Genotype-based comparison of bone microstructure in adult patients with classical osteogenesis imperfecta

Abstract

Summary

In 115 adult patients with classical OI, reduced bone structure was observed in patients with COL1A1 variants compared to those with COL1A2 variants. In addition, bone structure in the distal radius was differentially affected in individuals with quantitative COL1A1/2 variants in comparison to patients carrying qualitative variants.

Purpose

Classical osteogenesis imperfecta (OI) is genetically determined by pathogenic variants in the COL1A1/2 genes encoding collagen type I. Although genetic testing is widely used today, the genotype–phenotype correlation is still a matter of debate.

Methods

A retrospective analysis was carried out in a total of 115 adult OI patients who presented to our outpatient clinic. Only patients with available genetic test results and clinically diagnosed OI were included. Biochemical parameters, bone density (DXA), and bone microstructure (HR-pQCT) were assessed and compared between genotype variants.

Results

Overall, 71 patients showed (likely) pathogenic variants in the COL1A1 and 44 in the COL1A2 gene. Forty-six variants were predicted as causing quantitative and 58 as causing qualitative effects, whereas the potential effect of 11 variants could not be predicted. COL1A1 variants were associated with significantly lower trabecular bone mineral density (Tb.BMD) and bone volume fraction (BV/TV) compared to COL1A2 variants both at the radius (Tb.BMD: p = 0.008; BV/TV: p = 0.010) and tibia (Tb.BMD: p = 0.015; BV/TV: p = 0.028). In addition, trabecular parameters only in the distal radius were significantly (p < 0.05) reduced in patients with quantitative variants in comparison to patients with qualitative variants. OI type I patients had significantly higher cortical bone mineral density (Ct.BMD) than type III. No other microstructural differences were found between the types I, III, and IV.

Conclusions

Genetic stratification in adult OI patients revealed differences in bone microstructure, while these were absent when individuals were grouped according OI types (Sillence classification). Future studies are needed to further investigate the genotype–phenotype correlation in patients with classical OI.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00198-025-07774-w.

Introduction

Osteogenesis imperfecta (OI) is a genetically and clinically heterogeneous disease characterized by increased bone fragility, growth failure, and various degrees of severity and genetic manifestations [1–3]. The clinical stratification of OI is based on the 1979 Sillence classification system [4], which differentiates according to clinical features. OI type I is characterized by the occurrence of no (minimal) bone deformities, normal height, low-traumatic fractures, and blue sclerae. OI type III patients exhibit moderate to severe bone deformities, (severe) short stature, atraumatic fractures, and normal sclerae. OI type IV patients present with mildly reduced height, a moderate severity of bone deformities and fracture frequency, as well as normal sclerae. The most severe form of OI, type II, shows heavily deformed long bones and is typically lethal perinatally and thus is not prevalent in adult patients [4].

Advances in genetic research have helped to identify variants in the COL1A1 and COL1A2 genes encoding type I collagen as causative for classical OI types I–IV [5–7]. As a result, early diagnosis and genotype determination, together with the initiation of screening in the presence of a positive family history, frequent fractures, bone deformities, or other non-bone-related symptoms, are becoming increasingly common in clinical practice, leading to improved diagnostics [8–11].

Although significant progress has been made in understanding the genetic and molecular basis of OI, most studies to date have focused on pediatric populations. The characterization of adults with OI remains inadequate, and data regarding adult OI patients is still scarce today [6, 12].

Here, we provide a comprehensive clinical and biochemical characterization of 115 adult patients with classical OI types I, III, or IV. We focused on the investigation of biochemical parameters and bone microstructure, emphasizing stratification in genetic subgroups according to COL1A1 vs. COL1A2 variants as well as patients with quantitative vs. qualitative collagen type I variants.

Methods

Study design

This retrospective study was performed in accordance with the local ethics committee (PV5364) and the Declaration of Helsinki, and all patients gave written informed consent. A total of 115 adult OI patients (≥ 18 years of age) who presented to our specialized osteologic outpatient clinic (National Bone Board) between September 2010 and March 2023 were included. Only patients with available genetic test results were analyzed. All patients were clinically diagnosed with classical OI confirmed by detection of (likely) pathogenic variants in COL1A1 or COL1A2. Patients received biochemical analyses, bone densitometry via dual-energy X-ray absorptiometry, and bone microstructure measurements via high resolution peripheral quantitative computed tomography as part of the routine clinical assessment. The age at the appearance of the first fracture, as well as the number of vertebral and peripheral fractures, was determined by a medical history questionnaire. For further analyses, we compared clinically classified OI types I, III, and IV and analyzed patients with regard to the genotype. Therefore, patients carrying COL1A1 vs. COL1A2 variants and patients with quantitative vs. qualitative variants were compared regardless the clinical appearance or Sillence classification.

Biochemical analysis

Blood and urine samples were routinely analyzed at the local laboratory as described elsewhere [13]. Serum calcium, phosphate, alkaline phosphatase (ALP), creatinine, bone-specific alkaline phosphatase (b-ALP), osteocalcin, parathyroid hormone (PTH), 25-hydroxyvitamin D (25(OH)D), and deoxypyridinoline per creatinine in urine (DPD) were assessed and then compared with the reference ranges provided by the local laboratory.

Dual-energy X-ray absorptiometry (DXA)

Areal bone mineral density (aBMD, with T- and Z-scores) was measured using DXA (Lunar iDXA, GE Healthcare, Madison, WI, USA). The lumbar spine vertebral bodies L1–L4 and each side (left/right) of the hip (femoral neck and total hip) were evaluated, if applicable. For further analysis, the lowest T-score of the lumbar spine and the proximal femora with corresponding Z-score were used. For quality control of the DXA, daily calibration scans were performed with a special phantom according to the manufacturer’s recommendations. The accuracy tests included least significant change calculations according to the guidelines of the International Society for Clinical Densitometry (ISCD) [14].

High resolution peripheral quantitative computed tomography (HR-pQCT)

Patients received either first- (XCT1) or second-generation (XCT2) HR-pQCT (XtremeCT and XtremeCT II, Scanco Medical AG, Brüttisellen, Switzerland) scans of the non-dominant distal radius and the opposite distal tibia, using the manufacturer’s protocol of the standard in vivo scan for each HR-pQCT (XCT1: 59.4 kVp, 900 µA, 100 ms integration time, 82.0 µm voxel size; XCT2: 68.0 kVp, 1,470 µA, 43 ms integration time, 60.7 µm voxel size) [15]. Scanning area covers 110 slices for XCT1 and 168 slices for XCT2, which corresponds to a total length of the scanning area of 9.0 mm and 10.2 mm, respectively.

Volumetric bone mineral density (vBMD) was expressed as total BMD (Tt.BMD, mg HA/cm³), cortical BMD (Ct.BMD, mg HA/cm³), as well as trabecular BMD (Tb.BMD, mg HA/cm³). Microarchitecture parameters followed the standardized nomenclature of the IOF-ASBMR-ECTS working group [15] and included bone volume to total volume ratio (BV/TV), trabecular number (Tb. N, mm⁻¹), trabecular thickness (Tb.Th, mm), trabecular separation (Tb.Sp, mm), and cortical thickness (Ct.Th, mm). Geometric values included total bone area (Tt.Ar, mm²), trabecular bone area (Tb.Ar, mm²), and cortical bone area (Ct.Ar, mm²). To enable comparisons across sex and age groups, the HR-pQCT parameters were normalized using device-, age-, and sex-specific reference values for further analyses [16, 17], thereby improving the comparability of results independent of scanner type, age, or sex.

Variant classification

Variants in COL1A1 and COL1A2 predicted to result in premature stop codons, frameshifts, or splice site alterations, potentially leading to nonsense-mediated decay (NMD) and, thus, to haploinsufficiency, were categorized as quantitative. In contrast, variants involving single nucleotide changes that lead to structurally altered collagen proteins with dominant-negative effects on collagen stability and function were classified as qualitative. Nonsense variants located in the C-terminal region were further assessed using the NMD prediction tool AutoPVS1 [18, 19] and classified as either qualitative or quantitative based on the predicted likelihood of NMD escape. Variants in the C-terminus with ambiguous NMD prediction or in the last exon were evaluated as qualitative using the “last exon/50–55 nucleotide rule” [20, 21]. In general, mRNAs containing a premature termination codon (PTCs) are degraded by the NMD pathway, leading to haploinsufficiency. However, PTCs in the last exon or within ~ 50–55 nucleotides upstream of the final exon–exon junction often escape NMD due to the lack of downstream exon–junction complexes [21, 22]. Such variants can produce truncated collagen proteins that integrate into the triple helix and impair its stability and function via a dominant-negative effect. They are thus classified as qualitative and typically associated with a more severe phenotype [23]. Variants for which the molecular effect remained inconclusive and the “last exon/50–55 nucleotide rule” could not be reliably applied were classified as unclear (Suppl. Table 1).

Statistical analysis

JASP 0.19.1 (University of Amsterdam, Netherlands) and GraphPad Prism 10.4.1 (GraphPad Software, San Diego, CA, USA) were used for statistical analysis. Results were expressed as mean ± standard deviation (SD) and mean percentage of median (%median) of reference values for HR-pQCT parameters. To evaluate the normal distribution of the data, the Shapiro–Wilk test was used. Comparisons between two subgroups were conducted using an unpaired two-tailed t-test for parametric data and the Mann–Whitney U test for nonparametric data. When testing differences between more than two groups, a one-way analysis of covariance (ANOVA) with Tukey test was used for normally distributed data and the Kruskal–Wallis H test with Dunn test was used for non-parametric data. Differences in the distribution in subgroups were tested with the chi-square test.

Results

Characterization of OI cohort

The characteristics of the entire study cohort classified by Sillence classification system can be found in Table 1.

Table 1.

Overview of the total cohort of adult OI patients categorized by Sillence classification system. Classical adult OI patients were divided according the Sillence classification [4]. Total fractures were further classified into vertebral and peripheral fractures with mean values provided. Numbers in bold indicate statistical significance (p < 0.05)

		OI type I (n = 76)				OI type III (n = 14)				OI type IV (n = 25)				I vs III	I vs IV	III vs IV
Parameter		Mean	SD	Min	Max	Mean	SD	Min	Max	Mean	SD	Min	Max	p-value	p-value	p-value
Demographics
	Sex (male/female)	32/44				7/7				11/14				0.584	0.868	0.718
	Age (years)	43.8	15.5	18.0	82.0	38.1	14.7	19.0	69.0	46.2	15.3	19.0	70.0	0.413	0.771	0.257
	Weight (kg)	71.1	16.4	47.0	127.0	44.4	26.2	18.0	106.0	66.0	19.1	27.0	111.0	< 0.001	0.861	0.039
	Height (m)	1.64	0.10	1.47	1.89	1.18	0.29	0.8	1.74	1.61	0.19	1.25	1.89	< 0.001	> 0.999	< 0.001
	BMI (kg/m2)	26.4	5.1	17.7	47.2	30.2	8.9	19.0	45.1	25.1	5.9	17.3	41.8	> 0.999	0.504	0.232
DXA
	Femoral T-score	− 1.9	1.3	− 4.9	3.6	− 2.4	1.5	− 3.9	− 0.9	− 2.4	1.0	− 4.7	− 0.5	> 0.999	0.210	> 0.999
	Femoral Z-score	− 1.5	1.2	− 4.5	3.5	− 2.7	1.0	− 3.7	− 1.5	− 1.9	1.0	− 3.9	0.2	0.139	0.615	0.583
	Spinal T-score	− 2.9	1.4	− 5.1	3.2	− 3.6	1.3	− 5.4	− 1.5	− 3.1	1.1	− 4.3	− 0.6	> 0.999	> 0.999	> 0.999
	Spinal Z-score	− 2.2	1.5	− 4.6	2.7	− 3.1	1.3	− 4.9	− 1.2	− 2.6	1.0	− 3.9	− 0.6	0.538	> 0.999	> 0.999
Fractures
	Age at first fracture	8.1	10.1	0.1	59.0	3.0	2.7	0.0	5.0	19.5	21.0	1.3	56.0	0.292	0.190	0.008
	Vertebral fractures	3.2	4.5	0.0	15.0	6.1	5.9	0.0	16.0	1.0	3.1	0.0	15.0	0.515	0.077	0.019
	Peripheral fractures	11.3	8.2	0.0	55.0	12.2	4.5	1.0	20.0	7.8	6.2	0.0	20.0	0.624	0.165	0.026
	Total fractures	14.2	9.2	0.0	55.0	17.9	6.5	8.0	30.0	8.6	7.1	0.0	30.0	0.164	0.007	< 0.001
Bone specific therapy
	None (%)	53.1				33.3				32.0				0.298	0.025	0.498
	Anti-resorptive therapy (%)	44.9				55.6				64.0				0.462	0.033	0.394
	Osteoanabolic therapy (%)	2.0				11.1				4.0				0.388	0.727	0.670

SD standard deviation, BMI body mass index, DXA dual-energy X-ray absorptiometry

No significant differences in age, BMI, femoral T- and Z-scores, and spinal T- and Z-scores were found between the different OI types I, III, and IV. OI type III patients differed significantly in height (p < 0.001) and weight from patients with OI types I and IV (p < 0.001 and p = 0.039, respectively). In addition, the age at first fracture was significantly lower in patients with OI type III (p = 0.008) compared to those with OI type IV. Moreover, OI type III patients had a higher number of vertebral and peripheral fractures compared to patients with OI type IV (vertebral fractures: p = 0.019, peripheral fractures: p = 0.026) (Table 1 and Suppl. Fig. 1). No significant differences in fracture-related parameters were observed between OI type I and OI types III/IV patients.

Based on the Sillence classification, 76 patients were diagnosed with OI type I, 14 with type III, and 25 with type IV. Genetic stratification identified 71 patients with a COL1A1 variant and 44 with a COL1A2 variant. Furthermore, we classified the cohort according to qualitative or quantitative collagen type I impairment: 46 patients were classified as having a quantitative defect and 58 as having a qualitative defect; 11 patients could not be classified because the potential defect could not be predicted. The distribution of the clinical OI types within genetic subgroups is shown in Fig. 1.

Fig. 1.

Distribution of OI types according to genetic stratification. The distribution was based on COL1A1 (a) and COL1A2 (b) as well as quantitative (c) and qualitative (d) variants. OI type I is highlighted in blue, OI type III in red, and OI type IV in orange

After stratifying the study cohort according to the genotypes COL1A1 and COL1A2 variants or by quantitative and qualitative variants, no significant differences were observed in demographics, T- and Z-scores as well as fracture parameters (Table 2).

Table 2.

Overview of adult OI patients after genetic stratification. The subgroups were divided into patients with COL1A1 vs. COL1A2 variantsand quantitative vs. qualitative variants. Total fractures were further classified into vertebral and peripheral fractures with mean values provided

			COL1A1 (n = 71)				COL1A2 (n = 44)					Quantitative (n = 46)				Qualitative (n = 58)
Parameter			Mean	SD	Min	Max	Mean	SD	Min	Max	p-value	Mean	SD	Min	Max	Mean	SD	Min	Max	p-value
Demographics
	Sex (male/female)		31/40				19/25				0.960	20/26				27/33				0.969
	Age (years)		45.3	15.8	18.0	82.0	40.9	14.6	19.0	69.0	0.167	42.5	15.7	18.0	76.0	44.4	15.7	19.0	82.0	0.536
	Weight (kg)		68.9	15.0	39.0	127.0	63.5	25.9	18.0	111.0	0.164	70.5	16.0	50.0	127.0	64.7	22.9	18.0	111.0	0.155
	Height (m)		1.62	14.5	1.21	1.89	1,51	0.28	0.80	1.82	0.347	1.64	0.12	1.38	1.89	1.54	0.26	0.80	1.85	0.301
	BMI (kg/m2)		26.4	5.0	19.3	47.2	26.9	7.2	17.3	45.1	0.785	26.2	5.3	19.3	47.2	26.7	6.0	17.3	41.8	0.838
DXA
	Femoral T-score		−2.2	1.1	−4.9	0.0	−1.7	1.5	−3.8	3.6	0.157	−2.2	1.1	−4.9	0.0	−1.9	1.4	−3.8	3.6	0.602
	Femoral Z-score		−1.7	1.1	−4.5	0.7	−1.5	1.4	−3.7	3.5	0.626	−1.6	0.9	−3.3	0.2	−1.5	1.2	−3.6	3.5	0.798
	Spinal T-score		−3.0	1.4	−5.4	3.2	−2.9	1.4	−5.1	0.8	0.888	−2.8	1.5	−4.3	3.2	−3.0	1.2	−5.1	0.8	0.840
	Spinal Z-score		−2.2	1.5	−4.9	2.7	−2.5	1.3	−4.4	0.6	0.420	−2.2	1.5	−4.1	2.7	−2.2	1.5	−4.6	1.9	0.929
Fractures
	Age at first fracture		11.8	16.6	0.1	59.0	9.3	10.9	0.0	51.0	0.653	7.4	9.7	0.1	49.0	12.2	15.9	0.0	59.0	0.284
	Vertebral fractures		3.4	4.9	0.0	15.0	2.5	4.1	0.0	15.0	0.256	3.5	5.0	0.0	15.0	2.7	4.3	0.0	15.0	0.242
	Peripheral fractures		11.4	8.3	0.0	55.0	9.6	6.3	0.0	25.0	0.421	10.3	6.5	0.0	31.0	10.9	8.6	0.0	55.0	0.738
	Total fractures		14.3	9.6	0.0	55.0	12.0	7.5	0.0	30.0	0.285	13.4	8.3	0.0	31.0	13.5	9.7	0.0	55.0	0.884
Bone specific therapy
	None (%)		47.9				54.5				0.488	60.9				44.8				0.104
	Anti-resorptive therapy (%)		49.3				40.9				0.381	37.0				51.7				0.133
	Osteoanabolic therapy (%)		2.8				4.5				0.623	2.2				3.4				0.700

SD standard deviation, BMI body mass index, DXA dual-energy X-ray absorptiometry

Regarding bone active medication, it was found that such medication was more frequently given to patients with OI types III and IV, whereas the difference only reached the level of significance between OI types I and IV (Table 1). After genetic stratification, there were no significant differences in the frequency of bone active drugs (Table 2).

Biochemical parameters in OI (with and without vitamin D insufficiency)

After categorizing the total cohort into vitamin D insufficient (25(OH)D < 30 µg/l) and vitamin D sufficient (25(OH)D ≥ 30 µg/l), it was found that 60.9% of all patients were vitamin D insufficient. For the vitamin D insufficient group, significantly higher levels of ALP (p = 0.018) and b-ALP (p = 0.003) were observed (Fig. 2).

Fig. 2.

Comparison of laboratory parameters between vitamin D (25(OH)D) insufficient vs. sufficient group. The vitamin D insufficient group (< 30 µg/l) is marked in yellow while the vitamin D sufficient group (≥ 30 µg/l) is marked in green. Reference values from the local laboratory are shaded in gray. PTH, parathyroid hormone; ALP, alkaline phosphatase; b-ALP, bone specific alkaline phosphatase; DPD/Crea, deoxypyridinoline per creatinine in the urine. Significant differences in the group comparisons are indicated by exact p-values

When comparing the laboratory parameters of OI types I, III, and IV, it was found that OI type IV showed significantly higher 25(OH)D levels than OI type I (p = 0.032) (Suppl. Table 2). In terms of bone resorption, OI type III patients showed significantly elevated DPD/Crea values compared to OI type I (Suppl. Table 2). There were no significant differences in biochemical parameters between the subgroups of patients with COL1A1 vs. COL1A2 variants or in patients with quantitative vs. qualitative variants (Suppl. Table 3).

HR-pQCT bone microstructure stratified according to Sillence classification system

OI type I (n = 67) patients exhibited significant higher Ct.BMD in the distal radius compared to OI type III (n = 7) (p = 0.007). No other significant differences in bone microstructure were observed among OI types I (n = 67/66), III (n = 7/4), and IV (n = 18/17) at either the distal radius or distal tibia (Suppl. Table 2).

HR-pQCT bone microstructure comparison between COL1A1 vs. COL1A2

Patients with COL1A1 variants revealed a reduced trabecular microstructure compared to those with COL1A2 variants at both the distal radius and tibia (Fig. 3a). At the distal radius, both trabecular density and microstructure differed significantly between patients with COL1A1 (n = 59) and COL1A2 (n = 33) variants. Specifically, Tb.BMD, BV/TV, and Tb.N were significantly lower, while Tb.Sp. was significantly higher in the COL1A1 group (radius: Tb.BMD: p = 0.008; BV/TV: p = 0.010; Tb.N: p = 0.023; Tb.Sp: p = 0.020) (Fig. 3b and Suppl. Table 3). A similar pattern was observed at the distal tibia, where patients with COL1A1 (n = 58) variants again showed significantly lower trabecular parameters compared to those with COL1A2 (n = 29) variants (tibia: Tb.BMD: p = 0.015; BV/TV: p = 0.028; Tb.N: p = 0.014; Tb.Sp: p = 0.012) (Fig. 3b and Suppl. Table 3).

Fig. 3.

Differences in bone microarchitecture according to genotype. Reduced trabecular microstructure was detected in both distal radius and tibia of patients carrying COL1A1 or quantitative variants (a). Comparison of genetic subgroups with respect to patients with COL1A1 or COL1A2 variants (b) and patients with quantitative or qualitative variants (c). HR-pQCT results of the distal radius and tibia were compared to the median of device-, age- and sex-specific reference values (XCT1: S Hansen, et al. [16], XCT2: D.E. Whittier, et al. [17]). Ct.BMD, cortical BMD; Ct.Ar, cortical area; Tb.BMD, trabecular BMD; Tb.N, trabecular number; BV/TV, bone volume to tissue volume. Significant differences in the group comparisons are characterized by exact p-values

HR-pQCT bone microstructure comparison between quantitative vs. qualitative variants

At the distal radius, bone structure assessment revealed significant reduction only in trabecular number in patients with quantitative variants (n = 38) compared to patients with qualitative variants (n = 46) (radius: Tb.N: p = 0.025) (Fig. 3c and Suppl. Table 3).

When analyzing the bone microstructure at the distal tibia, no statistically significant differences were found between the quantitative group (n = 37) and the qualitative group (n = 40) in any of the assessed parameters (Fig. 3c and Suppl. Table 3).

Discussion

In classical OI, pathogenic variants of COL1A1 or COL1A2 impair collagen type I synthesis, which among other things results in brittle bone. Collagen type I forms a triple-helical structure composed of two alpha-1 chains (COL1A1) and one alpha-2 chain (COL1A2) and constitutes the primary structural component of the extracellular bone matrix [1]. Pathogenic variants in the COL1A1/2 genes lead to either a quantitative deficiency typically due to nonsense or frameshift mutations causing haploinsufficiency or a qualitative abnormality, often resulting from missense mutations that substitute glycine residues, which are crucial for the stability of the triple-helical structure of collagen type I [24]. The aim of the present study was to identify bone microstructural patterns in adult patients with classical OI after genetic stratification in both (1) COL1A1 vs. COL1A2 variants and (2) quantitative vs. qualitative COL1A1/2 variants. In addition, we investigated adult OI patients with vitamin D insufficiency vs. sufficiency.

Our findings show that vitamin D insufficiency, present in 60.9% of our cohort, is associated with significantly elevated ALP and b-ALP levels, which may suggest altered bone mineralization [25]. However, vitamin D status is only presented for a single time point at first visit, which is a limitation, as vitamin D status can fluctuate over time. Therefore, a single measurement may not fully capture long-term vitamin D status, deficiency, and/or osteomalacia. Although OI is clearly characterized by collagen type I defects, our data highlights the need of addressing vitamin D insufficiency in OI management to avoid additional mineralization defects of the brittle bone.

In our cohort, patients with OI types I and IV were significantly taller and showed a higher body weight compared to OI type III patients. This aligns with the established classification of OI types based on clinically characteristics by Sillence [4, 12, 26]. Patients with classical OI exhibited reduced bone mineral density within the osteoporotic range. Among the clinical subtypes, OI type III was associated with the lowest T- and Z-scores. However, this trend did not reach statistical significance.

HR-pQCT serves as a different diagnostic approach alongside traditional two-dimensional bone density measurements, offering detailed assessment of the three-dimensional bone architecture, including bone mineral density as well as cortical and trabecular bone parameters [27, 28]. Considering bone microstructure, patients with OI type I in our cohort exhibited significantly thicker cortical bone in the distal radius when compared to OI type III patients. Nonetheless, clinical application of HR-pQCT in OI remains limited due to bone deformities, prior surgeries (implants), fractures, or limitations related to body size [29]. Analyses of available bone microstructure data assessed by HR-pQCT confirmed reduced bone density, trabecular number, trabecular thickness, and cortical thickness in OI type I patients compared to healthy controls [30]. Similar results were shown for OI types III and IV [29, 31]. Currently, only limited studies are available investigating bone microstructure in adult OI patients with regard to the genotype [6, 12], and the genotype–phenotype correlation of patients with classical OI remains a topic of ongoing discussion [32–34].

In our cohort, genetically stratified subgroups did not differ in demographic, bone density, and fracture parameters. However, following genetic stratification, we identified significant microstructural differences between the patients carrying COL1A1 vs. COL1A2 variants as well as in patients with quantitative vs. qualitative COL1A1/2 variants. Previous studies predominantly compared adult patients with classical OI to healthy controls or individuals with early-onset osteoporosis (EOOP) and/or focused on analyses distinguishing quantitative from qualitative variants [6, 7]. As expected, adult patients diagnosed with classical OI show impaired trabecular bone microstructure in comparison to healthy controls, though not significantly different from EOOP patients [7]. In addition, individuals with quantitative COL1A1/2 variants exhibit significantly reduced trabecular parameters in the distal radius compared to patients with qualitative variants, while reductions in the tibia were less pronounced [6]. In our cohort, we could show that the COL1A1 group exhibited lower trabecular density and microstructure in both the distal radius and tibia (Tb.BMD, BV/TV, Tb.N, Tb.Sp) compared to the COL1A2 group. These findings suggest that genetic alteration in the two alpha-1 chains of collagen may lead to more pronounced microstructural deterioration than variants affecting the single alpha-2 chain.

Among the analyzed parameters, only trabecular number in the distal radius was significantly reduced in patients with quantitative compared to qualitative variants. Hald et al. found comparable results regarding bone microstructure in 85 adult OI patients with quantitative vs. qualitative variants, mainly showing a significant decrease in trabecular microstructure in the quantitative group predominant in the distal radius [6]. Despite this microstructural difference between the subgroups of our cohort, no significant differences in fracture frequency were detected.

While these classifications provide a useful instrument for genotype–phenotype correlation, it is important to acknowledge their limitations. Additional mechanisms of NMD escape may exist, and truncated proteins can still be produced at appreciable levels [35]. As such, qualitative effects may occur even in cases predicted to undergo NMD, and bioinformatic predictions alone by tools like AutoPVS1 may not adequately capture these complexities [18, 19]. Furthermore, evidence suggests that the so-called last exon/50–55 nucleotide rule does not universally apply to all C-terminal variants, as some of these have been associated with milder phenotypes typically attributed to quantitative effects [34]. Functional studies remain essential for definitive assessment of variant impact.

Our results emphasize that although HR-pQCT can detect trabecular and cortical deterioration in association with variant class, it does not fully reflect qualitative bone characteristics such as material composition, bone deformities, or fracture susceptibility in OI.

More importantly, detected microstructural differences cannot be directly translated to clinical severity of the disease and is, at this stage, rather descriptive. This is a commonality with the Sillence classification system, which is also not intended as a classification of severity. Therefore, bone microstructure alone may underestimate the severity of bone fragility in these patients. In this context, finite element analysis (FEA) has been proposed to assess intrinsic bone quality and predict mechanical performance. Thus, the application of FEA to HR-pQCT data may improve the understanding of bone fragility in patients with OI.

We are aware of the strengths and limitations of our study. The strength of our study is that a large cohort of adult OI patients has been examined by many osteologic parameters including biochemical, bone density (DXA), and bone microstructure (HR-pQCT) parameters. To our knowledge, no study has been performed to determine differences in microstructure in adult patients with classic OI matched for COL1A1 vs. COL1A2 variants. However, the patient number is still low for specific subgroups, in particular for OI type III patients. Additionally, there are incomplete data sets, in particular for OI type III patients, since arm-/leg-length or deformity or surgical implants impeded with respective analyses. Moreover, the measurements analyzed here were done in clinical routine following standard procedures and were, for example, not adjusted for body height by specific protocols. Finally, this retrospective cross-sectional study can per se only identify associations but not establish causality.

Conclusion

In conclusion, it can be stated that after genetic stratification, regardless of the Sillence classification, differences in the bone microstructure, in particular of the trabecular compartment, become evident between COL1A1 and COL1A2 as well as quantitative and qualitative COL1A1/2 variants. This indicates the importance of the respective variant and emphasizes its consideration in classical OI patients. Further research is essential to refine diagnostic methods, enhance treatment strategies, and deepen our understanding of disease progression. Future studies may aim to combine bone microstructure analysis with FEA-based mechanical performance in order to allow better determination of clinical severity and to deepen our understanding of genotype–phenotype correlations in OI.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Open Access funding enabled and organized by Projekt DEAL. Grant supporters: This project has received funding from the Deutsche Forschungsgemeinschaft (DFG) within the Clinical Research Unit CRU 5029 (ProBone) (Project no. 517063424).

Data availability

The data of the study are not publicly accessible for reasons of sensitivity but can be requested from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This retrospective study was performed in accordance with the local ethical guidelines (PV5364) and the Declaration of Helsinki. All patients gave written informed consent for retrospective analysis of the derived data from clinical routing measurements during their visit at our outpatients’ clinic.

Conflict of interest

RO has served as a speaker and advisory board member for Kyowa Kirin, Inozyme, Ipsen, Mereo, Pharmacosmos and UCB and has received institutional research grants from Kyowa Kirin, UCB and Inozyme. FB has received speaker fees from Alexion, UCB, and Diasorin and has received institutional research grants from UCB and Alexion. All other authors declare that there is no conflict of interest.