Assessing test validity

• Interpretation of results should begin with a review of test quality. Good test quality is important as suboptimal quality tests may impact negatively on the interpretation of results and hence on clinical decision making. Information regarding indicators of test quality is provided in the test-specific chapters and in Chapter 7.
  
• The identification of suboptimal quality results can be gleaned from examination of the raw test data, technical comments provided by the test operator or a combination of both.
  
• When a suboptimal quality test is obtained, a cautionary statement identifying the magnitude and direction of the impact of the suboptimal quality results should be included in the report. For example:

Results should be interpreted with caution as test performance for spirometry was suboptimal due to coughing at end expiration, and may result in potential underestimation of forced vital capacity (FVC).

Assessing the adequacy of reference values for the particular subject

• Lung function results are interpreted by comparing the obtained results to a known reference range.
  
• The reference range/equations chosen need to reflect the population(s) tested and the test methods used in the laboratory (1).
  
• 
  

  The reference range used for each test, as well as the limits of the variables (e.g. age, height, weight) of the reference equations, should be known to those reporting.

  
  

  If reference values are extrapolated beyond the limits of the variables (for example, a subject’s age is 85 years, but the age range of the reference set used is 8–80 years), then a cautionary statement should be included as there is uncertainty regarding the validity of the reference data. For example: Reference values for spirometry have been extrapolated for age and should be used with caution.

  
  
• 
  

  Lung function may be affected by race. Clear differences between Caucasian and African-American populations in the United States have been shown (2). Ideally, the subject’s race (or the race they identify within the case of mixed race) should be taken into account in selecting appropriate reference sets. There are, however, practical issues in identifying and using appropriate reference sets for multiple races, and appropriate reference sets for some tests do not exist.

  
  

  The Global Lung Initiative has published a multiethnic set of spirometry reference values (3), which goes some way to addressing the issue of race in reference values. At the time of writing, the Global Lung Initiative is working towards race-specific reference values for T_LCO also.

  
  

  A useful, but less than ideal solution for this problem, is the application of a race correction factor (e.g. 0.88 for FEV₁ and FVC (forced vital capacity)) to Caucasian reference values when testing non-Caucasian subjects (1). This method is by no means ideal and when a correction factor is applied, a cautionary statement should be used to inform the reader that the reference values have been adjusted for race. For example: Reference values have been adjusted for race and should be used with caution.

Determining normality or abnormality using upper and/or lower limits of normal

The normal range

• The normal range is defined by the range in which there is confidence for inclusion of 95% of the normal population.
  
• The 95% confidence limits are determined using the mean predicted value (MPV) calculated from the reference equations and the residual standard deviation (RSD) that describes the amount of scatter or variation around the MPV.
  
• The upper limits of normal (ULN) and lower limits of normal (LLN) can be calculated using the MPV and the RSD as follows:
  
  
  
  • For parameters that may have an abnormally high or low result (e.g. haemoglobin), the upper and lower 95% confidence limits are given by
    
    
    
    • ULN: MPV + 1.96RSD
      
    • LLN: MPV − 1.96RSD
      
    • The limits are set at the 2.5th and 97.5th percentiles (5% in total lie outside the normal range)
    
    
  • For parameters where it is possible to have only abnormally low results (e.g. FEV₁, FVC), the lower 95% confidence limit is given by
    
    
    
    • LLN: MPV − 1.64RSD
      
    • The lower limit is set at the 5% percentile (5% lie below the normal range)
    
    
  • For parameters where it is possible to have only abnormally high results (e.g. RV (residual volume):TLC (total lung capacity) ratio), the upper 95% confidence limit is given by
    
    
    
    • ULN: MPV + 1.64RSD
      
    • The upper limit is set at the 95th percentile (5% lie above the normal range)
  
  
• 
  

  A z-score expresses the number of standard deviations a measured result is from the mean and is calculated (measured value—MPV)/RSD. z-score values below the MPV are recorded as a negative number and values above the MPV as a positive number.

  
  

  Using the 95% confidence limits to set the upper and/or lower limits of normal.

  
  
  
  
  • Parameters that may have an abnormally low or high result: an abnormal result can be identified by a z-score either less than −1.96 or greater than +1.96, respectively.
    
  • Parameters with only abnormally low results: an abnormal result can be identified by a z-score less than −1.64.
    
  • Parameters with only abnormally high results: an abnormal result can be identified by a z-score greater than +1.64.

Determining normality or abnormality

• Limit the number of parameters used in the interpretation of lung function. The more parameters that are included in the test analysis, the more likelihood there is of returning an abnormal finding.
  
• When results are within normal limits, they should be reported as being within normal limits rather than being normal. There may be lung disease present that has not as yet forced any parameters of lung function outside the normal limits.
  
• When a result is abnormal, it is described as being reduced if it is below the LLN or elevated if it is above the upper limit of normal.
  
• Borderline results require careful consideration in interpretation and it is acceptable to describe a result as borderline.
  
• As the normal range is defined as the range in which there is confidence that 95% of the normal population will be included, 5% of the normal population will have an abnormal finding. This is a particularly important consideration when lung function is being tested in a general population in the absence of symptoms (e.g. pre-employment medicals, epidemiological surveys). In a doctor referred population dictated by specific symptoms, an abnormal finding is more likely to be a true abnormal finding.