To the best of our knowledge, our study is the largest multicenter series to describe characteristics of patients with recurrence and to compare with first episode in 1491 PSP. The main result is that the usual clinical risk factors considered to be linked to occurrence of a first episode of PSP were not found to be associated with recurrence of PSP in this population. The clinical presentation, sociodemographic data, and medical history were similar in both the first PSP and recurrence groups. Only the place of onset was identified in multivariate analysis as a factor associated with recurrence, with pneumothorax more frequently occurring at home in the case of recurrence as compared to first episodes. However, this result must be interpreted with caution in view of the number of missing data (missing data n = 423).

Risk factors for the occurrence of PSP are often considered as risk factors for recurrence [3, 9]. Tobacco smoking is the most important risk factor for occurrence of PSP, with a relative risk that is ninefold higher in women and 22-fold in men for smokers compared with nonsmokers [2]. Smoking is thus often presented as a risk factor for recurrence [3, 15]. Even though recurrence seems to occur earlier in smokers [9], we failed to find a significant relation between smoking and recurrence in our study, as in other studies [9, 13]. Furthermore, smokers could have a significant reduced risk of recurrence compared to nonsmokers [13]. An association between the use of cannabis and pneumothorax has been suggested in previous studies [1, 16–18], but no studies have explored the link between cannabis use and PSP recurrence. In our study, about 5% of patients with PSP declared that they were cannabis users, but no relation was found between use of cannabis and recurrent PSP. Others factors such as gender or age are considered as risk factors for recurrence of PSP. PSP appears to be more frequent in men, with a male/female sex ratio of 3.3 : 1, and PSP appears to occur at a higher age in women than in men [6]. Sadikot et al. found that recurrence was more frequent in women than in men [9]. However, the sex ratio did not differ between those with a first occurrence and those with recurrence in our study, in line with the findings of Uramato et al., who also failed to find a significant gender difference between recurrent and nonrecurrent patients after surgery for PSP [11]. Concerning age, although some studies have revealed that it could be a risk factor [10, 19], it was not found to be associated with recurrence of PSP in our study, as in previous reports [9].

This study presents some limitations. The retrospective design and the potential for bias are the main limits. Indeed, several variables such as smoking habits and cannabis use were self-declared and thus subject to bias. We cannot exclude that patients may have underreported their usage, particularly in the recurrence group. Furthermore, some potential factors associated with recurrence were not studied. Other authors have found that a low body mass index is a risk factor for recurrence [9, 13, 20]. We were unable to study this factor because it was not available in medical files of EDs. Furthermore, the relation between BMI and recurrence of PSP is already well demonstrated in these studies. A further limitation is the fact that we did not record precise information concerning the extent of the pneumothorax. Previous reports have suggested that recurrence within one year was associated with a smaller initial size of pneumothorax [21]. Evaluating the size of pneumothorax is not simple, since no official consensus classification exists. Comparisons using the British Thoracic Society (BTS), American College of Chest Physicians (ACCP), and Belgian Society of Pulmonology (BSP) classifications show concordant estimations in only 47% of patients [22]. Finally, our study was focused on clinical risk factors, whereas lung morphological aspects, genetic factors, and therapeutic management were not studied.

Despite its retrospective nature and limitations, this study offers the opportunity to analyze a large population of patients with PSP. Previous studies of the risk factors mainly included smaller samples. For example, Sadikot et al. included 153 patients with 83 recurrences [9]. More recently, Ouanes-Besbes et al. studied the impact of clinical presentation, treatment, and pulmonary scan findings on recurrence of PSP in 80 patients with 15 recurrences [12], while Olesen et al. included 234 patients [13]. While our study does not provide the same information as a cohort study, the statistical power is nonetheless important. We also conducted a multilevel analysis to take into consideration readmissions during the study period, thus accounting for the correlated configuration of the data. Furthermore, the hospitals involved in our study were spread over a large geographical area, in regions with different climatic conditions. However, this study remains the only one to date to include such a wide diversity of regions, climates, and hospitals. Thus, our recruitment stems from a range of different types of hospitals, with local management, and almost exhaustive inclusions. In light of these considerations, it is unlikely that selection bias could explain our results.

One way for physicians to modify the rate of recurrence seems to be the modalities of treatment. Surgery is clearly recommended for recurrence of PSP [23, 24] but it could be debated for a first episode. Video-assisted thoracoscopic surgery (VATS) represents the surgical treatment of choice [3]. Some authors propose immediate VATS after a first episode of PSP in patient who do not heal after simple aspiration [25, 26], because it could reduce the rate of recurrence. A CT-based lung dystrophy severity score has recently been proposed to predict recurrence and could be useful for clinicians to determine patients who could benefit from early surgery after a first episode [27].

In this cross-sectional study, admittances with first episode had similar characteristics to those with recurrence, and no clinical factor associated with recurrence was identified. The place of onset differed between groups. Only a large prospective follow-up of patients admitted for a first episode of PSP would provide a significantly higher standard of proof concerning risk factors of recurrence.