One of the difficulties in performing neonatal research is how to determine the most clinically appropriate outcomes. Bronchopulmonary dysplasia is often used as a dichotomous outcome variable, partly to facilitate the design of research, including combining it with other competing outcomes, such as death. If you die before 36 weeks you don’t have BPD, so that death would be counted as a good pulmonary outcome unless the 2 are combined in some way. Death is a truly dichotomous outcome (as long as zombies don’t really exist) so using ‘death or BPD’ as an outcome makes some sense as long as it is understood that this doesn’t mean they are considered equivalent!
The big downside of this is that lung injury is not a yes or no phenomenon. Lung injury in the preterm is a continuum, and deciding that a good outcome is stopping oxygen 1 day before the baby reaches 36 weeks, and a bad outcome is stopping oxygen 1 day later is patently too simplistic. Defining different severities of BPD is useful clinically, but needs also to be considered in study design, and is still too restrictive, with just 3 categories in the Jobe and Bancalari definitions.
Longer term follow up with evaluation of the functional effects of lung injury needs to be part of trials which address respiratory interventions.
A new report of respiratory outcomes of infants in the SUPPORT trial (Stevens TP, Finer NN, Carlo WA, Szilagyi PG, Phelps DL, Walsh MC, Gantz MG, Laptook AR, Yoder BA, Faix RG et al: Respiratory outcomes of the surfactant positive pressure and oximetry randomized trial (SUPPORT). The Journal of pediatrics) has 36 authors reporting on behalf of the trial study group. The collaborating centers followed up babies at 6 monthly intervals for 2 years, data were collected by asking the parents how the infant was doing, in terms of wheezing episodes, cough, re-hospitalization and so on, using a validated questionnaire. They found that :
patients in the CPAP group had lower rates of several important respiratory morbidities at 18-22 months CA, including respiratory illnesses diagnosed by a doctor, treatment with oxygen or diuretics at home, and a trend toward a lower rate of overnight hospitalization for breathing problems.
They also saw a lot of respiratory morbidity among the SUPPORT babies who did not have a diagnosis of BPD (including 7% of them who had home oxygen, which surprised me).
This is consistent with some of the data from the COIN trial which also seemed to show that clinically important longer term pulmonary outcomes in the babies who had CPAP and attempted avoidance of intubation were improved.
Neonatal Research 
This may also be consistent with the findings in 319 survivors aged 11-14 years by Zivanovic and Peacock et al that HFOV improved small airway function, as indicated by the primary outcome of forced expiratory flow at 75% of expired vital capacity and other measures, (NEJM 2014; 2021-30). The authors speculated that, because volumes in HFOV are lower than in mechanical ventilation, this may be the mechanism for a protective effect of HFOV. CPAP uses no (or maybe v. small volumes) compared with mechanical ventilation, so perhaps this provides a similar protective mechanism.
oops – should be NEJM 2014; 370: 2021-30
Further complicating this issue is the fact that not just the severity but the phenotype of BPD is quite variable – some infants have more of a pulmonary hypertension phenotype, others have more tracheobronchomalacia, others have marked wheezing (small airways), and a large proportion have patchy parenchymal lung disease. This reminds me of the situation many decades earlier when all leukemias were lumped together, and progress was achieved only after careful phenotyping identified different sub-categories each of which had different optimal therapies.