 Research
 Open Access
 Published:
Probability of sporadic lymphangioleiomyomatosis in women presenting with spontaneous pneumothorax
Orphanet Journal of Rare Diseases volume 18, Article number: 180 (2023)
Abstract
Background
Sporadic lymphangioleiomyomatosis (SLAM) is a rare lowgrade neoplasm of young women characterized by multiple pulmonary cysts leading to progressive dyspnea and recurrent spontaneous pneumothorax (SP). The diagnosis of SLAM may be delayed by several years. To reduce this delay, chest computed tomography (CT) screening has been proposed to uncover cystic lung disease in women presenting with SP. However, the probability to discover SLAM in this population has not been determined precisely. The aim of this study was to calculate the probability of finding SLAM in women presenting with (a) SP, and (b) apparent primary SP (PSP) as first manifestation of SLAM.
Methods
Calculations were made by applying the Bayes theorem to published epidemiological data on SLAM, SP and PSP. Each term of the Bayes equation was determined by metaanalysis, and included: (1) the prevalence of SLAM in the general female population, (2) the incidence rate of SP and PSP in the general female population, and (3) the incidence rate of SP and apparent PSP in women with SLAM.
Results
The prevalence of SLAM in the general female population was 3.03 per million (95% confidence interval 2.48, 3.62). The incidence rate of SP in the general female population was 9.54 (8.15, 11.17) per 100,000 personyears (py). The incidence rate of SP in women with SLAM was 0.13 (0.08, 0.20). By combining these data in the Bayes theorem, the probability of finding SLAM in women presenting with SP was 0.0036 (0.0025, 0.0051). For PSP, the incidence rate in the general female population was 2.70 (1.95, 3.74) per 100,000 py. The incidence rate of apparent PSP in women with SLAM was 0.041 (0.030, 0.055). With the Bayes theorem, the probability of finding SLAM in women presenting with apparent PSP as first disease manifestation was 0.0030 (0.0020, 0.0046). The number of CT scans to perform in women to find one case of SLAM was 279 for SP and 331 for PSP.
Conclusion
The probability of discovering SLAM at chest CT in women presenting with apparent PSP as first disease manifestation was low (0.3%). Recommending chest CT screening in this population should be reconsidered.
Introduction
Pulmonary lymphangioleiomyomatosis (LAM) is a rare lowgrade neoplasm which exclusively affects women of reproductive age. It is characterized by infiltration of the lungs by neoplastic smooth musclelike cells (LAM cells) leading to the development of multiple pulmonary cysts, which progressively replace the lung parenchyma [1, 2] and may lead to respiratory failure [3]. The most common initial manifestations include progressive dyspnea and multiple recurrent pneumothorax resulting from spontaneous rupture of pulmonary cysts [4]. The disease may either occur sporadically (SLAM) or in patients with tuberous sclerosis complex (TSCLAM), a rare genetic disorder [2].
Due to rarity of the disease, the diagnosis of SLAM is often delayed by several years after the first symptoms [5]. As spontaneous pneumothorax (SP) is a common inaugural presentation of the disease, an earlier diagnosis could theoretically be achieved by carrying out a chest computed tomography (CT) scan after a first episode of SP to reveal multiple pulmonary cysts [5, 6]. One study suggested that a screening CTscan in women who present with an inaugural (socalled “sentinel”) SP would allow earlier detection of patients with LAM with a favorable cost–benefit ratio [5]. In this study, it was estimated that 5–30% of nonsmoking women aged 25–54 presenting with an apparently primary SP (PSP), i.e. SP occurring in the absence of known lung disease, may actually have LAM as a hidden underlying cause [5]. However, this estimate has never been assessed precisely. The probability of having LAM in a woman with apparent PSP is therefore undetermined. The goals of this study were to determine the probability of having SLAM (a) in women presenting with SP (both primary and secondary), and (b) in women presenting with apparent PSP. Calculations were made through the Bayes theorem of conditional probability, which allows to determine the probability of an event based on prior knowledge of conditions related to this event. Each term of the Bayes equation was determined through metaanalyses of published studies, following a method previously used by our group to calculate the prevalence of Birt–Hogg–Dubé syndrome in the general population [7]. Our study was restricted to SLAM, because the issue of apparent PSP is less relevant in TSCLAM. Indeed, TSC frequently presents in infancy or childhood with neurological, mental or cutaneous manifestations leading to the diagnosis of this genetic disorder, and LAM is systematically looked for by chest CTscan in women once TSC is diagnosed.
Methods
Overview
The classical definitions of spontaneous pneumothorax (SP), primary SP (PSP) and secondary SP (SSP) were used in this study [8,9,10]. SP was defined as a pneumothorax occurring in the absence of precipitating external event such as trauma or iatrogenic cause. SSP was defined as SP occurring in the context of an underlying lung disease that predisposes to SP such as emphysema, fibrosis, LAM, or other cystic lung diseases. PSP was defined as SP occurring in the absence of underlying lung disease as a predisposing factor, i.e. without detectable cause [8, 10]. Apparent PSP was defined as a SP occurring in the absence of known underlying lung disease, although a hidden cause is present but is undiagnosed at the time of pneumothorax occurrence, which is therefore initially considered as PSP [11, 12]. In the present study, apparent PSP in patients with SLAM was defined as the first manifestation of SLAM, at a time when the disease was already present but not diagnosed.
In a first set of data analyses, we calculated the probability of having SLAM among women presenting with SP, i.e. both PSP and SSP, including in women with diagnosed SLAM as known cause of SSP. All terms of the Bayes equation were determined by metaanalyses of published studies. They included: (1) the prevalence of SLAM in the general female population, (2) the incidence of SP in the general female population, and (3) the incidence of SP in SLAM.
In a second set of data analyses, we calculated the probability of having SLAM among women presenting with apparent PSP, i.e. in women with “sentinel” pneumothorax as inaugural manifestation of SLAM in whom the disease was not yet diagnosed, and who could therefore benefit from a screening chest CTscan to reveal multiple lung cysts. All terms of the Bayes equation were determined by metaanalyses of published studies. They included: (1) the prevalence of SLAM in the general female population, (2) the incidence of PSP in the general female population, and (3) the incidence of apparent PSP in SLAM.
Literature search
A literature search was performed in November 2021 in the PubMed, Embase, Web of Science and Cochrane Library electronic databases, and was updated in April 2023. The search was limited to fulltext journal articles in English and French. Articles whose primary and secondary outcomes met the subjects of interest were selected. All articles were then reviewed to identify other studies of interest in the reference list.
To assess the incidences of SP and PSP in the general population, a search was performed with the Medical Subject Heading (MeSH) keyword “Pneumothorax/epidemiology.” To assess the probability of having SLAM among patients with SP and apparent PSP, and the probability of experiencing SP and apparent PSP in SLAM, a search was performed with the keywords “pneumothorax” and “lymphangioleiomyomatosis” combined with the Boolean operator “AND”. To assess the prevalence of SLAM in the general female population, a search was performed with the keywords “lymphangioleiomyomatosis” AND (“prevalence” OR “epidemiology”).
All search strategies were conducted and reported according to the PRISMA 2020 statement [13].
Statistics
To determine the probability of SLAM in women presenting with SP (both primary and secondary), the Bayes formula was written as follows:
where \(P({\text{S  LAMSP}})\) is the probability of a woman presenting with SP to be affected by SLAM. In the numerator of the Bayes formula, \(P({\text{S  LAM}})\) is the prevalence of SLAM in the general female population, and \(P({\text{SPS  LAM}})\) the prevalence of an SP episode in individuals suffering from SLAM. In the denominator, \(P({\text{SP}})\) is the prevalence of an SP event in the general female population.
As the two prevalences, \(P({\text{SPS  LAM}})\) and \(P({\text{SP}})\), are not directly measurable, we estimated them using the following formulas [14, 15]:
where \(IR_{{\text{SPS  LAM}}}\) is the yearly incidence rate of SP in the SLAM population and \(IR_{{{\text{SP}}}}\) the yearly incidence rate of SP in the general female population, \(\overline{D}_{{\text{SPS  LAM}}}\) is the average duration of an SP episode in the SLAM population, and \(\overline{D}_{{{\text{SP}}}}\) the average duration of an SP episode in the general female population.
These formulas are valid in a steady state setting, i.e. when the total population of affected and unaffected individuals remains constant over time, and provide good approximations when the two prevalences \(P({\text{SPS  LAM}})\) and \(P({\text{SP}})\) are small.
Assuming that the duration of an SP episode is similar in the SLAM population and in the general population, i.e. \(\overline{D}_{{\text{SPS  LAM}}} = \overline{D}_{{{\text{SP}}}} = \overline{D}\), one may substitute these quantities in the Bayes formula and get:
The value of \(\overline{D }\) was based on a recently published randomized trial on the treatment of PSP, which showed that the median time of recovery for a PSP treated conservatively was 30 days, whereas it was 16 days with interventional treatment [16].
As the incidences rates (IR) were not always reported, we additionally used the following relationship between the cumulative incidence (CI) and the incidence rate [17]:
where E is the number of SP events and PT the persontime product in personyears of followup. When the latter was not reported, it was computed by multiplying the number N of individuals at risk at the beginning of the followup period by \(\overline{T}\) the average followup duration.
When the cumulative incidence was reported (instead of the incidence rate), CI = n/N, where n is the number of individuals experiencing at least one SP event (i.e. one or several SP episodes), the number of SP events E was computed by multiplying the average number \(\overline{E}\) of SP episodes per individual by the number n of individuals experiencing at least one SP event. In addition, when the average followup duration \(\overline{T}\) was not reported, given the small number E of events in comparison to the number N of individuals, the persontime product \(PT\) was simply computed by multiplying the number of individuals by the duration of the followup period (the justification comes from this formula \(\overline{T} = (N  E) \cdot T/N + E/N \cdot T/2 \cong T\)). When the median observation time was reported (along with the sample size and interquartile range or range) instead of the mean, we used the Hozo et al. formula to compute the mean [18].
As the three components in the Bayes formula were provided by different studies, a separate metaanalysis for each component was conducted. The variance of \(IR\) was computed based on the Poisson distribution, and the logtransformation and delta method were applied to compute a 95% confidence interval (95%CI). For the prevalence of SP, the FreemanTuckey double arcsine transformation [19] was used to ensure confidence intervals covering the appropriate [0–1] support.
Also, as all studies on SP and PSP epidemiology published before July 2000 were much smaller and had smaller IRs than those published after July 2000, a randomeffects subgroup metaanalysis was carried out with the first subgroup defined by studies published before July 2000 and the second by those published after July 2000 [20]. The same approach was used for the metaanalysis of the prevalences \(P(SP)\), as they were computed based on the IRs. Finally, the pooled effect sizes estimated in each stratum (defined by publication date < July 2000, > July 2000) were used to compute the probability \(P({\text{S  LAMSP}})\) of a woman presenting with SP to be affected by SLAM in each stratum based on Bayes formula. The multivariate delta method was used to compute the variance estimate of the logit transform of \(P({\text{LAM}}{\text{SP}})\). As SP in SLAM is a relapsing phenomenon, relapses of SP, both in SLAM and in the general population, where taken into account for the calculation of SP incidence. If relapses were not counted in the original publication, a correction factor was applied based on a recent metaanalysis of the relapse rate in PSP [21]. To specifically determine the relapse rate in women, we performed a metaanalysis of the articles used in this study.
As a sensitivity analysis, we repeated all calculations with an arbitrary duration of SP in SLAM of 40 days instead of 30 days.
Once the probability of SLAM in women presenting with SP was determined, the number needed to treat (NNT), i.e. the number of chest CTscans to perform among women with SP to detect one case of SLAM was calculated as follows:
where \(P({\text{S  LAMSP}})\) was computed using data published after July 2000.
To determine the probability of SLAM in women presenting with apparent PSP, the Bayes formula was written as follows:
The same method as described above was applied, by using data on the incidence and prevalence of PSP instead of SP in the general population, and by taking into account only inaugural episodes of SP as the first disease manifestation in the SLAM population, at a time when the diagnosis of SLAM was not yet established. Relapses of PSP were deliberately not included in these calculations.
Results
Probability of SLAM in women with SP
Prevalence of SLAM in the general female population
The literature search identified 234 articles. Twentytwo original articles containing data on the prevalence of LAM were retrieved. No additional article was found after manual review. Additional file 1: Fig. S1 shows the flow diagram depicting the search strategy. Seventeen articles were excluded because the population number was missing or the TSCLAM cases were mixed with the SLAM cases. Thus, 5 original studies were kept for metaanalysis [4, 22,23,24,25]. Their characteristics are shown in Table 1. From reference [4] to which we have contributed with data from Switzerland, we used our own data to separate TSCLAM and SLAM, as this stratification was not available for other countries. By metaanalysis, the overall prevalence of SLAM in women was 3.03 (2.48, 3.62) per million (Fig. 1).
Incidence and prevalence of SP in the general female population
The Pubmed search retrieved 1046 articles. A total of 35 original articles reporting SP incidence in the general population were retrieved. No additional article was found after manual review. Additional file 1: Fig. S2 shows the flow diagram depicting the search strategy. One paper was rejected because the number of SP could not be related to population size [26]. Sixteen articles were excluded because population size and/or gender proportion were not given. Four other articles were rejected because they focused only on PSP and not on SP. Fourteen original studies were kept for metaanalysis [27,28,29,30,31,32,33,34,35,36,37,38,40]. Their main characteristics are shown in Table 2.
By metaanalysis, the overall incidence rate of SP in the general female population was 9.54 (8.15, 11.17) per 100,000 py. With time period stratification, the incidence rate of SP in women was 6.80 (5.29, 8.76) per 100,000 py before July 2000, and 11.61 (9.53, 14.13) after July 2000. Figure 2 shows the results of the overall metaanalysis. Additional file 1: Fig. S3 shows the analyses by < July 2000/> July 2000 stratification.
With a randomeffects model, and a 30 days SP duration, the overall prevalence of SP in the general female population was 8.40 (7.06, 9.74) per million women. It was 5.60 (4.13, 7.06) per million before July 2000, and 10.28 (8.54, 12.01) per million after July 2000. Results are detailed in Table 3 and Additional file 1: Figs. S4 and S5.
Incidence and prevalence of SP in women with SLAM
The search identified 341 articles. Twentyone original articles were retrieved. Additional file 1: Fig. S6 shows the flow diagram depicting the search strategy. One article was excluded because it focused on chest CT findings. Ten articles were excluded because the TSC and SLAM patients were mixed, and 2 could be kept after the values were recalculated to remove TSCLAM patients [24, 41]. Three articles did not contain the data needed for metaanalysis. Thus, 6 original studies kept for metaanalysis [23, 24, 41,42,43,44]. Their characteristics are shown in Table 4.
The annual incidence rate of SP in women with SLAM was 0.13 (0.08, 0.20). Figure 3 shows the results of the metaanalysis. The prevalence of SP among women with SLAM with an SP duration of 30 days was 0.012 (0.008, 0.016). Results are detailed in Additional file 1: Fig. S7.
Probability of SLAM in SP
To determine the probability of finding a case of SLAM among women presenting with SP, the above components were combined using the Bayes equation. Results are detailed in Table 3. For the calculation of SP prevalence, we assumed that the highest accuracy would be provided by studies on SP incidence published after July 2000 and by using a median pneumothorax duration of 30 days reflecting the natural history of the condition. Using these assumptions, we found a prevalence of SLAM in SP of 0.0044 (0.0029, 0.0066). It was 0.0065 (0.0025, 0.0166) when integrating studies performed before July 2000, and 0.0036 (0.0025, 0.0051) when using studies performed after July 2000. Only slightly higher figures were found when using an arbitrary SP duration of 40 days in SLAM instead of 30 days (Table 3). The number of CTscans to perform among women with SP to detect one case of SLAM was 279. As sensitivity analysis, considering the lower and upper boundaries of the confidence interval of \(P({\text{S  LAMSP}})\), this number might have varied between 195 and 400 (the uncertainty in the estimation of \(P({\text{S  LAM}})\) is so small, given the large numbers, that taking it into account does not change this result).
In summary, using an SP duration of 30 days and studies on SP incidence performed after July 2000, the probability of finding SLAM in women presenting with SP was 0.36%. The number of CTscans to perform among women with SP to detect one case of SLAM was 279.
Probability of SLAM in women with apparent PSP
Prevalence of SLAM in women
This parameter of the equation was the same as the one used above to calculate the prevalence of SLAM in women presenting with SP. The overall prevalence of SLAM in women was 3.03 (2.48, 3.62) per million (Fig. 1).
Incidence and prevalence of PSP in the general population
The same literature search was conducted and 1046 original articles were identified. A total of 11 original articles reporting PSP incidence in the general population were retrieved [27, 28, 30, 34,35,36,37, 46,47,38,49]. No additional article was found after manual review. Their main characteristics are shown in Table 5. Additional file 1: Fig. S8 shows the flow diagram depicting the search strategy.
By metaanalysis, the overall incidence rate of PSP in the general female population was 2.70 (1.95, 3.74) per 100,000 py (Fig. 4). With time period stratification, the incidence rate was 1.54 (1.14, 2.07) per 100,000 py before July 2000, and 3.45 (2.33, 5.09) after July 2000 (Additional file 1: Fig. S9).
With a randomeffects model, and a 30 days PSP duration, the overall prevalence of PSP in the general female population was 2.62 (1.46, 3.77) per million in women (Additional file 1: Fig. S10). It was 1.23 (0.86, 1.60) per million before July 2000, and 3.36 (1.92, 4.80) per million after July 2000 (Additional file 1: Fig. S11). Results are detailed in Table 6.
Incidence and prevalence of apparent PSP in women with SLAM
Seven datasets from 6 studies were identified [23, 24, 41, 43,44,45] and were used for the calculation of the incidence and prevalence of apparent PSP in women with SLAM (Table 4). The number of women in whom a pneumothorax constituted the first manifestation of SLAM at a time when the disease was undiagnosed was used as the number of apparent PSP in the study population. From reference [43] performed by our group, we reviewed our raw data to identify inaugural episodes of apparent PSP occurring before the diagnosis of SLAM.
By metaanalysis, the annual incidence rate of apparent PSP in women with SLAM was 0.041 (0.030, 0.055) (Fig. 5). With a PSP duration of 30 days, the overall prevalence of apparent PSP among patients with SLAM was 0.0033 (0.0026, 0.0041) (Additional file 1: Fig. S12).
Probability of SLAM in PSP
The 3 components were integrated into the Bayes equation to determine the probability of SLAM among women presenting with apparent PSP. The results are detailed in Table 6. We found a probability of SLAM of 0.0038 (0.0003, 0.0066) for a PSP duration of 30 days. It was 0.0079 (0.0033, 0.1617) when integrating studies before July 2000, and 0.0030 (0.0020, 0.0046) when using studies after July 2000. As sensitivity analysis, the calculation of the prevalence of SLAM was also performed for an arbitrary PSP duration of 40 days, which led to slightly higher figures only (Table 6). The number of CTscans to perform among women with PSP to detect one case of SLAM was 331. As sensitivity analysis, considering the lower and upper boundaries of the confidence interval of \(P({\text{S  LAMPSP}})\), this number might have varied between 219 and 502 (the uncertainty in the estimation of \(P({\text{S  LAM}})\) is so small, given the large numbers, that taking it into account does not change this result).
In summary, using a PSP duration of 30 days and studies on PSP incidence performed after July 2000, the probability of finding SLAM in women presenting with apparent PSP was 0.3%. The number of CTscans to perform among women with PSP to detect one case of SLAM was 331.
Discussion
In this study, epidemiological data on SP, PSP and SLAM were used to calculate the probability of finding a case of SLAM among women presenting with both primary and secondary SP, and among women presenting with apparent PSP, some of whom having in fact undiagnosed SLAM and experiencing an inaugural episode of (sentinel) pneumothorax. Calculations were based on the Bayes theorem of conditional probability, and metaanalyses of published studies to determine each component of the Bayes equation. We found a probability of SLAM among women presenting with SP of 0.36%, and the number of CTscans to perform to detect one case of SLAM was 279. The probability of SLAM among patients presenting with apparent PSP was 0.3%, and the number of CTscans to perform to discover one SLAM case was 331. To our knowledge, this is the first study to precisely determine these parameters.
In a previous publication addressing this issue [5], Hagaman et al. estimated the probability of finding LAM among nonsmoking women with sentinel SP, i.e. inaugural apparent PSP, to be 5–30%. Using a conservative value of 5%, these authors concluded that the NNT, i.e. the number of women with sentinel SP needed to screen to uncover on case of SLAM, was about 20. Based on these assumptions, the costeffectiveness of performing a systematic chest CTscan in nonsmoking women aged 25–54 presenting with inaugural apparent PSP was calculated using a Markow statetransition model. The authors concluded that this procedure was costeffective and should be encouraged to allow earlier diagnosis of LAM. However, these conclusions relied heavily on the probability of having SLAM among women presenting with apparent PSP, and this value has not been determined precisely, but only estimated. We calculated this parameter in the present study. Our findings sharply contrast with the results of Hagaman et al., and suggest that the probability of SLAM among women with apparent PSP is much lower than previously expected. This leads to question the cost/benefit ratio of systematically screening all women presenting with apparent PSP by chest CTscan. Indeed, unlike the NNT of 20 found by Hagaman et al., we found an NNT of 331, meaning that 331 women with apparent PSP need to be screened by chest CTscan to discover one case of SLAM. This has important implications in terms of costeffectiveness. Hagaman et al. used a threshold of 50,000 $ per qualityadjusted life year (QALY) to define the costeffectiveness of an intervention. In their sensitivity analysis, the lowest prevalence of LAM in the population tested was 0.8% and was associated with a cost of 85,291 $/QALY, meaning that the intervention was no longer effective at this prevalence. With the even lower probability of SLAM in women presenting with apparent PSP found in the present study (0.3%), the intervention does not appear costeffective. Furthermore, besides costeffectiveness, the likelihood of help to harm should also be considered, given the number of incidental findings at chest CTscan screening which generate additional, possibly invasive, diagnostic procedures, as shown in lung cancer screening studies [50]. Finally, the irradiation of the population exposed to chest CTscan screening should also be considered.
The sharp contrast between the findings of Hagaman et al. and the present study has several possible explanations. First, it is unclear how the prevalence of LAM was calculated in their study. The authors cite prevalence values between 0.6 and 3 per million based on the published literature [24, 51,52,53]. However, in some of these references, only SLAM was considered [51, 52], whereas others included both SLAM and TSCLAM [24, 53]. We chose to restrict our analysis to SLAM, as TSCLAM is frequently diagnosed on the basis of extrapulmonary symptoms manifesting early in life, and the event of sentinel PSP is less relevant for the diagnosis of TSCLAM. In addition, Hagaman et al. estimated the prevalence of LAM in the United States on the basis of the number of patients recorded in the registry of the LAM Foundation (n = 850) over a 15year period (1995–2009). However, it is not clear whether all these cases were truly diagnosed within this period. One can hypothesize that: (1) both SLAM and TSCLAM were included, and (2) that, at the opening of the registry in 1995, older cases of LAM were also included, thus leading to overestimation of prevalence. The prevalence of LAM appears in the numerator of the Bayes equation. Thus, when overestimated, it contributes to overestimate the probability of LAM among women presenting with SP or PSP.
Secondly, the number of SP in this LAM population was arbitrarily estimated to 3 per patient during a 3decades period, i.e. an incidence rate of (3 × 850)/(30 × 850) = 10% per year. This is roughly similar to the 13% found by metaanalysis in the present study. However, these SP included both SP occurring in women with known LAM (including repeated events) and inaugural apparent PSP in women with undiagnosed LAM. The latter subgroup is the true population of interest, which could theoretically benefit from a screening chest CTscan at first apparent (sentinel) PSP. When restricting the calculation to this specific subpopulation, we found an incidence rate of first apparent PSP in LAM of only 4.1%, i.e. lower than the 10% of Hagaman et al. The incidence rate of apparent PSP in SLAM appears in the numerator of the Bayes equation. Thus, when overestimated, it also contributes to overestimate the probability of SLAM among women with apparent PSP.
Thirdly, based on incidence values of SP reported in the general female population between 1.2 and 9.8/100,000/year [28, 32, 40], Hagaman et al. estimated that the incidence of SP in the female population aged 25–54 was between 0.16 and 1.3/100,000/year. We found higher values by metaanalysis of recent large epidemiological studies performed after July 2000, i.e. 11.61/100,000/year for SP (Table 3) and 3.45/100,000/year for PSP (Table 6). The incidence rate of SP or PSP in the general population appears in the denominator of the Bayes equation. When underestimated, it further contributes to overestimate the probability of SLAM among women with SP or apparent PSP. In turn, overestimating the probability of SLAM leads to underestimate the NNT to uncover one case of LAM by chest CTscan screening among women presenting with SP or PSP, and to overestimate the costeffectiveness of the intervention.
Other cystic lung disease manifesting with recurrent SP such as Birt–Hogg–Dubé syndrome or pulmonary Langerhans cell histiocytosis (PLCH) could theoretically also benefit from a screening chest CTscan at first episode of apparent PSP, and combining these diagnoses might reduce the NNT to uncover one case. However, one recent study by Cattran et al. found that LAM and PLCH taken together account for only 0.13% of SP occurring in the United States [26], which is lower than the figures found for SLAM alone in the present study. Only hospitalized patients were considered in the study by Cattran et al. [26], which might result in underestimated figures. Additionally, it is not specified whether the SP episodes occurring in LAM and PLCH in this study were sentinel events, or whether they occurred in already diagnosed cases, in whom a screening chest CTscan is no longer relevant.
In the present study, we used a thorough methodology previously developed by our group to determine the prevalence of Birt–Hogg–Dubé syndrome in the general population based on metaanalyses and the Bayes theorem [7]. Particular attention was paid to avoid or minimize all potential sources of bias. Studies included in metaanalyses were carefully selected using a standard methodology [13]. Mixing of SLAM and TSCLAM was avoided and only SLAM was considered for the reasons mentioned above. In contrast to the study by Hagaman et al., SP and apparent PSP were considered separately in the present study. Indeed, CTscan screening is only relevant in women presenting with apparent PSP and undiagnosed SLAM, whereas it is of no interest in known preexisting lung diseases, including SLAM, presenting with recurrent SP. Separating these 2 settings is therefore essential. Although we analyzed both for clarity and completeness, only the analysis of undiagnosed SLAM in apparent PSP is truly relevant to assess the value of CTscan screening. For the same reasons, relapses were taken into account in the calculations made for SP, whereas for apparent PSP as sentinel event in women with undiagnosed SLAM, relapses were deliberately not considered, and only the first event was taken into account. To determine the incidence of SP and PSP in the general population, we chose to rely on studies published after July 2000 to reflect more accurately the true incidence. Indeed, substantial differences in SP and PSP incidences were observed between studies performed before and after July 2000, the latter consistently showing a higher incidence. As a true increase in incidence over time appears unlikely, we believe that the observed differences are due to more comprehensive case finding and larger sample size in more recent studies, which were based on national registries or large medical care networks, allowing to retrieve data more precisely and at a larger scale than the small studies performed decades earlier at a regional level only (county, island, or a region smaller than a country). We thus considered that the true incidence of SP and PSP was better appraised in recent studies, and chose to rely more on data from this subgroup. The duration of PSP needed to calculate the prevalence of PSP in the general population was based on a recently published randomized controlled trial on conservative versus interventional treatment of PSP, thus allowing to determine the natural history of PSP [16]. Finally, Hagaman et al. considered only nonsmoking women in their calculations, to eliminate cases of SP related to smoking. However, patients with SLAM may also smoke, as shown in one large series where 37% of patients were active smokers or exsmokers at the time of SLAM diagnosis [52], a smoking prevalence similar to that of the general population. We therefore considered that women with a history of smoking should be maintained in the population at risk of having SLAM and we did not exclude these patients from our study.
Several terms of the Bayes equation determined for the purpose of the present study deserve comments. First, to our knowledge, we provide the first determination of SLAM prevalence by metaanalysis. Although only 5 studies were available [4, 22,23,24,25], little variation was observed between countries, suggesting that the value provided by the metaanalysis (3.03 cases per million) is close to the true disease prevalence, and that it is similar in various populations worldwide. Secondly, the annual incidence of SP in SLAM determined by metaanalysis of 6 studies was 13%. This is higher than the 8% found in one large study by our group, which specifically addressed this issue [43]. As the 5 other studies were not specifically designed to calculate this parameter, it is possible that some bias has occurred, although data remain in the same range of magnitude. In any case, this confirms that the incidence of SP in LAM is about 1000 times higher than in the general female population.
Our study has limitations. The number of epidemiological studies on SLAM was small. The number of studies allowing to determine the annual incidence rate SP and apparent PSP in SLAM was also small, as was the number of patients included in each study. Thirdly, the average duration of SP in SLAM is not known. We hypothesized that it was the same as the duration of PSP in the general population, but given the different nature of the disease, we could not rule out a longer disease course in SLAM. To overcome this difficulty, we used pneumothorax durations of 30 and 40 days in the calculations of SP and PSP prevalence in SLAM, and found little variability in the final probability of SLAM among SP and apparent PSP. This reinforces the validity of our findings.
In summary, our findings question the suggestion of Hagaman et al. to perform systematic screening of women with SP or PSP by chest CTscan in search of cases of SLAM, and we believe that more studies are needed to explore this issue. Indeed, current guidelines on SP and PSP [8,9,10, 54] do not recommend systematic chest CTscan at first episode, and suggest to perform it only in selected situations. However, our findings do not challenge to use of chest CTscan for diagnosis and clinical management of individual patients, and it remains an invaluable tool in this setting. It is also worthwhile reminding that, for LAM as for other diseases, screening is not equivalent to diagnosis. Indeed, although multiple, round, thinwalled cysts evenly distributed throughout the lung parenchyma at chest CTscan are highly suggestive of LAM, its diagnosis requires at least one additional feature such as increased vascular endothelial growth factor D, the presence of renal angiomyolipoma or lymphangiomas at imaging, chylous effusion, a histopathological proof of LAM, or characteristic features of TSC, in the appropriate clinical setting [55, 56].
Conclusions
This study is the first one to precisely determine the probability of finding SLAM among women presenting with apparent PSP. This probability determines the relevance of screening this population by systematic chest CT in search of SLAM. We found that the probability of finding SLAM among women with apparent PSP was only 0.3% with an NNT of 331, a very different result from that published previously [5]. This has major impact on the cost/benefit ratio and the likelihood of help to harm of this intervention. More studies are needed before recommending systematic chest CT screening in women presenting with apparent PSP in search of SLAM and other cystic lung diseases.
Availability of data and materials
Data used in this study are available from the authors upon reasonable request.
Abbreviations
 LAM:

Lymphangioleiomyomatosis
 py:

Personyears
 SP:

Spontaneous pneumothorax
 PSP:

Primary spontaneous pneumothorax
 P:

Prevalence
 IR:

Incidence rate
References
Johnson SR. Lymphangioleiomyomatosis. Eur Respir J. 2006;27:1056–65.
McCarthy C, Gupta N, Johnson SR, Yu JJ, McCormack FX. Lymphangioleiomyomatosis: pathogenesis, clinical features, diagnosis, and management. Lancet Respir Med. 2021;9:1313–27.
Xu KF, Xu W, Liu S, Yu J, Tian X, Yang Y, et al. Lymphangioleiomyomatosis. Semin Respir Crit Care Med. 2020;41:256–68.
Harknett EC, Chang WY, Byrnes S, Johnson J, Lazor R, Cohen MM, et al. Use of variability in national and regional data to estimate the prevalence of lymphangioleiomyomatosis. QJM. 2011;104:971–9.
Hagaman JT, Schauer DP, McCormack FX, Kinder BW. Screening for lymphangioleiomyomatosis by highresolution computed tomography in young, nonsmoking women presenting with spontaneous pneumothorax is costeffective. Am J Respir Crit Care Med. 2010;181:1376–82.
Gupta N, Langenderfer D, McCormack FX, Schauer DP, Eckman MH. Chest computed tomographic image screening for cystic lung diseases in patients with spontaneous pneumothorax is cost effective. Ann Am Thorac Soc. 2017;14:17–25.
Müller ME, Daccord C, Taffé P, Lazor R. Prevalence of BirtHoggDubé syndrome determined through epidemiological data on spontaneous pneumothorax and Bayes theorem. Front Med (Lausanne). 2021;8:631168.
Baumann MH, Strange C, Heffner JE, Light R, Kirby TJ, Klein J, et al. Management of spontaneous pneumothorax: an American College of Chest Physicians Delphi consensus statement. Chest. 2001;119:590–602.
MacDuff A, Arnold A, Harvey J, BTS Pleural Disease Guideline Group. Management of spontaneous pneumothorax: British Thoracic Society pleural disease guideline 2010. Thorax. 2010;65(Suppl 2):18–31.
Tschopp JM, Bintcliffe O, Astoul P, Canalis E, Driesen P, Janssen J, et al. ERS task force statement: diagnosis and treatment of primary spontaneous pneumothorax. Eur Respir J. 2015;46:321–35.
Johannesma PC, Reinhard R, Kon Y, Sriram JD, Smit HJ, van Moorselaar RJ, et al. Prevalence of BirtHoggDube syndrome in patients with apparently primary spontaneous pneumothorax. Eur Respir J. 2015;45:1191–4.
Bintcliffe OJ, Hallifax RJ, Edey A, FellerKopman D, Lee YC, Marquette CH, et al. Spontaneous pneumothorax: Time to rethink management? Lancet Respir Med. 2015;3:578–88.
Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71.
Freeman J, Hutchison GB. Prevalence, incidence and duration. Am J Epidemiol. 1980;112:707–23.
Bouyer J, Hémon D, Cordier S, Derriennic F, Stücker I, Stengel B, et al. Epidémiologie: principes et méthodes quantitatives. INSERM, editor. Paris: Lavoisier; 2009.
Brown SGA, Ball EL, Perrin K, Asha SE, Braithwaite I, EgertonWarburton D, et al. Conservative versus interventional treatment for spontaneous pneumothorax. N Engl J Med. 2020;382:405–15.
Spronk I, Korevaar JC, Poos R, Davids R, Hilderink H, Schellevis FG, et al. Calculating incidence rates and prevalence proportions: not as simple as it seems. BMC Public Health. 2019;19:512.
Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Med Res Methodol. 2005;5:13.
Barendregt JJ, Doi SA, Lee YY, Norman RE, Vos T. Metaanalysis of prevalence. J Epidemiol Community Health. 2013;67:974–8.
Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Introduction to metaanalysis. Chichester: Wiley; 2009. p. 421.
Walker SP, Bibby AC, Halford P, Stadon L, White P, Maskell NA. Recurrence rates in primary spontaneous pneumothorax: a systematic review and metaanalysis. Eur Respir J. 2018;52:1800864.
Cordier JF, Lazor R. Perspectives on lymphangioleiomyomatosis in France. In: Moss J, editor. LAM and other diseases characterized by smooth muscle proliferation. Lung biology in health and disease, vol. 131. New York: Marcel Dekker Inc.; 1999. p. 9–31.
Johnson SR, Tattersfield AE. Clinical experience of lymphangioleiomyomatosis in the UK. Thorax. 2000;55:1052–7.
Hayashida M, Seyama K, Inoue Y, Fujimoto K, Kubo K, Respiratory Failure Research Group of the Japanese Ministry of Health, Labor and Welfare. The epidemiology of lymphangioleiomyomatosis in Japan: a nationwide crosssectional study of presenting features and prognostic factors. Respirology. 2007;12:523–30.
Kristof AS, Zhi Li P, Major P, Landry JS. Lymphangioleiomyomatosis and tuberous sclerosis complex in Quebec: prevalence and healthcare utilization. Chest. 2015;148:444–9.
Cattran A, Meganathan K, Gupta N. Epidemiology and healthcare utilization of spontaneous pneumothorax and diffuse cystic lung diseases in the United States. Ann Am Thorac Soc. 2021;18:534–6.
Hallgrimsson JG. Spontaneous pneumothorax in Iceland with special reference to the idiopathic type. A clinical and epidemiological investigation. Scand J Thorac Cardiovasc Surg Suppl. 1978;12:1–85.
Melton LJ 3rd, Hepper NG, Offord KP. Incidence of spontaneous pneumothorax in Olmsted County, Minnesota: 1950 to 1974. Am Rev Respir Dis. 1979;120:1379–82.
Ferguson LJ, Imrie CW, Hutchison J. Excision of bullae without pleurectomy in patients with spontaneous pneumothorax. Br J Surg. 1981;68:214–6.
Primrose WR. Spontaneous pneumothorax: a retrospective review of aetiology, pathogenesis and management. Scott Med J. 1984;29:15–20.
SuarezVarel MM, MartinezSelva MI, LlopisGonzalez A, MartinezJimeno JL, PlazaValia P. Spontaneous pneumothorax related with climatic characteristics in the Valencia area (Spain). Eur J Epidemiol. 2000;16:193–8.
Gupta D, Hansell A, Nichols T, Duong T, Ayres JG, Strachan D. Epidemiology of pneumothorax in England. Thorax. 2000;55:666–71.
Chen CH, Kou YR, Chen CS, Lin HC. Seasonal variation in the incidence of spontaneous pneumothorax and its association with climate: a nationwide populationbased study. Respirology. 2010;15:296–302.
Bobbio A, Dechartres A, Bouam S, Damotte D, Rabbat A, Regnard JF, et al. Epidemiology of spontaneous pneumothorax: genderrelated differences. Thorax. 2015;70:653–8.
Schnell J, Koryllos A, LopezPastorini A, Lefering R, Stoelben E. Spontaneous pneumothorax. Dtsch Arztebl Int. 2017;114:739–44.
Hallifax RJ, Goldacre R, Landray MJ, Rahman NM, Goldacre MJ. Trends in the incidence and recurrence of inpatienttreated spontaneous pneumothorax, 1968–2016. JAMA. 2018;320:1471–80.
Hiyama N, Sasabuchi Y, Jo T, Hirata T, Osuga Y, Nakajima J, et al. The three peaks in age distribution of females with pneumothorax: a nationwide database study in Japan. Eur J Cardiothorac Surg. 2018;54:572–8.
Kim D, Jung B, Jang BH, Chung SH, Lee YJ, Ha IH. Epidemiology and medical service use for spontaneous pneumothorax: a 12year study using nationwide cohort data in Korea. BMJ Open. 2019;9:e028624.
Lee SH, Seo YW, Kwak SG. Influence of meteorological factors on development of spontaneous pneumothorax. Medicine (Baltimore). 2022;101:e31488.
Bense L, Eklund G, Wiman LG. Smoking and the increased risk of contracting spontaneous pneumothorax. Chest. 1987;92:1009–12.
Oprescu N, McCormack FX, Byrnes S, Kinder BW. Clinical predictors of mortality and cause of death in lymphangioleiomyomatosis: a populationbased registry. Lung. 2013;191:35–42.
Bee J, Bhatt R, McCafferty I, Johnson SR. A 4year prospective evaluation of protocols to improve clinical outcomes for patients with lymphangioleiomyomatosis in a national clinical centre. Thorax. 2015;70:1202–4.
Gonano C, Pasquier J, Daccord C, Johnson SR, Harari S, Leclerc V, et al. Air travel and incidence of pneumothorax in lymphangioleiomyomatosis. Orphanet J Rare Dis. 2018;13:222.
Cheng C, Xu W, Wang Y, Zhang T, Yang L, Zhou W, et al. Sirolimus reduces the risk of pneumothorax recurrence in patients with lymphangioleiomyomatosis: a historical prospective selfcontrolled study. Orphanet J Rare Dis. 2022;17:257.
Johnson J, Stewart I, Johnson SR. Disease monitoring using lung function trajectory in lymphangioleiomyomatosis: assessment in two national cohorts. Thorax. 2023;78:61–8.
WynnWilliams N. A ten years’ study of spontaneous pneumothorax in a community. Thorax. 1957;12:253–7.
Huang YH, Chang PY, Wong KS, Chang CJ, Lai JY, Chen JC. An agestratified longitudinal study of primary spontaneous pneumothorax. J Adolesc Health. 2017;61:527–32.
Olesen WH, Titlestad IL, Andersen PE, LindahlJacobsen R, Licht PB. Incidence of primary spontaneous pneumothorax: a validated, registerbased nationwide study. ERJ Open Res. 2019;5:1–9.
Ogata H, Matsumoto K, Nakatomi K, Nakashima N, Shoji F, Koto H. Seasonal and meteorological impacts on primary spontaneous pneumothorax. J Thorac Dis. 2018;10:3957–64.
National Lung Screening Trial Research Team, Aberle DR, Adams AM, Berg CD, Black WC, Clapp JD, et al. Reduced lungcancer mortality with lowdose computed tomographic screening. N Engl J Med. 2011;365:395–409.
Kitaichi M, Nishimura K, Itoh H, Izumi T. Pulmonary lymphangioleiomyomatosis: a report of 46 patients including a clinicopathologic study of prognostic factors. Am J Respir Crit Care Med. 1995;151:527–33.
Urban T, Lazor R, Lacronique J, Murris M, Labrune S, Valeyre D, Groupe d’Etudes et de Recherche sur les Maladies “Orphelines” Pulmonaires (GERM“O”P). Pulmonary lymphangioleiomyomatosis. A study of 69 patients. Medicine (Baltimore). 1999;78:321–37.
Oh YM, Mo EK, Jang SH, Yoo CG, Kim YW, Seo JW, et al. Pulmonary lymphangioleiomyomatosis in Korea. Thorax. 1999;54:618–21.
Schnell J, Beer M, Eggeling S, Gesierich W, Gottlieb J, Herth FJF, et al. Management of spontaneous pneumothorax and postinterventional pneumothorax: German S3 guideline. Respiration. 2019;97:370–402.
Johnson SR, Cordier JF, Lazor R, Cottin V, Costabel U, Harari S, et al. European Respiratory Society guidelines for the diagnosis and management of lymphangioleiomyomatosis. Eur Respir J. 2010;35:14–26.
Gupta N, Finlay GA, Kotloff RM, Strange C, Wilson KC, Young LR, et al. Lymphangioleiomyomatosis diagnosis and management: highresolution chest computed tomography, transbronchial lung biopsy, and pleural disease management. An official American Thoracic Society/Japanese Respiratory Society clinical practice guideline. Am J Respir Crit Care Med. 2017;196:1337–48.
Acknowledgements
Not applicable.
Funding
Open access funding provided by University of Lausanne. This work has been funded with internal resources.
Author information
Authors and Affiliations
Contributions
Study conception and design: RL, PT; data collection: AS, MEM; data analysis and interpretation: PT, RL, AS, MEM; manuscript drafting: AS, RL, PT, CD; critical manuscript revision: all authors; final manuscript approval: all authors.
Corresponding author
Ethics declarations
Ethics approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Additional file 1. Fig S1.
PRISMA flow diagram for SLAM epidemiology. Fig S2. PRISMA flow diagram for SP epidemiology. Fig S3. Incidence rate of SP in women stratified by period before/after July 2000. Fig S4. Overall prevalence of SP in women with average pneumothorax duration of 30 days. Fig S5. Prevalence of SP in women stratified by period before/after July 2000, with average pneumothorax duration of 30 days. Fig S6. PRISMA flow diagram for SP and PSP in women with SLAM. Fig S7. Prevalence of SP in women with SLAM with average duration of SP of 30 days, randomeffects model. Fig S8. PRISMA flow diagram for PSP epidemiology. Fig S9. Incidence rate of PSP in women of the general population stratified by period before/after July 2000. Fig S10. Overall prevalence of PSP in women of the general population, with an average PSP duration of 30 days. Fig S11. Prevalence of PSP in women of the general population stratified by period before/after July 2000, with average PSP duration of 30 days. Fig S12. Prevalence of apparent PSP in women with SLAM, with average duration of PSP of 30 days, randomeffects model.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Suter, A., Müller, ME., Daccord, C. et al. Probability of sporadic lymphangioleiomyomatosis in women presenting with spontaneous pneumothorax. Orphanet J Rare Dis 18, 180 (2023). https://doi.org/10.1186/s13023023027845
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13023023027845
Keywords
 Lymphangioleiomyomatosis
 Spontaneous pneumothorax
 Primary spontaneous pneumothorax
 Prevalence
 Bayes theorem
 Metaanalysis