RESULTS

Section:

ChooseTop of pageAbstractMETHODSRESULTS <<DISCUSSIONReferencesCITING ARTICLES

The entire cohort was characterized by elderly patients with severe airflow obstruction, moderate breathlessness, and increased mBODE (Table 1). At baseline, there were no differences in mBODE between the medical and surgical arms. Tables 2 and 3 give insight into the magnitude of mBODE differences between the two treatment arms over time and formally compare the two arms based on mixed model results that adjust for attrition of repeated measures. As shown in Table 3, surviving surgically treated (LVRS) patients presented negative slopes at 6, 12, and 24 months postrandomization on average, whereas medically treated patients had significantly higher (P value always less than 0.001) positive mBODE slopes at each time mark postrandomization. As such, LVRS patients showed a consistent pattern of mBODE decrease (improvement), whereas medically treated patients exhibited significant increases (deterioration) in mBODE. The distribution of change in mBODE among survivors at 6 (Figure 1), 12 (Figure E1A of the online supplement), and 24 months (Figure E1B) is illustrated by plot. It is evident that the entire cohort exhibits a wide range of scores at each landmark time, with negative changes in mBODE (improvements) more frequent in surgically treated patients. Significant missing data were seen, particularly in medically treated patients. Among those who survived to 6, 12, and 24 months after randomization (Table E1), subtle differences were observed in component variables of the mBODE at baseline in both the medical and surgical groups, making formal treatment comparisons among survivors over time inappropriate. When patients were grouped by survival duration, there was a clear gradient in baseline FEV₁ and baseline dyspnea, with those who survived 24 months having higher FEV₁ and lower dyspnea at baseline than those who did not survive as long.

Figure 1. The percentage of patients (x axis) experiencing changes in mBODE (modified Body mass index, airflow Obstruction, Dyspnea, and Exercise capacity) (y axis) 6 months after randomization to medical therapy or bilateral lung volume reduction surgery (LVRS). Patients who had died or were not able to provide data to calculate mBODE at each time point are also illustrated.

Download Figure | Download Powerpoint

[More] [Minimize]

The influence of changes in mBODE on subsequent mortality was assessed by computing individual trajectories on the basis of each patient's available mBODE patient history available at 6 (Figure 2), 12 (Figure E2), and 24 months (Figure E2) postrandomization. The mortality characteristics of the entire cohort (Figure 2A), surgically treated patients (Figure 2B), and medically treated patients (Figure 2C) 6 months after randomization are illustrated in Figure 2 by prior change in mBODE. Similar curves for 12 and 24 months after randomization are illustrated in Figure E2. It is evident that there are distinct groups with significantly different survival experiences in the entire cohort, in surgically treated patients, and in medically treated patients. Figure 3 illustrates how a decrease in mBODE of greater than 1 point at 6 months postrandomization in the entire cohort was associated with a significant decrease in subsequent mortality (hazard ratio [HR], 0.57; 95% confidence interval [CI], 0.41–0.78; P < 0.001), whereas an increase in mBODE of more than 1 point was associated with a significantly increased mortality (HR, 2.35; 95% CI, 1.71–3.23; P < 0.001). Patients with missing data at 6 months experienced an intermediate increase in mortality (HR, 1.45; 95% CI, 1.10–1.92; P = 0.01). Similar results were noted in the entire cohort at 12 and 24 months postrandomization. In the surgically treated cohort, a decrease in mBODE of more than 1 point at 6 months was suggestive of improved subsequent survival (HR, 0.72; 95% CI, 0.48–1.08; P = 0.110), with a similar trend at 12 and 24 months postrandomization becoming statistically significant. Hazards associated with an increase in mBODE of greater than 1 point or missing data were somewhat higher in the surgical arm compared with the medical arm; otherwise, hazards in both treatment arms were ordered very similarly to the overall cohort and were highly statistically significant. An exploratory analysis confirmed that a change of 1 point in mBODE better distributed survival curves than other thresholds, including a 2-point change in mBODE (data not shown).

Figure 2. Kaplan-Meier survival curves (starting at 6 months after randomization [F06]) for the entire cohort (A), surgically treated patients (lung volume reduction surgery [LVRS] arm) (B), and medically treated patients (C) in patients who survived and were able to provide data to calculate mBODE (modified Body mass index, airflow Obstruction, Dyspnea, and Exercise capacity) 6 (F06) months after randomization. Survival curves are segregated by groups who experienced a decrease in mBODE > 1 (blue), a rise in mBODE > 1 (red), no change in mBODE (−1 to 1 point change) (green), or missing data (black) from baseline to the starting time point. (A) Log rank = 46.9; degrees of freedom = 3; P value < 0.01. (B) Log rank = 53; degrees of freedom = 3; P value < 0.01. (C) Log rank = 7.58; degrees of freedom = 3; P value ≤ 0.06.

Download Figure (Part 1) | Download Figure (Part 2) | Download Figure (Part 3) | Download Powerpoint

[More] [Minimize]

Figure 3. Multivariate Cox survival models comparing the predictive value of change in mBODE (modified Body mass index, airflow Obstruction, Dyspnea, and Exercise capacity) at 6, 12, and 24 months postrandomization, adjusted for sex, ethnicity, baseline age, and baseline mBODE. LVRS = lung volume reduction surgery.

Download Figure | Download Powerpoint

[More] [Minimize]

Figure 4 illustrates event-free rates associated with the composite outcome of death or an increase in mBODE of more than 1 point. The median time to this composite event was significantly shorter in medically treated patients (3.65 yr; 95% CI, 3.21–4.37 yr) compared with LVRS-treated patients (5.68 yr; 95% CI, 4.93 yr–not available). Table 4 demonstrates indices of concordance for mBODE, 6MWT, and UCSD SOBQ. The index of concordance allows us to quantify the predictive ability of a survival model. As can be seen from the table, mBODE exhibits indices of concordance consistently higher than those obtained for FEV₁, 6MWT distance, and the UCSD SOBQ alone, indicating that the predictive ability of the mBODE for survival is better than any of the individual components. On the other hand, it appears that change in UCSD SOBQ likely accounted for much of the mBODE's predictive ability.

Figure 4. Kaplan-Meier curves comparing a composite outcome of survival or increase in mBODE (modified Body mass index, airflow Obstruction, Dyspnea, and Exercise capacity) greater than 1 point between medical and surgical treatment groups. The red line reflects lung volume reduction surgery (LVRS)–treated patients, whereas the blue line reflects medically treated patients.

Download Figure | Download Powerpoint

[More] [Minimize]

DISCUSSION

Section:

ChooseTop of pageAbstractMETHODSRESULTSDISCUSSION <<ReferencesCITING ARTICLES

Reducing mortality has become increasingly feasible in COPD. Therapeutic trials designed to demonstrate improved survival require large study design with prolonged periods of follow-up. As such, identifying reliable surrogates of survival in COPD has become an important concept. Multidimensional indices, such as the BODE index, evaluated at baseline have been shown to exhibit greater predictive values than the individual components of the index. Recently, three uncontrolled studies have suggested that the BODE index changes after therapeutic intervention and that this change appears to relate to subsequent mortality (5, 6). These studies have been limited by lack of control groups. The current analysis describes longitudinal changes in a highly characterized, large cohort of patients with severe emphysema, half of whom underwent a therapeutic intervention that resulted in significant improvement. As such, we report the following: (1) the mBODE is responsive to longitudinal changes with greater improvement in surgically compared with medically treated patients; (2) a decrease in mBODE of more than 1 point was predictive of lesser mortality in the entire cohort and, particularly, the surgically treated cohort; (3) an increase of mBODE of more than 1 point was predictive of greater mortality, a finding more consistently seen in the medically treated patients; (4) analysis of concordance suggests that the mBODE is more predictive of survival than the changes in its individual components; and (5) short-term changes in a composite endpoint (mBODE increase of >1 point or death) may serve as a valuable surrogate of mortality in future COPD studies.

We confirm that the mBODE changes over 6 to 24 months in patients with severe COPD. Interestingly, the entire cohort illustrated similar numbers of patients who exhibited increases or decreases in mBODE. Importantly, medically treated patients appeared to experience a lesser change in mBODE, with most of the change seen in increasing values, suggesting a worsening score in this multidimensional index. A separate report noted similar, subtle change in BODE in a cohort of patients, some of whom were treated with pulmonary rehabilitation (5). Although the influence of additional pulmonary rehabilitation in the current cohort cannot be directly assessed, our work expands previous data by providing insight into the natural history of longitudinal change in mBODE in a large cohort of patients with severe emphysema treated with maximal medical therapy. Furthermore, these data allow the calculation of the expected natural history of the mBODE in such patients over a 2-year time frame. These data can be used to define expected changes in mBODE in future therapeutic trials that wish to use this composite index as a treatment outcome.

Our data provide a contrast in that surgically treated patients exhibit a wider distribution of changes in scores, with a larger number of patients experiencing decreasing, or improving, scores. These data expand on the original observations of the NETT that confirmed functional improvements in surgically compared with medically treated patients with severe emphysema (9). Two groups have recently presented results of short-term (3 mo) and intermediate-term (12–24 mo) changes in BODE in a small group of patients with emphysema treated with LVRS that document a similar distribution of change, with the majority of patients exhibiting decreases in scores (6–8). Our data extend these observations by illustrating the natural history of change in mBODE over a 2-year period, and are strengthened by having a similar medically treated patient group for comparison. Direct survival comparison of surgically versus medically treated patients by mBODE change is limited by differential mortality over the period of measurements. Nevertheless, the totality of these data provides insight into the potential change in mBODE in patients with severe COPD treated with a therapy that provides marked functional improvements.

Our data provide important insight into how short- and intermediate-term changes in a composite index can be used to predict mortality in a large group of medically and surgically treated patients with severe COPD. In both medically and surgically treated patients, an increase in mBODE was associated with an increased risk of mortality in subsequent follow-up. This was noted as early as 6 months and out to 2 years after randomization to intervention. Furthermore, a greater than 1 point increase in mBODE was associated with a 1.9- to 2.1-fold increase in mortality in medically treated patients and a 2.5- to 3.6-fold increase in surgically treated patients. We also noted that a decrease in mBODE of greater than 1 point after 12 months was associated with improved subsequent mortality only in surgically treated patients. These data expand on results from other groups that suggest that a decrease in BODE is associated with decreased subsequent mortality in surgically treated patients (6) by illustrating that changes in mBODE are predictive of subsequent mortality from as short as 6 months of follow-up to 24 months of follow-up. An important finding of our study is the ability to describe the time to death or to an increase in mBODE in medically and surgically treated patients. These data can be used to construct clinical trials of therapeutic interventions that may improve mortality with a smaller sample size over a period as short as 6 months.

We also provide insight into the operating characteristics of a composite index in contrast to some of the individual components. An analysis of concordance demonstrates that the composite index was better able to predict subsequent mortality in the entire cohort as well as in the medically and surgically treated patients. A higher index of concordance was evident for the composite mBODE index at 6, 12, and 24 months after intervention. Further examination suggests that the majority of change in mBODE was contributed by improved dyspnea. Our data extend the findings of others that have suggested that change in 6MWD over 1 year predicts survival (13) by contrasting change in exercise capacity with longitudinal changes in dyspnea, FEV₁, and a composite index incorporating all of these factors.

Limitations of this study come from its highly selective sample population. All subjects were enrolled in NETT, a randomized trial of LVRS for patients with a predominantly emphysematous phenotype. As such, selection bias was imposed by requirements for a highly selected subset of patients with severe COPD. In addition, the mBODE was a modification of the original multidimensional index. On the other hand, the operating characteristic of the mBODE was remarkably similar with regard to prognostic ability (4) to the original index (3). In addition, the predictive ability of a greater than 1-unit change in mBODE was similar to that noted for changes in a broader spectrum of patients with COPD undergoing rehabilitation in a smaller, uncontrolled study (5). Finally, although no patients were lost to follow-up, we are missing data whenever a patient was unwilling or unable to complete any one component of the mBODE. This was particularly evident in medically treated patients. It is difficult to know if this led to any systematic bias in our reported results because the predictive ability of change in mBODE is conditional on the ability of patients to contribute data at the respective time points. Interestingly, we noted that those with missing data experienced a higher mortality than those with no change in mBODE, although those with rising mBODE generally appeared to exhibit even higher mortality.

In summary, we demonstrate that a composite index (mBODE) demonstrates short- and intermediate-term responsiveness to intervention in severe COPD. This was more evident in surgically treated than in medically treated patients. Furthermore, changes in mBODE at 6, 12, and 24 months of follow-up were predictive of subsequent mortality. An increase in mBODE of greater than 1 point proved predictive of mortality. Importantly, the changes in the individual parameters were not as predictive of subsequent survival as individual components. These changes provide support that a short- and intermediate-term change in a multidimensional composite index may be used as a surrogate for survival in clinical trials.