The mode of presentation in this study was in keeping with previous studies which have reported that vertebral fractures are the most common type of fracture associated with PAO, and that they usually affect the thoracolumbar spine [2, 3, 12, 13]. While Hadji et al. reported a mean of 3.3 fractures per patient in their case–control study of PAO [6], eleven (68.8%) of our patients had fractures of five or more vertebrae. One patient in this series had a sacral fracture. This type of fracture has been previously described in two patients with PAO, where the use of LMWH was identified as a predisposing factor [14, 15]. Another patient had ankle fractures, which is an uncommon but probably underreported presentation of PAO [16]. Rozenbaum et al. described a patient with PAO who had fractures at the ankle, hip and knees, but this was associated with a cytomegalovirus infection [17] which did not occur in our patient with ankle fractures.

Back pain was the most common presenting complaint occurring in 81.3% of patients. Height loss and kyphosis were presenting features in two patients (12.5%). Height loss in this study was less frequent than in the study of Smith et al. where it was recorded 41.7% of cases, whereas the prevalence of kyphosis was the same as reported by Smith [18].

The average age at presentation was 34.1 years which is slightly higher than the average age at pregnancy of women in Scotland, which was 31.1 years in 2021 [19]. This is similar to the study of Tuna et al. where the mean age at presentation was 34.2 years [20], but lower than that reported by Kyvernitakis et al. who recorded a mean age at presentation of 39.5 years and proposed that PAO was an age-related disease [21].

Most patients were affected during their first pregnancies, which is in keeping with previous studies [13], and the diagnosis was most commonly made postpartum which again is in keeping with previous studies [6, 13].

This is the first study to estimate the incidence and prevalence of PAO. Since this is the only specialist osteoporosis clinic in NHS Lothian, we think it is likely all individuals with suspected PAO who came to medical attention will have been referred to our clinic. While it is possible that women who had experienced vertebral or other fractures during and after pregnancy did not seek medical attention, we feel this is unlikely. With these assumptions, we estimate the incidence of PAO to be 6.8/100,000 pregnancies with a population prevalence of 4.1/100,000 women. This is approximately tenfold higher than previously estimated in the literature where a figure of 4–8 per million pregnancies has been quoted [3, 6], but to our knowledge, this estimate is not based on any firm data.

Eight of the sixteen patients (50.0%) used an antithrombotic agent to prevent thromboembolic disease which in most cases was LMWH. This is in keeping with previous reports which have identified heparin as a potential risk factor for PAO [2, 6, 15]. The proportion of individuals treated with antithrombotic agents in this series is considerably higher than the 17% background rate of thromboprophylaxis during pregnancy in NHS Lothian (personal communication—Allyn Dick, Clinical Auditor, Women and Children’s Services, NHS Lothian). The mechanisms by which heparin might predispose to osteoporosis are incompletely understood [22]. Preclinical studies have shown that unfractionated heparin (UFH) and LMWH reduce bone formation and that UFH, but not LMWH, stimulates bone resorption [23]. Irie and colleagues [24] reported that heparin binds to and inhibits osteoprotegerin, and through this mechanism, stimulates RANKL-induced osteoclastic activity. The authors of this work speculated that LWMH may be less likely to inhibit OPG because of its smaller size, but this was not proven experimentally. Several clinical studies have reported associations between heparin use in pregnancy, increased bone loss and fractures [22]. There is less information on the effects of LMWH on bone but prospective studies have shown numerically greater increases in bone loss in LMWH users during pregnancy than controls, albeit with differences that were not significant [25].

Another potential risk factor is corticosteroid use [20, 26]. Five of our patients used corticosteroids during pregnancy, but four of these were inhalers for asthma and one was a local corticosteroid injection for soft tissue rheumatism. Inhaled corticosteroid use in patients with asthma has been associated with a decrease in BMD but the changes were small and not associated with fractures [27]. One patient in this series used levothyroxine to manage their hypothyroidism during pregnancy. While over-replacement with thyroxine is associated with reduced BMD, we know of no data to suggest that hypothyroidism is a risk factor for osteoporosis. Two patients received treatment with anticonvulsants during pregnancy, one for epilepsy and another for schizophrenia, while another was treated with an LHRH agonist for many years before pregnancy to manage her endometriosis. Though anticonvulsant use has previously been described as a risk factor for PAO [6] and as a risk factor for fractures in general, LHRH agonist use has not. LHRH agonists are known to decrease the BMD of women with endometriosis, but this can completely resolve after the treatment is withdrawn [28].

A family history of osteoporosis has been reported to be more common in women with PAO than healthy controls [7], and in our series, almost one third reported having a family history of osteoporosis. Some patients who present with features of PAO carry pathogenic mutations in the COL1A1, COL1A2 and LRP5 genes [26]. In keeping with this, Hardcastle et al. [5] and Smith et al. [2] each identified one patient who presented with PAO as the first manifestation of osteogenesis imperfecta. We did not conduct genetic analysis as part of the present study, but we are planning to do so in the future.

Four patients in our series had complicated pregnancies, two with hyperemesis gravidarum. In this regard, Hadji et al. found a higher prevalence of coagulation disorders, congenital malformations and pregnancy-related disorders in cases of PAO than controls [6]; however, a recent study did not find an association between hyperemesis gravidarum and osteoporosis [29].

As in previous studies [9, 30, 31], various bone-targeted treatments were given including calcium and vitamin D supplements, bisphosphonates and teriparatide. In most cases, BMD increased between the time of the original diagnosis and follow-up, but there was no significant difference in response between treatment types. It should be acknowledged that we cannot make any firm conclusions about relative efficacy of these agents since the treatment was not randomly allocated, and the numbers were small. Furthermore, the improvements observed could be due in part, to the known recovery of BMD after PAO even without bone targeted treatment [32]. The efficacy of different treatments for PAO remains a subject of debate. In one cases series, teriparatide was found to be superior to calcium and vitamin D supplements [30, 31], whereas a systematic review by Hong and Rhee reported that teriparatide and bisphosphonates were associated with similar increases in BMD in PAO and were superior to no active treatment [33].

It has been reported that almost a quarter of patients with PAO will experience a subsequent fracture, and that this risk correlates with the number of fractures a patient has at diagnosis [21]. However, this was not observed in the present series where only one patient had a further fracture, and this did not occur in the context of a further pregnancy.

The high proportion of individuals with a family history of osteoporosis raises the possibility that genetic factors may play a pathogenic role as has been suspected previously [26], but it should be acknowledged that shared environmental risk factors could also contribute to the finding of a positive family history of osteoporosis in these women.

In conclusion, this case series represents a significant addition to the literature on PAO. It represents one of the largest case series published to date with details of clinical and pharmacological risk factors, mode of presentation and response to treatment. Additionally, it is the only report we are aware of in which an evidence-based estimate of the true incidence and prevalence of PAO has been provided.