. In addition we have found a number of subtle hardware and software idiosyncrasies in our DLCO testing systems that can have significant effects on results. Although many researchers acknowledge issues like these the fact is that they (and we) are dependent on commercially available testing systems. For these reasons I think it is time the ATS-ERS updated their recommendations on DLCO testing and when they do, they should set the bar higher for equipment manufacturers.

The DLCO test is a critical component of Pulmonary Function testing. Performing it correctly requires close adherence to standards. Even when accurate results are obtained however, they need to be compared to normal values. The choice of which reference equations are used to determine what’s normal has significant implications for the clinical assessment of patients. Unfortunately the selection process remains a conundrum and we need to remember that what’s normal for DLCO is not necessarily clear and that our choices are lines in the sand.

Male DLCO Reference Equation Study Parameters

Equation	Ethnicity	No.	Ages	Heights	BHT	FIO2		Hgb Corrected
[A]	Caucasian	123	18-91	157-194	Ogilvie	25%		Yes
[B]	Caucasian	98 M+F	n/a	n/a	Ogilvie	n/a		n/a
[C]	Caucasian	119	15-70	179.2 +/- 6.6	Jones-Meade	20.9-21.9%		Yes
[D]	Caucasian	300	20-79	n/a	Ogilvie	n/a		No
[E]	Chinese	259	18-80	148-182	Jones-Meade	21.1%		Yes
[F]	Hispanic	71	25+	n/a	ESP	21%		Yes
[G]	Caucasian	110	25-75	n/a	n/a	n/a		Yes
[H]	Caucasian	74	n/a	n/a	Ogilvie	n/a		Yes
[I]	Brazilian	50	20-80	n/a	n/a	21%		No
[J]	Caucasian	243	19-64	171 +/- 6.7	ESP	20%		No
[K]	Caucasian	83	18-86	161-196	n/a	21%		No
[L]	Caucasian	194	20-70	169.8 +/- 6.7	Jones-Meade	18%		No
[M]	Indian	130	15-40	166 +/- 7.6	n/a	n/a		Yes
[N]	Chinese	180	20-79	153-181	Ogilvie	21%		Yes

Male DLCO Reference Equations (height = cm, weight = kg)

Equation	Formula
[A]	DLCO = (0.1646 x Height) – (0.219 x age) – 26.34
[B]	DLCO = (0.0984 x height) – (0.177 x age) + 19.93
[C]	DLCO = ((0.1217 x height) – (0.057 x age) – 8.05) x 2.986
[D]	DLCO = (0.284 x height) – (0.246 x age) – 4.625
[E]	DLCO = (0.4192 x height) – (0.1963 x age) – 33.912
[F]	DLCO = (0.3551 x height) – (0.2741 x age) – 11.3527
[G]	DLCO = ((0.0958 x height) – (0.06 x age) – 3.77) x 2.986
[H]	DLCO = (1.062 x height) – (0.229 x age) + 12.9113
[I]	DLCO = (0.32 x height) – (0.13 x age) – 13.07
[J]	DLCO = (0.441 x height) – (0.1938 x age) – 31.3822
[K]	DLCO = ((0.14005 x height) – (0.074 x age) -10.803) x 2.986
[L]	DLCO = (0.3874 x height) – (0.1961 x age) – 21.8982
[M]	DLCO = ((0.126 x height) – (0.0004 x age^2) – 11.236) x 2.986
[N]	DLCO = (0.3028 x height) – (0.2323 x age) + (0.1132 x weight) – 21.8743

Female DLCO Reference Equation Study Parameters

Equation	Ethnicity		No.	Ages	Heights	BHT	FIO2	HGB Corrected
[A]	Caucasian		122	17-84	146-178	Ogilvie	25%	Yes
[B]	Caucasian		98 M+F	n/a	n/a	Ogilvie	n/a	n/a
[C]	Caucasian		185	15-70	164.4 +/- 6	Jones-Meade	20.9-21.9%	Yes
[D]	Caucasian		327	20-79	n/a	Ogilvie	n/a	No
[E]	Chinese		309	18-80	133-174	Jones-Meade	21.1%	Yes
[F]	Hispanic		99	25+	n/a	ESP	21%	Yes
[G]	Caucasian		102	25-74	n/a	n/a	n/a	Yes
[H]	Caucasian		130	n/a	n/a	Ogilvie	n/a	Yes
[I]	Brazilian		50	20-80	n/a	n/a	21%	No
[J]	Caucasian		469	19-64	157.5 +/- 5.7	ESP	20%	No
[K]	Caucasian		96	18-86	146-177	n/a	21%	No
[L]	Caucasian		167	20-70	158 +/- 6.4	Jones-Meade	18%	No
[M]	Indian		117	15-40	152 +/- 5.3	n/a	n/a	Yes
[N]	Chinese		126	20-79	147-172	Ogilvie	21%	Yes

Female DLCO Reference Equations (height = cm, weight = kg)

Equation	Formula
[A]	DLCO = (0.256 x Height) – (0.144 x age) – 8.36
[B]	DLCO = (0.1118 x height) – (0.177 x age) + 7.72
[C]	DLCO = ((0.0911 x height) – (0.043 x age) – 4.93) x 2.986
[D]	DLCO = (0.238 x height) – (0.153 x age) – 7.781
[E]	DLCO = (-0.0604 x height) – (0.9363 x age) + (height x age x 0.0055) + 33.061
[F]	DLCO = (0.1872 x height) – (0.146 x age) + 3.8821
[G]	DLCO = ((0.0958 x height) – (0.06 x age) – 4.91) x 2.986
[H]	DLCO = (0.16 x height) – (0.1111 x age) + 2.2382
[I]	DLCO = (0.18 x height) – (0.075 x age) + 0.20
[J]	DLCO = (0.1569 x height) – (0.0677 x age) + 5.0767
[K]	DLCO = ((0.07391 x height) – (0.037 x age) – 2.516) x 2.986
[L]	DLCO = (0.1369 x height) – (0.1233 x age) + (0.0917 x weight) + 1.8879
[M]	DLCO = ((0.032 x height) – (0.00068 x age^2) + 2.076) x 2.986
[N]	DLCO = (0.1398 x height) – (0.1691 x age) + (0.1124 x weight) – 1.1088

References:

Brusasco V, Crapo R, Viegi G. ATS/ERS task force: Standadisation of lung function testing. Standisation of the single-breath determination of carbon monoxide uptake in the lung. Eur Respir J 2005; 26: 720-735.

Brusasco V, Crapo R, Viegi G. ATS/ERS task force: Standadisation of lung function testing. Interpretive strategies for lung function tests. Eur Respir J 2005; 26: 948-968.

[A] Crapo RO, Morris AH. Standardized single-breath normal values for carbon monoxide diffusing capacity. Am Rev Resp Dis 1981; 123: 185-189.

[B] Gaensler EA, Smith AA. Attachment for automated single-breath diffusing capacity measurement. Chest 1973; 63: 136-145.

[C] Gulsvik A, Bakke P, Humerfelt S, Omenaas E, Tostenson T, Weiss ST, Speizer FE. Single breath transfer factor for carbon monoxide in an asymptomatic population of never smokers. Thorax 1992; 47: 167-173.

[D] Gutierrez C, Ghezzo RH, Abboud RT, Cosio MG, Dill JR, Martin RR, McCarthy DS, Moorse JLC, Zamel N. Reference values of pulmonary function tests for Canadian Caucasians. Can Respir J 2004; 11(6): 414-424.

[E] IP MSM, Lai AYK, Ko FWS, Lau ACW, Ling SO, Chan JWM, Chan-Yeung MMW. Reference values of diffusing capacity on non-smoking Chinese in Hong Kong. Respirilogy 2007; 12: 599-606

[F] Knudsen RJ, Kaltenborn WT, Knudsen DE, Burrows B. The single-breath carbon monoxide diffusing capacity. Am Rev Resp Dis 1987; 135: 805-811.

[G] Marsh S, Aldington S, Williams M, Weatherall M, Shirtcliffe P, McNaughton A, Pritchard A, Beaseley R. Complete reference ranges for pulmonary function tests from a single New Zealand population. New Zealand Med J 2006; 119: N1244.

[H] Miller A, Thornton JC, Warshaw R, Anderson H, Teirstein AS, Selikoff IJ. Single breath diffusing capacity in a representative of Michigan, a large industrial state. Am Rev Resp Dis 1983; 127: 270-277.

Neas LM, Schwartz J. The determinants of pulmonary diffusing capacity in a national sample of U.S. adults. Amer J Respir Crit Care Med 1996; 153: 656-664.

[I] Neder JA, Andreoni S, Peres C, Nery IE. Reference values for lung finction. III. Carbon monoxide diffusing capacity (transfer factor). Braz J Med Biol Res 1999; 32: 729-737.

[J] Paoletti P, et al. Reference equations for the single-breath diffusing capacity. Am Rev Resp Dis 1985; 132: 806-813.

Pesola GR, Sunmonu Y, Hugguns G, Ford JG, Measured diffusion capacity versus predicted equations estimates in blacks without lung disease. Respiration 2004; 71: 484-492.

[K] Roberts CM, MacRae KD, Winning AJ, Adams L, Seed WA. Reference values and prediction equations for normal lung function in a non-smoking white urban population. Thorax 1991; 46: 643-650.

[L] Roca J, Rodrigue-Roisin R, Cobo E, Burgos F, Perez J, Clausen JL. Single-breath carbon monoxide diffusing capacity prediction equations from a Mediterranean population. Am Rev Resp Dis 1990; 141: 1025-1032

[M] Vijayan VK, Kuppurao KV, Venkatesan P, Sankaran K, Prabhakar. Pulmonary function in healthy young adult Indians in Madras. Thorax 1990; 45: 611-615.

[N] Yang SC, Yang SP, Lin PJ. Prediction equations for single-breath carbon monoxide diffusing capacity from a Chinese population. Am Rev Resp Dis 1993; 147: 599-606.

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

One of the recognized abnormalities of a flow-volume loop is a sawtooth profile due to flow oscillations that are superimposed on either the maximal expiratory or inspiratory flow curve, or the tidal loop.

The sawtooth pattern occurs in only a small fraction of patients but it is quite noticeable when you see it. Estimates of the number of individuals with flow oscillation range from 1.4% to 13.4% with the higher estimates being observed primarily with inspiratory loops. Since many spirometry efforts are concentrated solely on exhalation this means that it may frequently go unrecognized. Recently, I had several reports with distinct sawtooth flow-volume loops come across my desk within a short time period and for this reason thought it might be interesting to re-visit this old subject. I call it old only because most of the research on sawtooth profiles was done in the 1970’s and 1980’s and not much has been published since then.

The sawtooth pattern was first noted in the late 1960’s when flow-volume loops came into common clinical use. To some extent these flow oscillations were initially dismissed because they were thought to be noise from either the spirometry equipment, the amplifiers or the X-Y recorder. For this reason many of the published flow-volume loop waveforms of that time were smoothed to remove these oscillations.

The first investigators to recognize flow oscillations as a real phenomena thought that they correlated primarily with Obstructive Sleep Apnea (OSA) and could be used as an aid to diagnosis. Research since then has shown that most patients with OSA do not have a sawtooth pattern on their flow-volume loop and that what a sawtooth pattern usually correlates with instead is with upper-airway abnormalities from a variety of conditions. Investigators have shown that as well as OSA sawtooth flow profiles can occur in:

• Snorers without OSA
• Airway tumors
• Use of inhaled steroids
• Upper airway stenosis
• Parkinson’s Disease
• Neuromuscular disorders with bulbar involvement
• Leewanhoek’s Disease
• Burn Injury of the upper airway

The upper airway has a number of structures that are capable of affecting airway resistance.

There are two general mechanisms that can cause flow oscillations: fluttering and tremors. Fluttering is partly due to turbulence and partly due to loose tissue. When the structures in the airway are flaccid due to low muscle tone or extra unsupported tissue, it can narrow the airway and when the airway is narrowed turbulence increases. Turbulence in turn causes the loose tissue to flutter and this shows up as flow oscillations. Fluttering is likely the mechanism for flow oscillations in OSA, tumors, burn injuries and other lesions to the upper airway. Flow oscillations are seen more frequently during inspiration is because during inspiration there is negative pressure in the airway which causes the airway to narrow further and increases the likelihood of turbulence. Flow oscillations due to turbulence tend to have a relatively high frequency (>20 Hz).

In extrapyrimidal disorders (Parkinson’s disease) both rhythmic and irregular tremors of the glottic and supraglottic structures have been observed with fiberoptic endoscopes. Vocal cord tremors have been noted in some Parkinson’s patients as well. In motor neuron disease with bulbar involvement tremors can occur in the laryngeal and pharyngeal muscles. Notably these abnormal tremors often persist during breathholding which means they are not caused by airway turbulence but by phasic innervation/denervation of the muscles. Flow oscillations due to tremors tend to be relatively low frequency (4-8 Hz).

Posture can be a factor in the ability to detect flow oscillations. Although most OSA patients with sawtoothing showed this pattern in both the upright and the supine position some OSA patients that did not show sawtoothing while upright did show it while supine.

Like fine art, everybody knows what flow oscillations are when they see them but coming up with an objective definition has been difficult. Various definitions include:

• “…three or more consecutive peaks and troughs, with an amplitudes of 50 to 500 ml/sec and a maximum width of 10% of FVC occurring … in the middle 80% of the expiratory or inspiratory limb…”
• “…three or more consecutive peaks and troughs of no greater than 300 ml during the middle half of the vital capacity…”
• “…a minimum of three oscillations should be present…”
• “…sawtooth pattern was considered to be present…if at least two out of three readers reported its presence.”
• “…flow oscillations were defined as changes in the direction of flow (ie, consecutive flow decelerations and accelerations) superimposed on the inspiratory or expiratory flow-volume curves…”

I doubt that these different definitions have made a significant difference in the ability of researchers to assess flow oscillations but without a consensus of some kind it will probably be difficult to develop a computer algorithm to detect them, assuming this ever becomes desirable.

Many of the mechanical and electronic issues that affected the fidelity of early flow-volume loops have long since been overcome. Even though flow oscillations have a relatively high frequency (usually between 4 and 40 Hz) the digital acquisition rate of even simple spirometry systems is usually high enough that flow oscillations can be accurately recorded. The current obstacle towards recognizing flow oscillation has more to do with the resolution used to display or print the flow-volume loop. It doesn’t matter how accurately a flow-volume curve was recorded if it is going to be printed as a 1-inch rectangle. I mention this because my lab’s software came with some sample report formats and one included incredibly tiny flow-volume loops where the overall contour of the loop would have been difficult to discern, let alone any flow oscillations.

The sawtooth profile on flow-volume loops is seen in many diseases primarily involving the upper airway (and in rare instances from synchronous tremors in the respiratory musculature). Although it has a clear association with upper airway obstruction it is not a specific sign of any disease and it’s presence or absence does not tend to indicate a greater or lesser degree of disease severity. For these reasons flow oscillation has limited clinical relevance. The best it can do is to point towards upper airway involvement when this has not been previously suspected and this alone is sufficient reason its presence should be noted when it is noticed.

References:

Amado VM, Costa ACGA, Guiot M, Viegas CA, Tavares P. Inspiratory flow-volume curve in snoring patients with and without obstructive sleep apnea. Braz J Med Biol Res 1999; 32: 407-411.

Bogaard JM, Hovestadt A, Meerwaldt J, Mech FGAvd, Stigt J. Maximal expiratory and inspiratory flow-volume curves in Parkinson’s disease. Am Rev Respir Dis 1989; 139: 610-614.

Hadjikoutis S, Wiles CM. Respiratory complications related to bulbar function in motor neuron disease. Acta Neurol Scand 2001; 103: 207-213.

Haponik EF, Smith PL, Kaplan J, Bleecker ER. Flow-volume curves and sleep-disordered breathing: therapeutic implications. Thorax 1983; 38: 609-615.

Hoffstein V, Wright S, Zamel N. Flow-volume curves in snoring patients with and without obstructive sleep apnea. Am Rev Respir Dis 1989; 139: 957-960.

Krieger J, Weitzenblum E, Vandevenne A, Stierle JL, Kurtz D. Flow-volume curve abnormalities and obstructive sleep apnea syndrome. Chest 1985; 87(2): 163-167.

Neukirch F, Weitzenblum E, Liard R, Korobaeff M, Henry C, Orvoen-Frija E, Kaffmann F. Frequency and correlates of the saw-tooth pattern of flow-volume curves in an epidemiological survy. Chest 1992; 101: 425-431.

Rendleman N, Quinn SF. The answer is blowing in the wind: a pedunculated tomour with saw tooth flow-volume loop. J Laryng Otology 1998; 112: 973-975.

Shore ET, Millman RP. Abnormalities in the flow-volume loop in obstructive sleep apnoea sitting and supine. Thorax 1984; 39: 775-779.

Vincken W, Elleker G, Cosio MG. Detection of upper airway muscle involvement in neuromuscular disorders using the flow-volume loop. Chest 1986; 90(1): 52-57.

Vincken WG, Cosio MG. Flow oscillations on the flow-volume loop: clinical and physiological implications. Eur Respir J 1989; 2: 543-549.

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

I had mentioned previously that my PFT Lab has been questioned why the percent predicted Residual Volume (RV) measurements on some research patients were coming out so much lower at my lab than at some other PFT Labs. At that time the researchers had not shared the results they had from these patients so we could only speculate that either our RV’s were actually lower or that the predicted RV values used by the other PFT Labs were different than ours. We finally got a copy of the PFT report for one of these patients and it turns out that both answers were correct.

First, the predicted RV from what I will call Lab X (I am not familiar with the lab nor with any of the physicians or technicians there) was 15% lower than ours. My lab is using the reference equations from Stocks et al which are recommended by the ERS. I was unable to determine which sets of reference equations Lab X was using. They weren’t listed on the report and the calculated values didn’t seem to match any of the reference equations I have on hand. Our lab software uses the predicted RV plus the predicted Forced Vital Capacity (FVC) (from NHANESIII) to calculate predicted Total Lung Capacity (TLC). It is possible that Lab X’s software calculates the predicted RV by subtracting their predicted FVC from a predicted TLC.

I am not, however, going to try to argue that my lab’s predicted values are better than those of Lab X, just that they are different. The ATS has not officially recommended any particular set of lung volume reference equations and I think it would be easy to argue that all current reference equations are flawed to one extent or another. I would like to know how they were derived at Lab X but that is just to satisfy my own curiosity. Using Lab X’s predicted RV, our measured RV was 179% of predicted where by our reference equations it was 152% of predicted. This explains part of the difference in percent predicted values but by no means all of it.

When I looked at the actual test results from Lab X however, I saw numerous errors in the lung volume test and test calculations.

For comparison, here are our results.

Regardless of anything I could say about the actual test quality (and these results are from a final report not the raw test data so there is not a lot I can say about it), it is evident that RV could not have been calculated correctly. The most significant errors in the results from Lab X are:

• a negative Expiratory Reserve Volume (ERV)
• an RV that is larger than the Functional Residual Capacity (FRC)
• an Inspiratory Capacity (IC) that is larger than the Slow Vital Capacity (SVC)

Really? And in what alternate universe is this kind of math okay?

So what went wrong? Since I don’t have access to the raw test data I have to speculate a bit but I am actually familiar with this error since I have seen it before.

What is happening is that Inspiratory Capacity (IC) is calculated from the inspiratory part of the Slow Vital Capacity maneuver while the SVC and ERV are being calculated from the expiratory part of the maneuver. When this is performed correctly:

IC and ERV are smaller than the SVC (and SVC = IC + ERV). If the patient does not exhale completely however, the SVC will be lower than the IC and the ERV will be negative (and oddly enough SVC will still be equal to IC + ERV).

Alternatively, if the test system or the patient is leaking, even though the patient may actually perform the maneuver correctly the SVC, IC and ERV can be measured incorrectly and again SVC will be smaller than the IC and the ERV will be negative.

Without the raw test data it’s not possible to say which of these scenarios is the correct one, but either one would explain the numerical results and in either case the RV would be calculated incorrectly and would be significantly overestimated. This means that the other reason our percent predicted values were coming out so much lower than Lab X’s was due to sloppy testing on their part.

One other thing I noticed on the report from Lab X was the DLCO. What immediately struck me was the exceptionally high VA (9.23 L, 239% of predicted). The patient has very severe airway obstruction (FEV1 <35% of predicted) and my experience is that VA is almost always underestimated (less than measured TLC) when airway obstruction is present and the amount of underestimation usually correlates with the degree of obstruction but in this report the VA was 176% of the measured TLC. When tested in my lab, the patient’s VA was much smaller (2.89 L, 72% of predicted and 58% of our measured TLC).

This says to me that something was very wrong with Lab X’s DLCO test. Again, all I have is the final report so it is not possible to say what the specific error or errors are, but since DLCO scales almost linearly with VA, this means that the DLCO was also likely significantly overestimated and in this case both Lab X’s VA and DLCO were 3.1 times larger than my lab’s.

So what’s wrong with this picture? The fact is that the ERV was negative, the RV was larger than the FRC, the IC was larger than the SVC and the DLCO’s VA was 176% of the measured TLC. These results were accepted first by the testing software. They were then approved by the technician performing the test. Next, they were reviewed and approved by the physician interpreting the test. Finally, they were reviewed by the research coordinators who in turn gave my lab flak about our test results.

Really? We’re the ones that got the flak?

Lung volume measurements are much more difficult than they look. I’ve been in this field for more than forty years yet I continue to find subtle (and not so subtle) errors caused by software, equipment, patients and technicians. The errors in this patient’s results were not subtle however. They are actually incredibly basic errors in math and test performance yet at every step along the way the results were approved and passed on and I really can’t think of a single good reason why. It would be easy to point the finger at the technician that performed the tests but there is more than sufficient blame to be spread at every level. The simple fact is that none of the individuals involved in this patient’s tests, and that includes the programmers, technician, lab manager, clinical director, interpreting physician and research coordinators, did their job correctly.

I am not a fan of PFT Lab licensure because I think there are already too many badly-written regulations and too much bureaucracy. I am not a fan of credentialing because I don’t think credentials necessarily prove you know how to do your job. And I’ve seen more than enough bad examples to give me skeptical view of both. After seeing this report from Lab X however, I may have to change my opinion. Licensure and credentialing won’t necessarily make problems like this go away but it at least might make them less likely.

References:

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

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

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

I’ve noticed for a while that there has often been more variance between DLCO tests than I’d like to see. Some of this is of course attributable to differences in the way the patient performs each test. I am not overly surprised to see tests with different inspired volumes, different breath-holding times, different inspiratory times etc. etc. produce different results (in fact I am surprised that so many tests that have been performed differently frequently end up with almost identical results).

All too often though, I see tests that look like they were performed identically and yet have noticeably different results. For this reason I have been paying attention to small details to see if I can understand why this variance has been happening. I am well aware that there are “hidden” factors such as airway pressure (Valsalva or Mueller maneuvers) and cardiac output that can affect pulmonary capillary blood volume and therefore the DLCO. It is quite possible that much of the test-to-test variation is a result of these kinds of factors but I’ve also found several test system software and hardware errors that have lead to differences as well.

I am annoyed to say that I’ve found what could either be another system error or possibly a patient physiological factor that can lead to mis-estimated DLCO results. I’m annoyed not because I found it but because I’ve been looking at the DLCO test waveforms for a long time and never noticed this problem before. Of course since I’ve noticed it I now see it frequently.

To understand this error you have to know that during the DLCO test the gas is sampled from a port just downstream of the patient mouthpiece. Just before the test has actually started the analyzer is sampling either room air or the patient’s exhaled air. As part of setting up for the DLCO test the CO & CH4 analyzer is checked with calibration gases but not actually calibrated. The “real” calibration occurs in a separate calibration module and is usually performed first thing in the morning. The CO and CH4 zero offsets and gains obtained from the “real” calibration are used in all subsequent calculations. At the beginning of the DLCO test the patient is supposed to exhale to RV and then inhale to TLC. The DLCO test waveforms include about a second’s worth of data from the gas analyzer just before the patient inhalation.

Since this part of the test waveform comes just before inhalation the CH4 and CO traces should be zero, yet as this example shows they are actually above it. This is not a large amount of offset, but for CH4 it actually works out to be about 3.5% of full scale. Assuming that this offset is real then for this test the exhaled CH4 would likely be overestimated by somewhere between 1.5 and 2.0 percent of what it should be, which means that VA and therefore DLCO would be underestimated by about the same amount. Realistically this is probably not enough to make a significant difference in the results but it bothers me because it shouldn’t be there in the first place.

So where is this offset coming from? I see two possibilities and they aren’t mutually exclusive.

First, since at this point in the test the gas analyzers are sampling the end of the patient’s exhalation, this could be residual gases from previous tests. In favor of this interpretation is the fact that the above example is from a patient’s third DLCO test. This patient’s first test showed no offset and the second test about half the offset seen here. Additionally, the CO waveform is offset less than the CH4 waveform which also makes sense since CO can be absorbed in the lung far more easily than CH4.

What goes against this interpretation is that the offsets are usually a lot more random and often don’t follow test order. I have seen noticeable offsets in CH4 and CO in a patient’s first test and no offsets in their second or third tests. I’ve also see tests where the analyzer traces did not appear at all and when I looked at the numerical data it was because they were negative values. As I mentioned, the analyzer’s zero offsets are often obtained many hours before a given test and I’ve yet to see an analyzer that didn’t drift at least a bit so is quite possible the offset is due to analyzer drift.

As I said, these two interpretations are not mutually exclusive. Depending on how long an interval there is between DLCO test there may well be some residual CH4 and CO in the patient’s exhaled air. I have previously advocated that a patient’s exhaled CO level should be checked before each DLCO test since elevated PACO levels will decrease the measured DLCO. I hadn’t thought about CH4, but if CO is being retained between tests then of course so will CH4.

Since analyzer drift is probably never going to go away, I do not understand why the analyzer isn’t re-calibrated for every test. The calibration and pre-test analyzer check procedures go through exactly the same steps and take exactly the same length of time. There may well be a valid reason why it isn’t routinely re-calibrated, but so far nobody has been able to explain this to me and it is not documented anywhere either.

[The lack of documentation is a point that bothers me a lot. The only reason I even know that there is a difference between a “real” calibration and a pre-test analyzer check is from inspecting the database and getting my suspicions confirmed in a face-to-face conversation with one of the engineers that developed the software. I’ve found a number of small details in the DLCO and other tests that are not documented yet have the potential to have a significant effect on the results. I am concerned about this because I see research papers where the investigators say they did a such-and-such test using equipment from the so-and-so manufacturer yet nowhere do I see that they made any attempt to determine whether the equipment is actually producing accurate results. I am not necessarily asking that manufacturers openly publish the proprietary code for their test systems but maybe it’s time for the ATS-ERS to not just establish the procedural standards for tests but the computer algorithms for them as well.]

I have verified that this error is “real” at least in the sense that the values displayed in the DLCO test waveforms are matched by the numerical data. Since the numerical data has already been processed for the calibrated zero and gain this reflects either a signal from the patient’s residual gases or from analyzer drift. It’s not a big error and it’s unlikely to make a significant difference in the reported results, but many (many, many, many) years ago when I was a Cub Scout and we were packing our knapsacks for hikes we were told to “mind the ounces and the pounds would mind themselves”. The same principal applies here. Even though they are small each detail matters and enough of them acting together can end up reducing test results to garbage.

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

The plethysmographic technique for measuring lung volumes determines FRC first and then uses a slow vital capacity maneuver to calculate TLC and RV. FRC is defined as the volume of the lung at the end of a normal exhalation. The two components of the SVC maneuver that are used to calculate TLC and RV are the Inspiratory Capacity (IC) and the Expiratory Reserve Volume (ERV) and they too are measured relative to FRC. It would therefore seem to be important to have an accurate notion as to where FRC is in relation to TLC and RV when measuring lung volumes.

Plethysmography measures lung volumes by having a patient pant against a closed shutter and measuring pressure changes. The test is usually performed by having the patient breathe tidally for a period in order to determine where end-exhalation (FRC) is located, closing the shutter and performing the measurement, then returning to tidal breathing and performing the SVC maneuver. A critical assumption in this process is that the FRC baseline does not change while the shutter is closed.

Patients are incredibly ingenious, however, and often find a way to shift their FRC baseline during plethysmography. This is often due to a leak of some kind and when this occurs, both TLC and RV can be mis-estimated.

If the baseline shifts upwards, then TLC and RV will both be overestimated.

If the baseline shifts downwards, then TLC and RV will both be underestimated.

This is something that “everybody” knows about and is documented to some extent in the ATS-ERS guidelines and our manuals for the test equipment. At the moment, the software used in my PFT Lab does not allow us to correct for FRC baseline shifts so technicians are taught to recognize when a significant baseline shift has occurred and to not use the results from that test.

When FRC baseline shifts occur then as the last two examples show it is often quite evident and easy to recognize, but just as often it is much less clear less clear that a baseline shift has occurred and if it has, whether it is significant or not. Closing the shutter often affects how patient breathe and they may shift their baseline for this reason and not because they leaked. Eventually they may return to their original baseline, but in this example is this what happened or not?

 

Since shutter closure can affect a patient’s breathing pattern and FRC, if they drift off baseline after the shutter has re-opened is it because a change in lung mechanics or did they leak and are actually returning to their “real” FRC?

 

Finally, what is a significant change in FRC baseline? I’d rather only report the results from tests that had no baseline shift but you often have to work with what you can get. That makes this an important question and there does not appear to be a clear answer. The ATS-ERS guidelines recommend that to be reported serial measurements of FRC by plethysmograph should differ than no more than 5 percent. FRC baseline shifts however, do not affect the measurement of FRC, but of TLC and RV. 5 percent, however is probably a reasonable value to apply to TLC. Does that mean that a baseline shift less than 5% of TLC is acceptable?

 

The baseline shift in this example is approximately 0.16 L, which is 3.5% of the TLC and probably acceptable, particularly if it is the best you can get from a patient, but since there are no guidelines this is a guess. I’d also have to mention that estimating the error is also problematic because the best you can do within the test module is eyeball the graph and guestimate the volume shift (I was able to use a graphics program and apply a tape measure but that was a separate program and required a screen dump).

At a very rough estimate I’d say that about half of our plethysmographic lung volume measurements do not have any baseline shifts. Although the ATS-ERS guidelines ask for three acceptable quality tests with repeatable FRC measurements the reality is that we have to work with imperfect test results and for a variety of reasons there is a certain fraction of patients that just don’t seem to be able to perform plethysmography correctly.

I have to ask why the post-shutter FRC is not re-assessed by our lab’s testing software. It seems to me that it would be relatively simple to check a patient’s end-exhalation level both before and after shutter closure and to show any discrepancy between the two measurements. I would also appreciate the ability to “fix” TLC and RV calculations that have been incorrectly calculated from mis-estimated IC and ERV measurements caused by a FRC baseline shift. Finally I’d like to see the ATS-ERS guidelines have a better appreciation of potential testing errors from all lung volume measurement techniques with a consensus about what level of error is acceptable.

Plethysmography is still considered by many to be the gold standard in lung volume measurement. There are a number of reasons, FRC baseline shifts being one of them, that plethysmography cannot be considered to be any more accurate than helium dilution or nitrogen washout lung volumes. I am concerned that many of the possible error modes in plethysmography (and helium dilution and nitrogen washout) are not as well appreciated as they should be and that all too often results are accepted because “the computer said they were okay”.

References:

Brusasco V, Crapo R, Viegi G. Standardisation of the measurement of lung volumes. Eur Respir J 2005; 26: 511-522.

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

I’ve had several exercise tests come across my desk lately where the patient had an elevated Ve-VCO2 slope. An elevated Ve-VCO2 slope during exercise is usually taken as a sign of pulmonary vascular disease however these patients had a normal DLCO so I have been reviewing the literature to try to get a better understanding of what the Ve-VCO2 slope is trying to tell us in these cases.

Although the majority of the literature on Ve-VCO2 response indicates that it is likely due to some form of pulmonary vascular disease (micro-fracturing of the pulmonary capillaries, increased pulmonary vascular resistance, V-Q mismatching) there are some investigators that feel that in some individuals it is more likely due to an increased ventilatory chemosensitivity to CO2. It has been over 25 years since I last performed a CO2 response test and at that time there was no particular consensus on how the test should be performed. Since chemosensitivity may have a distinct bearing on Ve-VCO2 slope I thought it would be a good idea to also review the literature on CO2 response and see what has happened in the meantime.

After spending some time reading a couple dozen research papers it doesn’t seem as if much has changed. The CO2 Response landscape remains without an overall consensus and if anything has become more confusing, not less. There are two major approaches to measuring CO2 response and each of these approaches has at least two ways of analyzing the test data.

CO2 response was initially measured by having a subject breathe a series of gas mixtures containing specified concentrations of carbon dioxide (sometimes in room air, sometimes in oxygen). Ventilation was measured after alveolar gas had washed out and a steady state PetCO2 had been achieved, usually around 5 minutes. This was a laborious process, requiring not only only the time to make the measurement but for the subject to return to normal before the next level was tested. A re-breathing technique was first proposed by Read in 1967 and most research since that time has been based in one way or another on this approach.

The basic idea is that a large capacity anesthesia bag is filled with a mixture of 7% CO2 and 93% O2. The volume of gas in the bag is determined by the patient’s vital capacity plus 1 liter. The patient starts by breathing room air and the end-tidal CO2 is measured. The patient is then switched into the bag and as they re-breathe the gas mixture the CO2 concentration rises. As the level of CO2 increases, the patient’s minute ventilation increases as well. The test generally lasts around four minutes and is terminated when the CO2 concentration reaches 10% or the patient can no longer tolerate the sensation of dyspnea.

The first CO2 re-breathing response tests systems were primarily mechanical in nature. This is an example of a typical bag-in-a-box system:

More recently, researchers have simplified the test system by re-purposing metabolic measurement systems originally intended for exercise testing:

 

The primary measurements in these systems is end-tidal CO2 (PetCO2) and minute ventilation (Ve). CO2 response is calculated by:

 

Measuring the response to CO2 in this way has been criticized because patients with lung disease may be limited in their ability to increase their ventilation. For this reason, correction factors based on the patient’s maximum predicted minute ventilation have been developed. For example:

 

When expressed this way the results from patients with lung disease who would otherwise be expected to have an intact respiratory drive appear to more closely match those of patients with normal lungs.

Other investigators have argued that ventilation alone is not an accurate indication of respiratory drive and have instead advocated the use of inspiratory pressure. One way in which respiratory drive manifests itself is by the amount of force muscles exert during inspiration. When a person’s airway is occluded without their knowledge there is a period between the time an inspiration begins and the point at which their nervous system realizes that no airflow is occurring and stops inspiring. This period is approximately 200 to 300 milliseconds long and inspiratory pressures measured during this time are supposed to reflect true respiratory drive. Investigators have settled on the inspiratory pressure 100 milliseconds after the beginning of inspiration (P_0.1 or P₁₀₀) because it is long enough for the inspiratory muscles to be completely activated yet well before the time inspiration ends because of the occlusion.

 

The testing system consists of a pressure sensor, a fast-acting shutter valve, a two-way valve and the re-breathing bag. The shutter valve is designed to be as quiet as possible and is kept out of the patient’s range of vision. This is because if the patient anticipates that the shutter valve will be closed they will alter their inspiration. P_0.1 is measured by waiting until end-exhalation, closing the shutter valve, measuring the inspiratory pressure 100 milliseconds after the beginning of inspiration which is taken as the point where airway pressure goes negative, and then re-opening the shutter valve 200-300 milliseconds after that. P_0.1 is measured first while the patient is breathing room air. The patient is then switched into the re-breathing bag. P_0.1 is measured at irregular intervals (again to prevent anticipation of the shutter valve closing) until CO2 reaches 10% or the patient can no longer tolerate the sensation of dyspnea. CO2 response is then calculated as:

 

As with the ventilatory approach to measuring CO2 response, this method has been criticized because patients with neuromuscular disease or who for other reasons (i.e. flat diaphragm in COPD) have a limited inspiratory force would appear to have a lower respiratory drive. For this reason correction factors based on the patient’s MIP have been developed.

 

An individual’s response to CO2, however it is measured, is a critical component in their ventilatory response to exercise. Why after all these years of study hasn’t it become a standard clinical test? First, because as I’ve already mentioned there is no consensus on how to measure and how to calculate it. There are two major techniques (three if you count the stead-state approach) and within each of these techniques at least two ways to calculate the results and correct for underlying conditions. The test systems used to measure CO2 response also differ not just between techniques, but within each technique as well.

More importantly, there is no clear consensus on what a normal response is. A review of an admittedly limited number of articles for Ve/PetCO2 response has shown that normal values range anywhere from 1.3 +/- 1.0 to 3.8 +/- 1.0. L/min/mmHg. For P_0.1 I’ve found a range of 0.17 +/- 0.11 to 0.60 +/- 0.10 cmH2O/mmHg. To confound this even more, individual CO2 response has also been shown to depend on, among other things, fatigue, meal composition and the time of day it is measured. Because of the lack of an accepted normal range, studies usually compare groups of normals subjects to patients with a specific condition. A problems with this is that in many cases the range of normal values is significantly greater than the between-group differences for normals and patients with such diverse problems as obesity, sleep apnea, CHF and COPD.

Finally, it should be remembered that the true drive to respiration comes from arterial PCO2 (and pH) and that end-tidal CO2 is just a convenient substitution. This fact alone creates a level of uncertainty for the results from almost all CO2 response studies.

I have no doubt that CO2 affects ventilation. I have no doubt that certain medical conditions can affect an individual’s response to CO2. I have no doubt that an individual’s response to CO2 can affect their exercise ventilation. The only thing that is clear however, is that when it comes to measuring CO2 response, there is no clarity.

What I am left with is a great deal of uncertainty about how much the exercise Ve-VCO2 slope is affected by an individual’s response to CO2 and how much it is affected by the individual’s underlying conditions. At the moment I am going to have to leave it that the only time I can truly suspect that an elevated CO2 response is a factor for an elevated Ve-VCO2 slope is only when there is no evidence for any form of pulmonary vascular disease, but it’s going to have to be left as a suspicion because I don’t see a way to verify it.

References:

Altose MD, McCauley WC, Kelsen SG, Cherniak NS. Effects of hypercapnia and inspiratory flow-resistive loading on respiratory activity in chronic airways obstruction. J Clin Invest 1977; 59: 500-507.

Berkenbosch A, Bovill JG, DeGoede AD, Olievier ICW. Ventilatory CO2 sensitivities for Read’s rebreathing method and the steady-state method are not equal in man. J Physiol 1989; 411: 367-377.

Elliot MW, Mulvey DA, Green M, Moxham J. An evaluation of P 0.1 measured in mouth and oesophagus during carbion dioxide rebreathing in COPD. Eur Respir J 1993; 6: 1055-1059.

Mador MJ, Tobin MJ. The effect of inspiratory muscle fatigue on breathing pattern and ventilatory response to CO2. J Physiol 1992; 455: 17-32.

Read DJC. A clinical method for assessing the ventilatory response to carbon dioxide. Aust Ann Med 1967; 16: 20-32.

Rebuck AS. Measurement of ventilatory response to CO2 by rebreathing. Chest 1976; 70(1): S118-S121.

Sin DD, Jones RL, Man GC. Hypercapnic ventilatory response in patients with and without obstructive sleep apnea. Do age, gender, obesity and daytime PaCO2 matter? Chest 2000; 117: 454-459.

Tomita T, et al. Attenuation of hypercapnic carbon dioxide chemosensitivity after postinfarction exercise training: possible contribution to the improvement in exercise hyperventiatiion. Heart 2003; 89: 404-410.

Whitelaw WA, Derenne JP, Milic-Emili J. Occlusion pressure as a measure of respiratory centre output in conscious man. Respir Physiol 1975; 23: 181-199.

Zwillich CW, Sahn SA, Weil JW. Effects of hypermetabolism on ventilation and chemosensitivity. J Clin Invest 1977; 60: 900-906.

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

Well, everything, actually. Surprisingly enough they are intimately related to each other.

One of my current projects is to develop specifications for computer software designed to analyze spirometry results. Determining the “real” start of a spirometry effort using back-extrapolation is a critical part of accurately measuring FEV1 and all other timed values (FEV3, FEV6, TET). The ATS-ERS statement on spirometry includes recommendations for the back-extrapolation process, but this explanation shows its roots in old-school volume-time oriented spirometry:

“For manual measurements, the back extrapolation method traces back from the steepest slope on the volume-time curve. For computerized back extrapolation it is recommended that the largest slope averaged over an 80-ms period is used.” 

I was thinking about how to write software to do this when it occurred to me that the steepest slope of the volume-time curve is by definition the peak flow. This is probably something like re-inventing the wheel because I am sure this has been noticed before (probably by all the programmers that have done this before me) but I’ve never seen it written up this way.

The software algorithm for back-extrapolation turns out to be very simple. Asssuming that the spirometry effort has already been digitized and that an array of paired flow and volume data points is available, the software performs a rolling 80-millisecond average of through the flow data and stops when the highest average expiratory flow (Peak Flow) has been found. Using the volume and time from the center of the averaging period, the back-extrapolated time zero is:

 

and the extrapolated volume is the volume in the array at time zero.

This got me to thinking about some of the details of the back-extrapolation process. The ATS-ERS back-extrapolation specifications are based on time (80 milliseconds), not on volume. This means the volume over which the back-extrapolation slope is averaged scales with the peak flow and not with the FEV1 or FVC or any other spirometry value.

This also means that the back-extrapolation slope volume is small when peak flow is low and large when peak flow is large. When the slope volume is low the potential for error in the calculated slope probably increases. This probably doesn’t make a significant difference in the determination of time zero but it adds a level of uncertainty I wasn’t aware of.

The back-extrapolation precess makes sense when you are thinking about a spirometry effort in terms of volume and time, but when a spirometry effort is viewed as a flow-volume loop, it is evident that the peak flow itself occurs within a short period of time and that flow rates can change significantly within the volume used to define the slope of the back-extrapolation.

 

It’s possible that this volume when viewed in a volume-time curve would look flat, but it is quite evident that it isn’t. This makes it far from clear to me exactly what slope it is that we are back-extrapolating.

Does any of this make the ATS-ERS back-extrapolation process incorrect? Probably not. I fully understand the need for back-extrapolation. The start of a forced vital capacity maneuver cannot be instantaneous and there needs to be a standardized way of determining a starting point for measuring time. Given the threshold for an acceptable extrapolated volume (5% of the FVC or 0.15 L, whichever is greater) and the wide variety of expiratory flows and volumes that are seen in spirometry this process performs reasonably well. What is clear however, is that the back-extrapolation process has its roots in a time when a felt-tip pen attached to a water-seal spirometer traced a volume-time curve on graph paper and we had to put a ruler next to the tracing and draw a line to back-extrapolate the “real” beginning of a spirometry effort. That process depended on the eye of the technician drawing the line and although the process is now computerized some of the parameters (like the 80 millisecond averaging period) are still relatively arbitrary.

References:

Brusasco V, Crapo R, Viegi G.  ATS/ERS Standardisation of Lung Function Testing: Standardisation of Spirometry.  Eur Respir J  2005; 26: 319-338.

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

A physician associated with my PFT lab has become an investigator for a device study intended for patients with severe COPD. One of the major criteria for patients to be able to enroll in this study is a severely elevated Residual Volume (RV). Patients who have met this criteria at other PFT labs in New England have been referred to this study but when they have been re-tested in my lab their Residual Volumes are coming out lower and almost none of these patient have met this criteria. We have been asked why this is the case because they are now having difficulty finding patients that qualify for the study.

We have not been given access to the original PFT reports for these patients and have not been able to actually compare results on a case by case basis. For this reason we can only offer two possible reasons. First, that my lab may not be using the same reference equations for RV that other labs are. Second, that these patient’s RV’s may have been overestimated at other labs because of errors in testing.

To compare predicted RV’s I was able to find a dozen different reference equations for RV in adult males and females. These equations are mostly for Caucasian populations, but I was also able to find at least one reference equation each for Black, Asian, Indian, Iranian and Brazilian populations as well.

When the reference equations are graphed there are some immediately noticeable outliers. Some of these outliers make sense. For example Indians [L] and Asians [A][B] have been noted to have a lower FVC and a lower TLC than Caucasians by a number of investigators and this also appears to be the case for their respective RV reference equations.

The reference equations, [E]h [E]p, for an elderly subject group also appear to be outliers. I have seen this before in reference equations for spirometry and DLCO and I am not sure if it is an indication that the studies that include younger subjects do not analyze their data correctly (differences get smeared out when you analyze by simple linear regression) or whether the elderly subject group itself is an outlier. In this case, the results from helium dilution and plethysmographic lung volumes are at significant disagreement with each other and for this reason alone I consider these studies to be suspect.

I also suspect there may have been some typographical errors in the published Iranian reference equations [F] both because they are extreme outliers and because the Residual Volumes calculated from the equations don’t match the numbers discussed in the text of the article. I left this in however, because this is not the first time I have found typographical errors in published reference equations and it is yet another potential error source we should all be on guard about.

 [Something we should all be concerned about is that we have limited ability to check on the reference equations used by our test systems. Who among us actually has a copy of Crapo’s original paper on lung volumes (Bull Eur Physiopathol Respir 1983; 18: 419-425)? I don’t, and have had to rely on articles that I can access that discuss it in detail and there’s no guarantee that transcription errors haven’t occurred somewhere along the way. Even when a source article is available on-line, it is often behind paywalls and who can afford to repeatedly pay $30-$50 per article just to check the dozens of reference equations on their test systems? Here’s a thought: why don’t the manufacturers of our test equipment include copies (PDF’s, of course) of all the articles their reference equations come from? They could get a bulk rate from the publishers and it would just be part of the cost of the test systems rather than having to come out of our operating budget.]

I am most concerned about the reference equation for Black males [H]. As far as I have been able to determine this is the only study ever performed regarding lung volumes other than FVC in blacks. It was based on a population of 77 black miners in England and lung volumes were estimated radiographically. This one study appears to be the primary basis for the assumption that the black predicted values for lung volumes are 12% less than Caucasians. The subjects were matched with Caucausians whose lung volumes were measured by the same technique so the 12% difference may be correct, but it is still an extremely limited subject population both in age, locale, profession and gender (no females).

Looking solely at the remaining Caucasian reference equations, for males of average height and weight, the values are actually clustered fairly tightly. For adult males there is a range of 1.43 [D] to 1.68 L [J] at age 20 (~10%) and 2.48 [D] to 2.72 L [J] at age 70 (~6%). There is greater variance for the female reference equations, with a range of 0.86 [D] to 1.63 L [J] at age 20 (>40%) and 1.86 [D] to 2.46 L [J] at age 70 (>28%).

My lab uses [K] as our reference equation. This equation is recommended by the ERS and by extension, to some extent by the ATS. For both males and females throughout the age range, the values derived from this equation are pretty much in the middle of the pack. That doesn’t necessarily mean that it is more accurate, just that the difference between it and the other reference equations is about half the whole range. To me this means that regardless of which reference equations are used at other PFT Labs, the possible differences in predicted RV are likely to be small.

Having said that, predicted RV isn’t necessarily derived directly from a reference equation for RV. In looking at our test systems I was interested to note that predicted TLC is not derived from its own reference equation but from [predicted RV + predicted FVC]. This makes sense in a way because the predicted SVC derived from research into lung volumes will not likely match the predicted FVC. FVC has been studied in far more detail with much larger populations and its reference equations are far more likely to be accurate than those for SVC. In addition, a PFT report that showed different predicted values for FVC and SVC would be confusing and for this reason alone most everybody defaults to using the predicted FVC rather than a predicted SVC. However, it is just as easy to derive RV from TLC – FVC as it is to derive TLC from RV + FVC and I could argue that this is actually the more correct approach.

In my experience, patients find it much easier to perform the Inspiratory Capacity (IC) part of a vital capacity maneuver than the Expiratory Reserve Volume (ERV) part. The ERV part is more difficult and more uncomfortable and for this reason patients often terminate their exhalation earlier than they should. Since TLC = FRC + IC and RV = FRC – ERV this means that, in my opinion at least, the measurement of TLC and the reference equations for predicted TLC are more likely to be accurate than the measurement of RV and the reference equations for RV. For this reason it may actually be better to derive RV from TLC – FVC than to derive TLC from RV + FVC. Since I don’t know which test systems the prospective study patients had previously been tested on and since I don’t know how all manufacturers derive their predicted RV, I can’t say how closely our predicted RV actually comes to that of these patient’s prior tests.

The notion that TLC tends to be more accurate than RV leads me to the next possible reason for the discrepancy in percent predicted RV between my lab and other labs and that is that RV was overestimated when measured elsewhere. Since my lab is of course perfect in every way and a shining beacon of rectitude that means that our RV measurements are always exceptionally accurate and the results from all other labs are always suspect.

{Sigh}.

I wish that was the case but I think we probably have as many problems measuring lung volumes as everybody else does. I do think however, that we are always very aware of the problems that can be encountered during lung volume measurements. A fair amount of time is spent training new staff on the common problems with our test equipment and updating all staff when new problems are discovered. I also review most, if not all, of the raw data for the lung volume tests and correct the selected values whenever I think it is necessary. The selection of reported FRC and SVC values is often a judgment call that is not black and white but shades of gray and I will email a technician the reasons I made a change whenever I think they have made a choice I didn’t agree with. I’d like to hope that all of this means that our lung volume test results are at least as accurate as those from the best labs anywhere.

Accurate TLC and RV measurements have to start with an accurate FRC measurement. Any kind of a leak during a helium dilution or nitrogen washout test will cause FRC to be overestimated. When FRC is measured plethysmographically in the patient population (severe COPD) that are the prospective subjects for the study too fast a pant will cause FRC to be overestimated. When FRC is overestimated then RV will also be overestimated.

Even when FRC is measured accurately, an accurate RV depends on an accurate ERV measurement. Patients with severe COPD have difficulty with this part of the SVC maneuver in particular because underlying SOB will cause them to terminate exhalation early and because too fast of an exhalation can cause airway collapse and an increase in gas trapping (this is why SVC is often significantly larger than FVC in patients with COPD). When ERV is underestimated then RV will be overestimated.

Most errors, therefore, in performing lung volume measurements tend to lead to an overestimated RV.

Since we weren’t given access to these prospective study patient’s prior PFT reports, I cannot definitively answer why my lab tends to get a lower percent predicted Residual Volumes than other labs. I’d like to think it was solely because our testing quality is better, but that is vanity speaking and it could be just as well due to differences in the predicted RV. It was interesting to find out that our predicted TLC was derived from predicted RV and predicted FVC. This fact is not documented in the manual and it took some significant digging into our system’s software in order to find this out. I don’t necessarily blame the manufacturer for the lack of documentation since users are allowed to select numerous different reference equations and it is not possible to anticipate all possible selections, but I would like to have seen it easier to find this out.

Male adult RV reference equations data set:

Source:	Number:	Ethnicity:	Age Range:	Height Range:	Technique:
[B]	101 M+F	Chinese	21-55	N/A	He Dilution
[C]	305	Caucasian	18-88	168 +/- 8	He Dilution
[D]	251 M+F	Caucasian	15-91	N/A	He Dilution
[E] h	132	Caucasian	65-85	166 +/- 6 cm	He Dilution
[E] p	132	Caucasian	65-85	166 +/- 6 cm	Pleth
[F]	845	Iranian	6-85	114-190 cm	Pleth
[G]	300	Caucasian	20-80	Not reported	Pleth
[H]	77	Black	34.9 +/- 11.9	177.9 +/- 7.4	Radiographic
[I]	50	Brazilian	20-80	155.5 – 185 cm	N2 Washout
[J]	300	Caucasian	20-70	170 +/- 6.6 cm	Pleth
[K]	N/A	Caucasian	18-70	155-195	N/A
[L]	760 M+F	Indian	15-65	N/A	N/A

Male adult reference equations: 

Source	Equation:
[B]	(0.035 x height) + (0.020 x age) – (0.019 x weight) – 3.91
[C]	(0.043 x height) – (0.012 x weight) + (0.024 x age) – 5.654
[D]	(0.022 x height) + (0.021 x age) – 2.84
[E] h	(0.02261 x weight) + (0.03978 x age) – 2.703
[E] p	(0.0000003551 x height^3) + 0.315
[F]	(0.011 x height) + (0.04 x age) – 0.521
[G]	(0.020 x height) + (0.021 x age) – 2.443
[H]	(0.0276 x height) + (0.05 x age) – 4.73
[I]	(0.0197 x height) + (0.0141 x age) – 2.08
[J]	(0.022618 x height) + (0.020664 x age) – 2.688
[K]	(0.0131 x height) + (0.022 x age) – 1.23
[L]	(0.019 x height) + (0.007 x age) – 1.945

Female adult RV reference equation data sets:

Source:	Number:	Ethnicity:	Age Range:	Height Range:	Technique:
[A]	197	Asian	17-56	142.2-175.3 cm	He Dilution
[B]	101 M+F	Chinese	21-55	N/A	He Dilution
[C]	286	Caucasian	18-88	156 +/- 7 cm	He Dilution
[D]	251 M+F	Caucasian	15-91	N/A	N/A
[E] h	189	Caucasian	65-85	152 +/- 6 cm	He Dilution
[E] p	189	Caucasian	65-85	152 +/- 6 cm	Pleth
[F]	642	Iranian	7-73	119-179 cm	Pleth
[G]	327	Caucasian	20-80	Not reported	Pleth
[I]	50	Brazilian	20-80	145.2 – 175 cm	N2 Washout
[J]	182	Caucasian	20-70	158 +/- 6.6 cm	Pleth
[K]	N/A	Caucasian	18-70	145-180 cm	N/A
[L]	760 M+F	Indian	15-65	N/A	N/A

Female adult reference equations:

Source	Equation:
[A]	(0.0224 x height) + (0.010 x age) – 2.307
[B]	(0.020 x height) + (0.009 x age) – (0.016 x weight) – 1.336
[C]	(0.023 x height) + (0.014 x age) – 2.790
[D]	(0.020 x height) + (0.020 x age) – 2.84
[E] h	(0.02266 x height) – 2.07
[E] p	(0.01678 x height) – 1.039
[F]	(-12.64 x age) + (12.821 x age^0.997) + (0.003 x height) – 0.77
[G]	(0.020 x height (cm)) + (0.015 x age) – 2.314
[I]	(0.0259 x height) + (0.0091 x age) – 3.15
[J]	(0.01131 x height) + (0.01651 x age) – 0.562
[K]	(0.0181 x height) + (0.016 x age) – 2.00
[L]	(0.014 x height) + (0.007 x age) – 1.167

References:

[A] Ching B. Horsfall PAL. Lung volumes in normal Cantonese subjects: preliminary studies. Thorax 1977; 32: 352-355

[B] Chuan PS, Chia M. Respiratory function test in normal adult Chinese in Singapore. Singapore Med J 1969; 110: 265-271.

[C] Cordero PJ, Morales P, Benlloch E, Miravet L, Cebrian J. Static Lung Volume: Reference values from a Latin population of Spanish descent. Respiration 1999; 66: 242-250.

[D] Crapo RO, Morris AH, Clayton PD, Nixon CR. Lung volume in healthy nonsmoking adults. Bull Eur Physiopathol Respir 1983; 18: 419-425

[E] Garcia-Rio F, Dorgham A, Pino JM, Villasante C, Garcia-Quero C, Alvarez-Sala R. Lung volume reference values for women and Men 65-85 years of age. Am J Respir Crit Care Med 2009; 180: 1083-1091.

[F] Golshan M, Amra B, Soltani F, Crapo RO. Reference values for lung volumes in an Iranian population. Introuducing a new equation model. Arch Iranian Med 2009; 12(3): 256-261

[G] Gutierrez C, et al. Reference values of pulmonary function tests for Canadian caucasians. Can Respir J 2004; 6: 414-424.

[H] Lapp NL, Amandus HE, Hall R, Morgan WKC. Lung volumes and flow rates in black and white subjects. Thorax 1974; 29: 185-188.

[I] Neder JA, Andreoni S, Castelo-Filho A, Nery LE. Reference values for lung function tests. I. Static Volume. Braz J Med Biol Res 1999; 32: 703-717

[J] Roca J, et al. Prediction equations for plethysmographic lung volumes. Respir Med 1998; 92: 454-460.

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

[L] Upwadia FE, Sunavala JD, Shetye VM. Lung function studies in in healthy Indian subjects. J Ass Phys India 1987; 36: 491-496

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

We’ve been aware of this particular issue for several years. My PFT lab has had some turnover lately and the newer staff aren’t familiar with this problem so it has re-appeared in some of the reports I have been reviewing.

Our lab has a mix of flow-based and volume-based test systems. This problem is peculiar to only the volume-based systems that have a vertically mounted volume-displacement spirometer and is due in part to mechanical issues but also to some underlying assumptions made by the test software.

During the set-up process for a DLCO test, the spirometer bell and the systems tubing and patient manifold are first flushed and then the spirometer bell filled with the DLCO test gas mixture. The balloon valves in the patient testing manifold are then closed to keep the spirometer bell in position and do not open again until the start of the DLCO test.

It is usually obvious if any of the balloon valves in the manifold are leaking. Either the testing system cannot complete the set-up process or the spirometer bell can be seen to be dropping when it should be remaining stationary. This requires that the balloon valve have a fairly significant leak, however, and a small leak may not be so evident.

Usually we have the patient perform the DLCO test as soon as the set-up is done. When this is the case a small leak has little effect on the DLCO test. Occasionally however, and for any number of good reasons (the patient has to go the bathroom or wants a drink of water, etc.) there may be a prolonged period between set-up and testing. When this period of time is long enough there may be a significant leak from the spirometer bell even when the balloon valves have passed all of our regular leak tests.

The problem occurs because the test software assumes that the position of the spirometer bell at the beginning of the DLCO test is the same as it was at the end of the set-up process. For this reason, when there is a leak and the spirometer bell position drops the inspired volume will be overestimated because the software does not take into account the new position of the spirometer bell. Since Alveolar Volume (VA) is a function of inspired volume:

VA will be overestimated when the inspired volume is overestimated. Since the calculated DLCO scales with VA:

this means that DLCO will be overestimated when the inspired volume is overestimated (and vice versa).

The lab equipment is regularly tested for leaks, but valves can (and all too frequently do) fail. Tubing can also develop leaks and any of these leaks can run the gamut from large and obvious to small and subtle. The work-around for this problem has been first to make the staff aware it can happen and what effect it has on the DLCO test (as well as lung volumes and spirometry, of course). Second, either to always perform the DLCO test shortly after the set-up has finished or if too much time has passed before the patient is ready to perform the test, abort it and start over.

The real solution would be to have the DLCO testing software re-check the position of the spirometer bell at the beginning of the test and either re-establish the zero volume baseline or stop the test if the bell has dropped too far. This error was passed on to the manufacturer of our test equipment when we first noticed it but so far a fix has not appeared in our software updates.

This problem is unique, as far as we can tell, to our volume-based testing systems. A larger point though, is that anything that can cause the inspired volume to be mis-measured and either over- or under-estimated will have a corresponding effect on the calculated DLCO. We do not see baseline shifts in our flow-based test systems but that doesn’t mean that inspired volume hasn’t been underestimated during the test itself due either to a patient or system leak (on these systems I have a lot of trouble seeing how inspired volume could be over-estimated other than with a mis-calibrated pneumotach). When I review DLCO test results I often see inspired volumes that are less than the FVC, sometimes substantially so. That may be just because that is the way the patient performed the test but unlike the pre-test leak with our volume-based test systems there is nothing particularly evident about a leak during inhalation.

Leaks are the bane of a Pulmonary Function Lab’s existence. They are a chronic problem and any lab that doesn’t think it has a problem with leaks probably isn’t looking hard enough. Depending on when and where they occur leaks can cause the results from any test to be over- or under-estimated. The only answer is to be familiar with your equipment’s failure modes and how they show up in test results. Having said all of this, I also have to say that equipment manufacturers often don’t make it particularly easy to determine if their test equipment is leaking or to find the source of the leak. From the equipment manufacturers point of view, as long as a leak isn’t from a manufacturing flaw it’s not their problem. I would argue that more care should go into valve design and the ability to more rigorously test for leaks, but these don’t seem to be factors that sell test systems and that’s at least partly our own fault.

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