Purpose The best criterion for diagnosing airway obstruction in COPD, fixed ratio (FR: FEV1/FVC<0.7) or lower limit of normality (LLN), remains controversial. We compared the long-term evolution of COPD patients according to the initial obstruction criteria. Patients and Methods Between 2005 and 2008, we evaluated 1728 subjects over 45 years of age with smoking history, pertaining to a primary care center. A total of 424 patients were obstructive by FR, after a bronchodilator test. Of those, 289 patients met obstruction criteria for both FR and LLN and were considered concordant patients (FR+LLN+), while 135 patients were obstructive by FR but non-obstructive by LLN and were defined as discordant patients (FR+LLN-). Results Forty-eight patients (11.3%) were lost in follow-up, and 158 died (37.3%). After a median time of 120.4 months (IQR 25–75%: 110.2–128.8), 215 patients were spirometrically reevaluated. The annualized loss of FEV1/FVC was greater in discordant (FR+LLN-) patients [0.54 (0.8) vs 0.82 (0.7); p = 0.008], while 81% became concordant (FR+LLN+) during the follow-up. Hospitalization for COPD exacerbations was more frequent in concordant (FR+LLN+) patients (1.57±3.51 vs 0.77±2.29; p = 0.002). Adjusting for age, concordant (FR+LLN+) patients had greater COPD mortality (HR: 2.97; CI 95%: 1.27–7.3; p = 0.02). Conclusion LLN seems to be less useful for COPD diagnosis in primary care. Discordant (FR+LLN-) patients lost more FEV1/FVC during their evolution and tended to become concordant. LLN predicted COPD hospitalizations and mortality more poorly.