For several years, the grading of chronic obstructive pulmonary disease (COPD) severity proposed by the Global Initiative for Obstructive Lung Disease (GOLD) has been based on Forced Expiratory Volume in 1 second (FEV1) only. This classification could not adequately predict clinical outcomes1,2 and therefore other prognostic indices, such as the BODE index3 have been proposed to stratify the severity of COPD. In 2011, the GOLD proposed a new classification of COPD severity that in addition to the level of FEV1 reduction also took into account severity of symptoms and frequency of exacerbations,4 thus generating a panel of four different stages of disease severity (A through D). This new classification, however, does not seem to have better prognostic capacity,5 and the GOLD group has recently proposed a new classification that merges the GOLD 2007 and GOLD 2011 classes.6 In this document, the severity of obstruction expressed as reduction of FEV1 is still considered a very important prognostic factor at the population level.

One potential issue with the GOLD classification is that FEV1 is expressed as the percent of the value predicted by estimating equations (FEV1PP). This approach may introduce bias, and it has been shown that GOLD classification can lead to misclassification of older patients.7,8

Due to these limitation, alternative ways of standardizing FEV1 have been proposed, such as FEV1 standardized residuals, FEV1 divided by height squared or cubed (FEV1/Ht3) or expressed as a function of the sex-specific first percentile.9 FEV1/Ht3 may be more informative than FEV1PP to predict clinical outcomes even in elderly patients.9–11 Our hypothesis is that part of the suboptimal prognostic capacity of the GOLD stratification comes from the use of equations developed in population that are different from the one that gives raise to the COPD cases. In this study, we evaluated how well the GOLD classification stratifies the mortality risk when FEV1 is expressed as percentage of the value predicted by an equation developed in the same population that gave rise to cases, externally developed equations, or as FEV1/Ht3.

The mean age of our sample was 73.2 (SD: 6), men were 57.8%. Characteristics of the population according to GOLD stages determined using the FEV1-SARA are reported in Table 1. Participants with bronchial obstruction were predominantly males, especially in classes III and IV. Beside having a higher BODE index, participants in these classes tended to have lower body mass index, higher exposure to cigarette smoking, and higher prevalence of cerebro-vascular disease. We did not observe a linear relationship between GOLD stages and prevalence of comorbidities; for example the prevalence of ischemic heart disease, heart failure, and stroke was higher in GOLD class III than in GOLD class IV.

The FEV1 predicted by the ERS equation was on average 150ml smaller compared to the SARA equation. Consequently, 21 patients classified in class I, 8 patients classified in class II, and 3 patients classified in class III using the ERS equation were classified in class II, III, and IV, respectively, according to the SARA equation. This notwithstanding, the general pattern across GOLD classes based on FEV1-ERS was not different from that observed using FEV1-SARA. Likewise, no clear specific relationships in clinical characteristics were found after stratification by quartiles of FEV1/Ht3 (Table 2).

In participants with COPD, the average distance walked in 6′ was 67.6m. In the same group, 15% reported at least one exacerbation in the preceding 3 months, and 24.5% had a modified Medical Research Council dyspnea index>2. The global SGRQ score was 39. We found no differences in the capacity of the different severity grading methods to stratify exercise capacity expressed by the distance walked in 6′ (Fig. 1, panel A). We found that the other important clinical outcomes (exacerbations, dyspnea, and quality of life) differed only across the first three GOLD classes, while participants in class IV did not differ from patients in class III. Using quartiles of FEV1/Ht3, instead, the linear association was evident across all categories (Fig. 1, panels B–D).

Fig. 1.

Distribution of walked distance (A), rate of exacerbations (B), dyspnea (C), and St. George Respiratory Questionnaire (D) according to COPD severity rated by GOLD or quartiles of FEV1/Ht3.

(0.22MB).

Data on vital status were gathered after a median of 13 years (range: 0.5–15). The risk for mortality was 19.5% at 5 years and 53,7% at the end of follow-up. Fig. 2 shows the Kaplan–Meier curves for 15-year mortality risk stratified by stages. Considering the 5-years mortality, all the classification systems showed a fairly linear relationship with mortality. Nonetheless, as reported in Table 3, the incremental changes in risk associated with each class were not statistically significant, except for the increase in risk from stage II to stage III of both GOLD classifications. The goodness of fit of the quartiles of FEV1/Ht3 was somewhat worse compared to the GOLD classifications, but the overall discriminative capacity expressed by the c statistics was 0.6 for all models. When the models were adjusted for age and sex, the incremental association between the GOLD stages defined using the SARA predicting equation was only marginally affected, while for GOLD stages defined using the ERS predicting equation the linear association was more evident, with significant increases of hazard ratio in group II vs. I, group III vs. II and group IV vs. III. No significant incremental risk was observed for quartiles of FEV1/Ht3.

Fig. 2.

Kaplan–Meier curves for 15-years risk of mortality, according to COPD severity rated by GOLD, considering FEV1-SARA (panel A) and FEV1-ERS (panel B), or quartiles of FEV1/Ht3 (panel C).

(0.28MB).

Considering the 15-years mortality risk (Fig. 2), the linear association across severity stages is more evident for the two GOLD classifications, as also confirmed by the significant increments in the hazard ratio, shown in Table 3. Although the linear association was evident also for quartiles of FEV1/Ht3, the increment in the hazard ratio was significant only for stage III compared to stage II. The goodness of fitness of the models obtained with the GOLD classification system was marginally better compared to that of the model including quartiles of FEV1/Ht3, but the overall discriminative capacity did not change between the three classification systems. After correction for age and sex, no substantial changes were observed.

Finally, when only deaths from respiratory causes were taken into account, the overall pattern did not change, although the estimates were more unstable due to the relatively low number of events (46 overall, 28 in the first 5 years).

Our data indicate that the GOLD classification system is superior to a stratification based on quartiles of FEV1 standardized by height in predicting mortality in older persons with COPD. This finding is partly in contrast with other reports9–11 showing that FEV1/Ht3 is superior to FEV1PP in stratifying the mortality risk. One possible explanation for this discrepancy is that previous studies on this issue have compared groups of FEV1PP and FEV1/Ht3 based on percentiles, and not GOLD classes vs. percentiles of FEV1/Ht3. One exception is a study by Miller et al.18 that compared GOLD classes vs. arbitrarily defined groups of FEV1 divided by height squared (FEV1/Ht2), in which the authors found that FEV1/Ht2 could better stratify mean survival compared to GOLD classes. In this study, however, the cut-off to define bronchial obstruction was set at FEV1/FVC<0.89, and the authors did not provide information on the mortality risk across groups.

A possible explanation of our findings is that the GOLD classification uses pragmatic cut-off of FEV1PP based on consensus, clinical experience, and previous epidemiological data,19 and in this study we compared it with a distribution-based classification. The use of arbitrary cut-off (as in the GOLD classification) may better fit the risk of mortality compared to distribution-based cut-off (i.e. quartiles) if the relationship between FEV1 and risk was not linear, while the opposite would be true if the relationship was linear. An alternative explanation is that FEV1PP, on which the GOLD classification is based, better captures the underlying pathophysiologic changes compared to FEV1/Ht3. Several studies, however, starting from the seminal observation by Fletcher and Peto,20 indicate that FEV1 standardized using height is a good indicator of reduction of pulmonary function with respect to the mortality risk. Further studies investigating different cut-off for FEV1/Ht3 are needed to better explore this issue.

To our knowledge, this is the first study comparing the discriminative capacity of GOLD classification and quartiles of FEV1/Ht3 with respect to walking speed, dyspnea, rate of exacerbations, and disease-specific quality of life in COPD. In contrast to what we observed for mortality, quartiles of FEV1/Ht3 could better stratify more specific clinical indicators (dyspnea, exacerbations, disease-specific quality of life) compared to the GOLD classification, while no grading system could stratify an overall indicator of exercise capacity such as the distance walked in 6′. The same observation made above about mortality risk may apply for these findings: the distribution-based classification may have a better fit because the relationship is linear. Furthermore, these outcomes were evaluated at the same time of the spirometry, therefore the discriminative capacity of the classification systems are relevant to the actual clinical conditions, not to a future outcome such as mortality. Based on these data, the GOLD classification seems better suited to identify people at generic risk for mortality, but not patients with worse health status related to pulmonary problems. In keeping with this hypothesis, the discriminative capacity of GOLD stages is better for very long-term compared to medium-term survival, and there was no difference in the discriminative capacity for mortality from any cause or mortality from pulmonary causes.

Results obtained using predicting equations developed in the same population that gave rise to the cases differed marginally from those obtained using equation developed in a different population. Thus, although guidelines recommend the use of internally developed standards,21 the use of externally developed standards seem to be acceptable. Indeed, we found a relatively small discrepancy between GOLD classes assigned by FEV1-SARA or FEV1-ERS.

Limitations of this study deserve consideration. An important limitation is that we do not have follow-up data on pulmonary and physical function and symptoms, therefore we cannot evaluate which classification system better identifies people with worse prognosis in terms of clinical course of disease. At the same time we cannot discriminate if some subjects has changed GOLD class during the follow-up. We used measured height, and this could introduce a bias as vertebral fractures are frequent in the older population (especially women) and COPD.22 Due to this problem, people with vertebral fractures might have been misclassified as having a FEV1/ht3 higher than the real. Since vertebral fractures are a risk factor for adverse outcomes in COPD,23 this bias may have caused a reduction in the discriminative capacity of FEV1/ht3. Furthermore, even if a classification basing on quartiles of FEV1/Ht3 is statistically logical considering the absence of indications from epidemiological studies, it may occur than this would not be the best method.

The observed frequency of exacerbation refers to the last 3 months period and, thus, cannot be compared to that reported in other studies. for example by Hurst et al.24 Indeed, we based our definition on two SGRQ items that have a time frame of only 3 months; therefore our data are not directly comparable to those reporting yearly exacerbation rates. Finally, the loss of many patients to follow-up may have biased our results on mortality. However, patients followed up and patients lost to follow up had comparable clinical characteristics.

In conclusion, our study indicates that the severity of bronchial obstruction according to GOLD classes may stratify mortality risk better than quartiles of FEV1/Ht3, whereas the second seems to be more suited to stratify clinical outcomes, such as dyspnea, walking speed and quality of life. Concerns about the appropriateness of using externally developed reference values to calculate FEV1PP do not seem to be confirmed, at least for GOLD classification.

Each have substantially contributed to data collection, manuscript drafting and revision. All authors read the final manuscript and approved it for publication.

The authors declare they have no conflict of interest.