Since the pioneering work of Hogg and associates (³), the small airways have been recognized as the major site of airflow limitation in COPD, accounting for up to 90% of the total resistance to flow in the lungs of patients with well-established disease. Small airway walls contain neutrophils, macrophages, and both CD4⁺ and CD8⁺ T lymphocytes (but not eosinophils) and their levels progressively increase with disease severity (Figure 2) (⁴).

In concert with these inflammatory changes, the thickness of the airway wall structural components (epithelium, lamina propria, smooth muscle, and adventitia) also increases with disease severity. These findings suggest that it may be possible to track COPD disease progression through serial invasive evaluation of airway histopathology through endobronchial biopsies (assuming that histopathologic changes in the large airways reflect similar changes in the small airways) and transbronchial biopsies that sample the small airways and distal lung. Disease progression has been tracked in this way in asthma, but to date, this has not been performed in COPD (⁵). While evaluation of airway changes may be an accurate surrogate of disease progression in COPD, the invasive techniques required are not practical or feasible beyond small-scale experimental investigation. Similar concerns limit the usefulness of bronchoalveolar lavage. Thus, reliance must be placed on noninvasive pulmonary biomarkers.

Induced sputum is a relatively easy and well-tolerated procedure that can reflect inflammatory processes in the airway. Evidence for the utility of this approach in assessing inflammation has been demonstrated by comparing samples taken from smokers with those from nonsmokers (⁴, ⁶). However, induced sputum also has limitations. For example, (a) it predominantly samples the larger airways and may not reflect inflammation of the small airways; (b) the sputum induction procedure itself can induce neutrophilic inflammation; (c) solubilization of the sputum may interfere with radioimmune assays for the detection of cytokines and chemokines; and (d) little is known concerning its long-term reproducibility or its correlation with COPD severity and progression.

Exhaled air provides another readily accessible and repeatable noninvasive means for monitoring inflammation in COPD. The major exhaled gases that reflect lung inflammation include nitric oxide (NO), carbon monoxide (CO), and volatile hydrocarbons (alkanes, pentanes, ethane), the latter being markers of lipid peroxidation due to oxidative stress (⁷). Exhaled NO levels are commonly elevated in uncontrolled asthma, but are often normal or only slightly increased in COPD, presumably due to the conversion of NO into peroxynitrite and nitrate due to oxidative stress. On the other hand, the alveolar fraction of exhaled NO (obtained by measuring exhaled NO at different expiratory flow rates) appears elevated in COPD (⁸). The relationship between exhaled NO and COPD severity, progression, and acute exacerbations, however, remains to be determined. Exhaled CO (an oxidative breakdown product of heme and thus a marker of oxidative stress) is easily measured but may not be a useful biomarker as it is only slightly elevated in COPD and its measurement is confounded by active and passive smoking (⁹). Exhaled ethane levels are elevated in COPD and correlate with disease severity (¹⁰); however, the current assay methodology is too complex for routine assessment.

Exhaled breath condensate is another noninvasive and simple sampling technique for inflammatory mediators in the lung, including oxidative products, leukotrienes, cytokines, and pH (which reflects tissue acidification due to inflammation) (⁶). Limitations associated with exhaled breath condensate biomarkers include wide variability, due largely to dilution from water vapor during condensation, and low concentrations of the biomarkers that approach the detection limits of the assays. These limitations require correction and validation before exhaled breath condensate can be reliably used to assess disease progression in patients with COPD.

Biomarkers in plasma and serum (interleukin [IL]-6, IL-8, tumor necrosis factor [TNF]-?, fibrinogen, and Creactive protein [CRP]) have been studied for their relationship to disease severity in COPD (¹¹). A recent meta-analysis demonstrated that neither serum CRP nor serum TNF-? levels are statistically significantly different between healthy subjects and patients at different COPD stages (¹¹). Data on the relationships between serum IL-6, IL-8, or fibrinogen and COPD are insufficient and inconclusive.

In summary, biomarkers obtained from noninvasive collection methods offer future promise for assessing disease progression. However, methodological improvements and validation of the relationship between the biomarkers and disease severity are required before these markers can be widely adopted as surrogates of COPD disease progression.

Subtle physiologic changes that may reflect early pathology involving peripheral airways occur relatively early in the course of cigarette smoking (¹²), but may or may not progress to clinically significant physiologic impairment. Abnormalities in airflow (measured by FEV₁/FVC) are considered the physiologic hallmark of COPD by which the disease is defined. It is well known that FEV₁ declines with age in all adults aged 25 years and older; however, the rate of decline is more rapid in the susceptible smoker than in age-matched healthy nonsmokers (Figure 3) (¹³). Just as changes in FEV₁ lag behind inflammatory and tissue changes, the development of symptoms is also delayed in comparison with significant decrements in FEV₁. In fact, dyspnea may not become evident until FEV₁ has declined to 50?60% of predicted normal (¹⁴). Thereafter, symptoms continue to increase at a rate that inversely correlates with the subsequent annual change in FEV₁ (¹⁴).

The rate of FEV₁ decline is not easily modified by interventions. It is therefore unclear whether this index is resistant to change or if the approaches to its modification that have been assessed so far are ineffective. To date, only smoking cessation unequivocally decreases the rate of FEV₁ decline (Figure 4) (¹³, ¹⁵, ¹⁶). Pharmacologic interventions, such as inhaled corticosteroids (ICS), which are the ?gold standard? anti-inflammatory agents for controlling inflammation in asthma, have failed to alter the rate of FEV₁ decline reported in large trials lasting 3 years or more (¹⁵, ¹⁷?¹⁹). Meta-analyses of these trials either confirmed the original negative results (²⁰, ²¹) or identified a modest but statistically significant 7.7 mL per year improvement in the annual rate of FEV₁ decline ? the clinical significance of which is uncertain (²²). Similarly, the antioxidant and mucolytic N-acetylcysteine failed to demonstrate any impact on the rate of FEV₁ decline over 3 years, despite the anticipated value of this agent given that oxidative stress is believed to play an important role in COPD pathogenesis (²³).

Only a few trials appear to have shown that pharmacological intervention alters the rate of FEV₁ decline and, thereby highlight the usefulness of this index in characterizing disease progression. The TOwards a Revolution in COPD Health (TORCH) trial evaluated the ability of pharmacologic interventions (salmeterol/fluticasone combination, fluticasone alone, salmeterol alone, or placebo) to alter all-cause mortality in approximately 6,000 patients with moderate to very severe COPD over the course of 3 years (²⁴). The decline in the rate of FEV₁ was not a primary or secondary endpoint and centralized spirometry was not employed; however, post-hoc analysis suggests that the rate of decrease in postbronchodilator FEV₁ between 6 months and 3 years was slightly but significantly less with salmeterol/fluticasone (39 mL per year; p < 0.001) and with either fluticasone alone or salmeterol alone (both 42 mL per year; p = 0.003), compared with placebo (55 mL per year) (Figure 5), with no differences among active treatment arms (²⁵). Similarly, post-hoc evaluation of FEV₁ data from a 1-year placebo-controlled study of tiotropium in COPD revealed a rate of decline in predose FEV₁ with tiotropium of 12.4 mL per year (n = 518) compared with a 58 mL per year decline with placebo (n = 328) (p = 0.005) (Figure 6) (²⁶). A lower and nonsignificant difference in the slope of decline in postbronchodilator FEV₁ was also noted.

Based on these data, a large-scale, multinational trial (Understanding Potential Long-term Impacts on Function with Tiotropium [UPLIFT]) of nearly 6,000 patients with moderate-to-severe COPD (FEV₁ <70% of predicted normal after maximal bronchodilation) was conducted to determine whether the long-acting antimuscarinic tiotropium unequivocally decreases the rate of both pre- and postbronchodilator FEV₁ decline in COPD over a 4-year period in the context of freely prescribed medications for the treatment of COPD except inhaled anticholinergics (²⁷). The study also examined whether treatment-related reduction in the rate of FEV₁ decline is associated with concomitant improvements in patient-centered outcomes (health status, exacerbations, hospitalizations for COPD), as well as all-cause mortality. The findings indicated that although tiotropium was associated with sustained improvements in lung function, quality of life, and exacerbations during a 4-year period, it did not significantly reduce the rate of decline in FEV₁ (²⁷). These negative findings need to be interpreted in the light of the fact that over 70% of UPLIFT participants used an inhaled corticosteroid and/or long-acting beta-agonist during the course of the trial (²⁷), thus making it more difficult to demonstrate an impact of tiotropium on top of any potential effect of these concomitant therapeutic agents on decline in FEV₁.

One limitation that affects all of the long-term studies, including TORCH and UPLIFT, is that many patients fail to complete the studies, with significantly more patients withdrawing from the placebo arms than the active treatment arms (²⁵, ²⁷). It has been observed that the patients who withdraw are often those who are experiencing more rapid deterioration in lung function, thereby resulting in the remaining patients essentially representing ?healthy survivors? (²⁵, ²⁷). The higher drop-out rate for individuals in the placebo arm with poorer lung function potentially minimizes the differences in lung function decline observed between placebo and active interventions (²⁵, ²⁷).

Death is the ultimate consequence of disease progression and thus reduced mortality and improved survival describe an impact on disease progression. Numerous factors are reported to influence mortality in COPD including FEV₁ (⁵⁸), dyspnea (⁵⁹), disease duration (⁶⁰), carbon dioxide and oxygen arterial tensions (⁵⁶), cardiac status (⁵⁴), body mass index (⁶¹), serum albumin level (⁶²), functional status (⁶³), exercise limitation (⁶³, ⁶⁴) and comorbidities (⁵⁵). The correlations between these cofactors and long-term mortality are variable.

Interventions that impact on mortality most probably do so via improvements in aspects of lung function and patient-centered outcomes, although there may be other, as yet unknown ways in which interventions affect this endpoint (Figure 10). Smoking cessation has beneficial effects on subsequent mortality, as demonstrated by an 18% reduction in all-cause mortality after 14.5 years (⁶⁵). Long-term home oxygen therapy has been shown to reduce mortality in patients with persistent hypoxemia, as has lung volume reduction therapy in highly selected patients with emphysema (⁶⁶?⁶⁸). In addition, a recent meta-analysis showed that pulmonary rehabilitation after a COPD exacerbation reduced the risk of mortality (pooled relative risk 0.45 [95% confidence interval (CI) 0.22?0.91]) (⁶⁹). Further research is required to define the long-term effects of these interventions.

The effect of a pharmacologic intervention on mortality was recently investigated in the TORCH study. In this study, patients received either salmeterol/fluticasone in combination, salmeterol alone, fluticasone alone, or placebo, for 3 years (²⁴). Of 6,112 patients in the efficacy population, 875 died over the 3-year study. The hazard ratio for all-cause mortality in the combination therapy group compared with the placebo group was 0.825 (95% CI 0.681?1.002; p = 0.052), corresponding to a reduction in the absolute risk of death of 2.6% (17.5% relative reduction), which approached statistical significance. Further factorial analysis indicated that the effect on mortality of the combination therapy appeared to be entirely due to salmeterol, and that the effect was highly significant (p = 0.004) (⁷⁰).

This underscores the important role that airflow limitation plays in disease progression and, consequently, the need for it to be managed. Despite the lack of a statistically significant reduction in mortality, combination therapy did result in significantly fewer exacerbations and improved HRQoL and lung function compared with placebo. Clinical findings from the study, however, suggest that monotherapy with corticosteroids should not be recommended, but that LABAs used alone or in combination may provide benefit. It is clear from the findings of the TORCH study that further investigation is required into the effect on mortality of pharmacotherapy in COPD.

While UPLIFT focused primarily on the rate of decline in FEV₁ as a primary endpoint, mortality was a key prespecified secondary endpoint. As in TORCH, vital status was ascertained in nearly all patients over the protocol-defined treatment period, permitting an intention-to-treat analysis of the impact of the study medication on mortality. The hazard ratio for all-cause mortality over the 1,440-day protocol-defined duration of the study (vital status known in 95% of all randomized subjects) was 0.87 (95% CI 0.76?0.99; p = 0.034), while that over this same time period plus 30 days (1470 days), as prespecified in the analysis plan, (vital status known in only 75% of all subjects over the latter time period) was 0.89 (95% CI 0.79?1.02; p = 0.086) (²⁷).

Thus, both of the recent long-term trials of pharmacotherapy in COPD (TORCH and UPLIFT) revealed an impact on mortality that came tantalizingly close to achieving statistical significance. The mechanism(s) of this effect need to be further explored but could involve several factors, including reductions in exacerbations and in respiratory failure, as well as improvements in ventilatory mechanics, including a reduction in hyperinflation, that may have indirect cardiac benefits through, for example, a decrease in cardiac afterload.

Since mortality is an end result of COPD, it can be a useful endpoint to define an impact on disease progression within a population receiving a study medication. However, mortality is not suitable for assessment of disease progression in individual patients. It is further complicated by comorbid conditions associated with COPD that can involve organ system dysfunction sufficient to contribute to mortality independently or additionally, but not exclusively due to COPD. Therefore, mortality is appropriate to define the end result of disease progression; however, it is not appropriate for describing disease progression in individual patients or in clinical practice.