In this multi-centre study we found that in a realistic clinical setting the smallest detectable change for the volume measurements, VC and VA, was smaller when measured by pneumotachograph than by mass flow sensor. A similar result was found for the smallest detectable difference for DLCO, but did not reached statistical significance.
These differences were mainly due to a lower measurement error (lower variance, Figure 2) of the pneumotachograph measurements. The measurement error in this study was reflected in absolute values by the standard error of measurement as well as the coefficient of variation. The coefficient of variation values in this study were similar to the values found in long term repeatability measurements in healthy subjects reported by Pennock et al.  and Jensen et al.  These studies showed that the measurement error of the simple VC assessment was smaller than the measurement error of the much more comprehensive single breath diffusion measurement. Our study confirmed these findings.
Although reliability of pulmonary function tests is commonly estimated by the coefficient of variation Hankinson et al. recommend expressing pulmonary function test variability in absolute terms . The coefficient of variation provides information about the measurement error related to the mean value of a sample of repeated measurements over time. This mean value of one individual patient is usually not known, as in clinical practice usually one single measurement for an individual is available. For that reason the smallest detectable change, which is important for clinical decision making, cannot be estimated by the coefficient of variation. Nonetheless, from our study it can be seen that a coefficient of variation difference for VC of 1.6 percent between the measurent devices (Figure 1) doubles smallest detectable change from 0.25L to 0.53L (table 2). This means for clinical practice that a difference between two consecutive VC measurements in time smaller than 250 ml measured by pneumotachgraph and smaller than 530 ml when measured by a mass flow sensor can not be distinguished from measurement error.
An additional important advantage of the standard error of measurement (Figure 2) is the possibility to analyze different sources of error variance (between-apparatus, between occasion and random). This knowledge provides direct information of the main cause of error. Quality control management of a pulmonary function laboratory should, when possible, act on this knowledge. In case of a large between-apparatus variation one should investigate and solve the cause of this undesirable systematic difference. Our results show that for all pulmonary function parameters random variation is the main source of error for both measurement devices, with the largest values in the mass flow sensor. Other errors are relatively small. Consequently, total measurement variation, which is the sum of all error sources, for all pulmonary function parameters is, except for the DLCO/VA, are higher in the mass flow compared to the pneumotachograph data. Random variation can be due to (subject- biological, coincidence (or circumstance) and unsystematic instrument error. Since there are no differences in the way the pulmonary function tests are performed on both devices, we expect the biological and circumstance variations to be comparable between both devices. Thus, unsystematic error differences between the measurement devices seem to be the most likely explanation for the difference in random error between the mass flow sensor and pneumotachograph measurements. An increased between apparatus variation for the VC found in the mass flow sensor measurements suggests a higher systematic difference between the mass flow sensor compared to the between pneumotachograph apparatuses. Increased occasion variation points to a lower degree of repeatability for VC and DLCO of mass flow devices, compared to the pneumotachograph devices. It is unlikely that the differences in measurement error for the pulmonary function parameters between mass flow and pneumotachograph measurements are due to differences in absolute values because mean outcome results from the three "mass flow sensor" hospitals were not different from the values measured in the five "pneumotachograph" hospitals.
In clinical practice knowledge of the smallest detectable change in patients would be most informative. Apparently, this knowledge is difficult to obtain because one needs repeated measurements from long term, exceptionally stable, patients. Nonetheless, Pennock et al. showed a larger coefficient of variation in healthy subjects compared to patients in pulmonary function measurements. Therefore, we speculate that the smallest detectable difference in patients is at least as large as the values estimated by the healthy subjects in this study.
In summary our results show that the total measurement error of one single VC or VA measured by pneumotachograph is lower compared to these measurements obtained by mass flow sensor. Consequently, the smallest detectable significant change between individual VC and VA measured by pneumotachograph are smaller than when measured by mass flow sensor measurements. Therefore, the pneumotachograph is the preferred instrument to estimate lung volume change over time in the individual patient.
Although pneumotach performs better than the mass flow sensor in terms of smallest detectable difference, the latter has the advantage that it can be used during exercise and is probably less influenced by temperature or pressure.