Interpreting pulmonary function tests.

The first two articles that I wrote for FOCUS Journal, published in the Winter and Spring 2004 editions, addressed test quality and the importance of separating the signal from the noise in measurements. This article will focus on what the signal means once it has been isolated.

Interpretation of all observed data involves comparisons with reference data. The patient coming to your laboratory is most likely there because a physician has reviewed their history and physical examination information, compared it to his/her expectation of "normal" and determined the patient has a health problem and that the deviation from health is likely related to their lungs. The patient's measured lung function values are compared against other lung function tests on healthy individuals and, if the values are outside the defined "normal" range they are then compared to patterns known to be associated with various lung diseases.

In this article, we will focus on the comparisons of data measured in the laboratory to reference values. The most accurate and precise comparison would be to compare a single subject's values against their own values measured at an earlier time when they were healthy (formally called comparison to self). This kind of testing is most often done in an occupational or health care setting when the subject will be exposed to substances or medications with known pulmonary toxicity and where serial measurements may allow for an earlier intervention if injury occurs. Comparison to self is best because changes as small as 5% can be detected. The next best comparison would be to a healthy identical twin.

[ILLUSTRATION OMITTED]

For most people such comparisons are rarely available. As described in our scenario above, most patients seen in pulmonary function laboratories have been sent there because a clinician has observed something that suggests an abnormality. The values measured in a pulmonary function lab are then compared first with average values drawn from a reference population of healthy people. The distribution about these average predicted values typically varies by [+ or -]20%. There are a number of reference equations available and predicted values can vary widely, as can interpretations based on them. As we have seen previous articles, it is important to minimize the noise. In this case, we minimize noise by selecting reference values that will reduce interpretative errors.

The reference equation or set should be a deliberate choice. It requires attention to the conditions of testing, the characteristics of the reference population and, most importantly, comparability to the testing conditions and the population in the your laboratory. As pulmonary function machines become more automated, it is easy to fall into the default choices of the manufacturer. Keep in mind that you do have a choice and that the choice is important.

The technical conditions of the testing should be comparable. If both the reference study and your laboratory use established standards for testing, the conditions for technical comparability will be met. Matching populations for biological variability is more difficult. Most biologic variability is related to height, age, and gender. Ethnicity can account for about 10% of the biologic variability. In the U.S., the most accurate equations now available for most laboratories come from the NHANES III study. These equations were built on data from a large representative sample of the United States with over sampling for the two most populous minorities (Mexican and African Americans). The measurements were made according to ATS standards with excellent spirometers and careful monitoring of test quality. If you pick these equations, be sure your tests are performed to ATS standards.

One of the important aspects of the reference equation is determining the lower (or in some cases upper) limits of the normal range. For most large reference sets, the limits are given as confidence intervals when the distribution is clearly Gaussian (the familiar bell shaped curve) or in percentiles when the distributions are skewed as they are in the mid or instantaneous flows. While fixed lower limits of the normal range (such as <80% predicted or an FEV1/VC ratio <70%) are easier to use than statistical lower limits, they turn out to be problematic. False categorizations occur and increase as height and age diverge from their average values. For example, in healthy men, the FEV1/VC ratio falls below 70% at about age 42; in women it happens at about age 52.

Choosing an appropriate lower limit of normal will not solve all the problems of interpretation. The appropriate interpretative choice is relatively easy when patients' data are clearly within or clearly outside the normal limits. The interpretation is more difficult for patients whose values are on, or very close to, the lower limit. Those borderline values should be interpreted with great caution and proper categorization will likely require comparison with data other than that available from the PFT. The complete clinical picture, including the prior probability for disease separate from the pulmonary function testing is crucial. For example, a patient who reports forty pack years of smoking, dyspnea on exertion and cough and who has an FEV1/VC ratio just above the lower limit of normal will be more confidently classified as having airway obstruction than will a person who has no symptoms and similar borderline values.

Interpretative statements should be clear, concise and helpful. Standardized interpretative schemes are available from the American Thoracic Society and European Respiratory Society. A combined statement will be available soon.

The Caveats

Keep it simple. Limit the initial interpretation to just three variables: Vital capacity (VC) (forced (FVC) or slow (SVC)), FEV1, and FEV1/VC. The number of pulmonary function variables presented by your automated pulmonary function machine presents yet another pitfall: false positive results. Increasing the number of values used in an interpretative scheme has the paradoxical effect of making the interpretation less reliable. The more tests included, the greater the chance of a false positive test. Additional tests may be helpful when the primary signal values are near the lower limits of the normal range. In this situation, other spirometric values may help to establish a pattern. But while a pattern may help, it is no substitute for clinical information about the patient.

Know the limits of epidemiological data. It is sometimes tempting to apply data shown in epidemiological studies to individual patients. For example, the FEF25-75% helps distinguish healthy smokers from healthy nonsmokers in large groups of individuals; this suggests it could be an early marker of small airways disease. For individual patients, however, it does not reliably detect small airways disease, partly because the test is too noisy.

Use the FEV1/VC ratio and VC to start the interpretation. (1) If both are normal, the spirometry is normal. If the ratio is low, airway obstruction is present and severity is categorized by the FEV1 as a percent of predicted (>70% is mild and <50% is severe). If the ratio is normal and VC is low, there is about a 50% chance that the patient has restriction (low TLC). Depending on the clinical information, you may wish to measure TLC in this setting.

Summary: The quick course in interpretation.

Pick your reference equations carefully. In most cases they should be those based on Hankinson's analysis of NHANES III data. (2) Keep your interpretative scheme simple and watch out for borderline values.

References:

1. Am Thoracic Society. Lung function testing: selection of reference values and interpretative strategies. Am Rev Respir Dis 1991; 144:1202-1218.

2. Hankinson, J.L. et al. Spirometric reference values from a sample of the general U.S. population. Am J Resp Crit Care Med 1999; 159:179-187.

by Robert Crapo MD