HBM prevalence on DXA databases
In total, 335,115 historical DXA scans were screened across 13 databases, collected over a combined total of 110.2 years, the earliest from 1992. DXA scans of all those with T- or Z-scores ≥ +4 from ten centres were inspected by both CG and JT; 49.4% were considered to have artefactually raised BMD due to degenerative changes (Table 1); 9.7% of DXA scans had evidence of other artefacts to explain their high BMD or were unverifiable. Of the remaining cases, 5.8% did not meet our Z-score threshold for defining HBM. After screening DXA databases at the other three NHS centres, local investigators identified a further 86 HBM cases as meeting our entry criteria. The final prevalence of HBM is shown in Table 2. When results from searching Hologic and Lunar databases were combined, the overall prevalence of HBM was 0.181%. Indication for DXA referral was examined in a subgroup of 22% of scans at the largest centre in Hull (Online Resource Table 1). The most common indication was a suspicion of osteoporosis based upon height loss or low trauma fracture (28.8%), which also accounted for 35.3% of indications for DXAs which were found to have a T-/Z-score ≥ +4. Treatment monitoring prompted 17.1% of overall referrals but only accounted for 4.8% of referrals for DXA in individuals found to have high BMD.
Descriptive analyses of HBM index cases and their relatives and spouses
We recruited 258 (41%) of HBM cases into our subsequent study of the detailed phenotype of HBM, identified from a total of 15 sites in England and Wales (Fig. 1). These cases were similar to those not recruited, except non-participants were shorter and had slightly lower left hip sBMD (Online Resource Table 2). Eight hundred ninety-three relatives were invited to participate, of whom 236 (26.4%) were recruited. Two hundred seventeen spouses/partners were invited to participate, of whom 61 (28.1%) were recruited; two individuals invited two partners (Fig. 1). Of the 258 HBM index cases recruited, 103 (39.9%) contributed one or more relatives into the study; 94.6% of index cases, 86.6% relatives and 93.3% spouses were able to be examined. Participants ranged in age from 18 to 90 years, and all but three were Caucasian.
The majority of index cases were female and spouses male, whilst relatives showed a more even gender distribution (Table 3). Most female index cases and spouses were post-menopausal, whereas just over half of female relatives had passed the menopause because relatives were generally younger than index cases and spouses. Despite their similar proportions of post-menopausal females, a greater proportion of index cases had taken oestrogen replacement compared to spouses. Index cases were shorter than relatives and spouses, likely reflecting differences in gender distribution. BMI was higher amongst index cases compared to relatives and spouses.
BMD Z-scores showed a Gaussian rather than a bi-modal distribution in all three groups (Fig. 2). As expected, mean Z-scores of the total hip and L1, both separately and combined, were considerably higher in HBM cases than spouses, whereas mean values in relatives were higher than spouses but lower than HBM cases (Table 3). This was despite Z-scores in spouses being elevated in comparison with the DXA scanner manufacturer’s reference population. Although L1 area initially appeared greater in spouses compared to index cases, following adjustment for age at time of DXA, gender, years since menopause, oestrogen replacement use, height and weight, L1 area was greater in index cases than spouses, with relatives showing intermediate results. Similar findings were seen irrespective of whether results were restricted to centres with Hologic or Lunar scanners (data not shown).
Clinical characteristics associated with unexplained HBM
To analyse clinical characteristics associated with HBM using logistic regression (which enabled adjustment for confounders), relatives were assigned as cases or controls based upon the Z-score +3.2 threshold (see Fig. 2). When comparing BMD between HBM cases (258 index, 94 affected relatives and three affected spouses) and controls (142 unaffected relatives and 58 unaffected spouses) categorised in this way, HBM cases had greater summed L1 and total hip Z-scores than controls, 6.98 (6.76, 7.20) vs. 1.04 (0.74, 1.35), p < 0.001. Cases were older (mean difference [95% CI] 7.7 [5.2, 10.3] years), more often female (272 [76.6%] vs. 93 [46.5%]), and women were more often post-menopausal (218 [82.9%] vs. 48 [54.5%]), with a history of oestrogen replacement (128 [52.7%] vs. 15 [19.2%]), p < 0.001 for all. After adjusting for these differences, HBM cases had a greater mean BMI than controls (2.2 [1.3, 3.1] kg/m2, p < 0.001). HBM cases had increased odds of an enlarged mandible (four HBM cases having prognathism), a broad frame, misshapen or extra bone at the site of tendon and/or ligament insertions, together with a larger shoe size (adjusted mean difference 0.4 of a UK size; Table 4). Whilst there was no difference in the reporting of dental overcrowding, there was a trend towards increased odds of reporting a history of structural oral/dental abnormalities amongst HBM cases. The odds of reporting visual or auditory problems, hearing aid use or abnormal vision or hearing being found on examination were similar amongst cases and controls. Equally, the odds of reporting spinal stenosis, or an operation for spinal stenosis, were similar amongst cases and controls (adjusted OR 0.98 [0.39, 2.45], p = 0.959, adjusted for gender and age). Similarly the odds of cranial nerve palsy were no higher amongst HBM cases compared with controls (adjusted OR 1.38 [0.51, 3.70], p = 0.522). There was a weak trend towards increased reporting of carpal tunnel syndrome amongst HBM cases. Renal calculi and osteomyelitis were no more commonly reported amongst cases than controls and were infrequent.
Interestingly, HBM cases had increased odds of reporting sinking when trying to swim (Table 4). Further adjustment for body weight, height and history of chronic obstructive pulmonary disease, asthma and smoking (as proxies for lung capacity) did not materially affect this association. Whilst fracture history was no different between cases and controls, HBM cases had reduced odds of reporting a family history of fracture. HBM cases were more likely to report current or previous experience of pain in their mandible, skull/head (including self-reported migraine) and limb bones in general. Unadjusted results suggested increased odds of joint pain in cases compared with controls; however, this was not apparent after adjustment. HBM cases had increased odds of reporting reduced exercise tolerance which persisted after adjustment for age and gender but partially attenuated after further adjustment for weight (adjusted OR 2.58 [1.39, 4.78], p = 0.003). On examination, there was no objective evidence of gait abnormality. However, after adjustment for age, gender, menopause and weight, the odds of reporting a previous joint replacement were the greater amongst cases than controls–47 (13.2%) vs. 8 (4.0%), OR 2.69 (1.10, 6.60), p = 0.031. After adjusting for age and gender, the odds of reporting a history of cancer were similar amongst cases and controls (OR 1.64 [0.84, 3.19], p = 0.145).
When considering five cardinal features associated with HBM after age and gender adjustment: (a) BMI >30, (b) broad frame, (c) sinking when swimming, (d) mandible enlargement on examination and (e) extra bone identifiable on clinical examination, 70% of HBM cases had two or more of these features, whilst 42% had four or more (18% having all five), so that the positive predictive value of four or more features was 78.0. When the frequency of clinical features was compared between index cases vs. all relatives and spouses combined, odds ratios were only partially attenuated (Online Resource Table 3). Mean laboratory values were similar between cases and controls, other than HBM cases had a lower platelet count than controls (267.9 [260.1, 275.8] vs. 275.1 [264.4, 285.8], respectively, mean difference 16.5 [3.6, 29.4] × 109/L, p = 0.012); platelet count remained within the reference range in 95.3% of the study population.
Other potential causes of raised BMD
In index cases with unexplained HBM, although no other cause of HBM was evident from initial analysis of DXA database scan images, this diagnosis was re-evaluated using additional information provided by clinical history, examination, X-rays and blood tests. No HBM cases had the clear dysmorphic features of previously reported extreme skeletal dysplasias such as pycnodysostosis or Camurati–Engelmann disease. Excessive oestrogen replacement implant use has been associated with substantial increases in BMD . Eighteen female HBM cases reported oestrogen replacement implant use of whom five had affected first-degree relatives based upon the +3.2 Z-score definition described above, suggesting a genetic basis to their HBM. Three index cases gave a history of lithium treatment (reported to increase BMD in mice ), two of whom had relatives with HBM, whilst one did not. No cases reported treatment with recombinant parathyroid hormone or strontium ranelate. None of the index cases who reported ever having fractured had radiological features consistent with osteopetrosis  nor evidence of pancytopenia. One HBM case had treated acromegaly, one myelofibrosis and one reported investigations for possible ankylosing spondylitis. Three cases were identified with serum phosphate level of <0.70 mmol/L and bridging osteophytes of the lower thoracic and upper lumbar spine, of whom one also had evidence of new bone formation at the pelvis and upper femorae. Whilst radiological features were consistent with diffuse idiopathic skeletal hyperostosis, in combination with low phosphate, these may represent the enthesopathy reported in X-linked hypophosphataemic rickets , although other clinical features, such as short stature, a family history or a history of fragility fractures, were absent. Thus, in 20 individuals recruited with unexplained HBM, more detailed clinical assessment gave a possible explanation for their raised BMD, but analyses of clinical characteristics were unchanged after their exclusion (Online Resource Table 4), as were fracture analyses (data not shown).