Criteria for osteoporosis diagnosis: a systematic review and meta-analysis of osteoporosis diagnostic studies with DXA and QCT

Summary

Background

This study aims to evaluate the difference between dual-energy X-ray absorptiometry (DXA) vs. quantitative computed tomography (QCT) in the diagnosis of osteoporosis (OP), and to discuss the impact of this difference on OP clinical practice, as well as the potential reasons for the disparities.

Methods

In this systematic review and meta-analysis (PROSPERO ID: CRD42024599632), we searched Medline, Embase, Scopus, Web of Science, and Cochrane Library databases for studies published from the inception to September 14, 2024. Studies using both QCT and DXA to diagnose OP in the same population were included, and 2 independent reviewers extracted the data by following the PRISMA statement. Pooled odds ratios (ORs) with 95% confidence interval (CI) were estimated using random effects model under heterogeneity.

Findings

A total of 19 studies with 3939 cases were included. The meta-analysis results indicated that QCT identified significantly more OP patients than DXA in the same population (OR: 4.91, 95% CI: 3.19–7.54; p < 0.0001). Subgroup analysis revealed that the higher diagnostic rate of QCT over DXA in diagnosing OP is significantly stronger in males (OR: 8.45, 95% CI: 3.80–18.77; p < 0.0001) than in females (OR: 2.11, 95% CI: 1.53–2.90; p < 0.0001). Furthermore, among the population aged ≥65 years (OR: 6.01, 95% CI: 3.45–10.47; p < 0.0001), the differences between QCT and DXA in diagnosing OP were significantly greater than those in the population aged <65 years (OR: 2.27, 95% CI: 1.55–3.33; p < 0.0001).

Interpretation

The incidence of OP examined by QCT is significantly higher than that of DXA. The different prevalence rates obtained from the two diagnostic techniques will inevitably complicate the prevention and treatment of OP, as well as the selection of cases in drug clinical trials. This study highlighted this confusion and aims to motivate relevant organizations or institutions to address this challenge.

Funding

There was no funding for this study.

Research in context.

Evidence before this study

The BMD measured by DXA is the recognized gold standard for diagnosing osteoporosis (OP). In recent years, another effective and alternative radiographic method, QCT, has also been used to determine BMD for the evaluation of OP. Many studies have found significant differences in the prevalence of OP identified by QCT compared to that identified by DXA. However, it is still uncertain whether this discrepancy is due to an underestimation of OP by DXA or an overestimation by QCT. Herein, we searched for all studies comparing the prevalence of osteoporosis in the same population using QCT and DXA that were indexed in the Medline, Embase, Scopus, Web of Science, or Cochrane Library databases up to September 14, 2024. We conducted a systematic review to compare the diagnostic performance of QCT and DXA for osteoporosis in the same population and analyzed the possible reasons for the observed differences.

Added value of this study

Our meta-analysis findings further emphasized that the number of OP cases identified by QCT is significantly higher than that detected by DXA in the same population, and indicated that the magnitude of this difference is related to age and gender. The certainty of the evidence produced in our review, along with the results of sensitivity analyses, supports the robustness of the higher prevalence of OP detected by QCT compared to DXA.

Implications of all the available evidence

There exists a significant discrepancy between QCT and DXA in the diagnosis of OP, and this inconsistency may pose substantial challenges to clinical practice. We highlight this difference and discuss the possible reasons for it, aiming to raise awareness among relevant organizations and institutions to address this challenge.

Introduction

Osteoporosis (OP) is a systemic metabolic bone disease characterized by decreased bone mass and bone microstructure disruption, which increases bone fragility, thereby increasing the risk of fracture.¹ The most serious OP outcome is osteoporotic fracture, also known as fragility fractures. The estimated risk of osteoporotic fractures in individuals aged ≥50 years was 158 million in 2010, and it will double by 2040.² In 2013, the International Osteoporosis Foundation (IOF) reported that an osteoporotic fracture occurs every 3 s worldwide.³ Fragility fractures, especially hip fractures, are a major global health concern and are associated with high disability and mortality.⁴ The European Union report on OP estimated that the mortality rate of hip fragility fracture was higher than road accident and was equivalent to breast cancer.⁵ Furthermore, a European survey estimated of 21 disability-adjusted life years (DALYs) per 1000 people aged ≥50, significantly higher than the estimates for stroke or COPD.⁶ Fragility fractures substantially increase the economic burden on both patients and society. A systematic review found that the cost of hip fracture in the first year ($43669) was higher than that of acute coronary syndrome ($32345) and ischemic stroke ($34772).⁷ Burge et al.⁸ revealed that in the United States, total expenditure on fragility fractures was $17 billion in 2005, which is expected to increase to $25 billion by 2025. The total cost of fragility fractures in 27 European Union countries was 98 billion euros in 2010 and is expected to reach 120 billion euros by 2025.⁵ Fragility fractures can be prevented by early detection, diagnosis, and treatment of OP. Bone mineral density (BMD) is a crucial indicator of OP, which can predict fracture risk and monitor late-stage outcomes.

Currently, the World Health Organization (WHO) has recommended Dual-energy X-ray absorptiometry (DXA) as the gold standard technique for diagnosing OP.⁹ DXA employs a posterior-anterior projection method to assess trabecular and cortical bone structures, and measures BMD in g/cm². In DXA, OP diagnosis is indicated by the BMD value that is ≥ 2.5 SD below the young adult mean value (T-score ≤ −2.5), and osteopenia is identified by the BMD value that lies between 1 and 2.5 SD below the young adult mean value (−2.5 < T-score < −1).¹⁰ Another valid and alternative radiologic method for determining BMD is Quantitative computed tomography (QCT),^11,12 which quantifies the volume of trabecular BMD, expressed in g/cm³.¹³ The International Society for Clinical Densitometry (ISCD) and the American Society of Radiology (ACR) recommend that vertebral BMD should be measured by QCT, where an absolute value ≤ 80 mg/cm³ indicates OP and a value between 80 and 120 mg/cm³ indicates osteopenia.^14,15

Currently, several studies including cross-sectional studies and case–control studies have compared the diagnostic performance of QCT and DXA for vertebral OP; however, the results are not consistent.16, 17, 18 Therefore, selecting the best diagnostic technique for the same group or individual is challenging. It remains unclear whether this is due to underestimation by DXA or overestimation by QCT. Therefore, this study discusses the difference in OP diagnostic rates between QCT and DXA by Meta-analysis and investigates the clinical influences on this difference.

Methods

Literature search

We prospectively registered this review on PROSPERO (CRD42024599632) and have followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.^19,20 We searched Medline, Embase, Scopus, Web of Science, and Cochrane Library databases from inception to September 14, 2024. The Medical Subject Headings (Mesh) and keywords used for screening are presented in Appendix 1. In addition, the reference lists of articles and reviews were hand-searched to trace relevant articles to supplement the electronic query. All articles did not restrict by language, geography, or publication date. The titles, abstracts, and full texts were screened by 2 investigators (JY and WY) independently.

Eligibility criteria

The included criteria included articles in which (1) all the participants received both QCT and DXA tests of the lumbar spine within a year, (2) the DXA diagnostic criteria for OP followed the 1994 WHO recommendations: peak BMD in the normal population as the criterion, with a T-score ≤ −2.5,²¹ (3) the QCT diagnostic criteria for OP followed the recommendations of the International Society for Clinical Bone Density (ISCD) and the American College of Radiology (ACR), i.e., lumbar BMD <80 mg/cm^3,14; (4) the detection rate of OP could be extracted directly or calculated based on raw DXA and QCT lumbar BMD data from the literature, (5) Newcastle–Ottawa Scale (NOS) score ≥6. The exclusion criteria were as follows: (1) reviews, letters, editorials, case reports, meta-analyses, and animal studies, (2) the study population was duplicated in another study included in meta-analysis; when studies of duplicate populations were identified, the study that included more patients was selected.

Data extraction

Study variables were extracted by two authors (JY and YZ) independently, and in case of any disagreement, a third author (WY) was adjudicated. Extracted information included the first author’s name, publication year, country, race, number of subjects, demographic characteristics (age and sex), and the number of patients with lumbar OP screened by QCT and DXA, respectively. The judgment of race was based on the racial classification system proposed by Blumenbach, including Caucasian, Mongolian, Ethiopian, American, and Malay.²²

Quality assessment

The quality of each study was evaluated by two authors (JY and YZ) independently according to the Newcastle–Ottawa Scale (NOS) for assessing the quality of nonrandomized studies in meta-analyses.²³ The scale is used to assess research biases and assigns scores based on nine items within three areas: appropriate selection of participants, appropriate measurement of exposure, and appropriate control of confounding factors. Based on the characteristics of the included studies, we developed scoring criteria for each item in NOS (Appendix 2). As accepted in the literature,^24,25 if the NOS scale score is at least 7, the quality of the study is determined to be high. Otherwise, the quality of the study is considered to be low.

Statistical analysis

Review Manager 5.3 software (Cochrane Community, UK) was used for Meta-analysis. Concerning the dichotomous data, the odds ratio (OR) with a 95% CI was calculated using the Mantel-Haenszel model.

Data were pooled using a random effects model (REM) to allow for differences between studies. The selection of random effects models is based on the heterogeneity of devices, populations, and technologies identified in the included studies. Using I² statistics to evaluate heterogeneity in the included studies, I² values of 25%, 50%, and 75% were considered low, moderate, and high heterogeneity, respectively.²⁶ Pooled ORs and 95% CIs were estimated to compare the diagnostic differences in OP prevalence between QCT and DXA. To determine publication bias, Begg’s and Egger’s tests were performed using Stata 18.0 software. Publication bias was considered absent only if both tests yielded p > 0.05. To assess the robustness of the main findings, the data were pooled after serially excluding each study included in the main analysis. To identify the sources of heterogeneity, a meta-regression analysis was conducted to examine the effects of race, gender, and age on the ability of QCT and DXA to identify OP. If p < 0.05, it is considered that the factor has a significant effect, and further subgroup analysis was performed on this factor.

Ethics statement

Not applicable since this is a systematic review and meta-analysis of publicly available data.

Role of funding source

There was no funding for this study.

Results

Study selection and study characteristics

The initial search based on the established search strategy yielded 2321 citations, of which 1751 remained after duplicate removal. Moreover, after screening the title and abstract, irrelevant literature was excluded, which left 79 studies. Then, a full-text review was performed, after which 24 articles were remained. A final detailed review based on the inclusion and exclusion criteria found 1 article lacked DXA lumbar detection data for OP,²⁷ 1 study had a different population who underwent DXA and QCT test,¹⁷ 1 article did not elucidate the QCT criteria for OP diagnosis,²⁸ and 1 of the two studies was excluded because of possible duplication of subject populations.^16,29 Using the NOS scale, the quality of the remaining 20 studies was assessed (Appendix 3). One of the studies³⁰ was excluded due to a NOS score <7. Finally, 19 studies were selected for analysis. The original search review process for the final article selection was depicted by the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) flow diagram (Fig. 1).³¹

Fig. 1.

PRISMA flow diagram.

The basic characteristics of the included studies are shown in Table 1. Total sample size of the 19 studies was 3939 cases, including 2106 females and 1833 males. Among 3939 individuals, QCT identified 2044 patients with OP (51.89%), and DXA identified 833 patients with OP (21.15%).

Table 1.

The basic characteristic of the included studies.

	Year	Country	Race	Subjects (n)	Age (years)	Sex (M/F)	OP patients, n (%)
							QCT	DXA
Lange U et al.32	2005	Germany	Caucasian	84	43.9 ± 8.4	53/31	36 (42.9)	11 (13.1)
Li X et al.33	2012	China	Mongolian	59	78.7 ± 8.4	59/0	18 (30.5)	5 (8.5)
Li N et al.34	2013	China	Mongolian	140	63.2 ± 8.1	0/140	65 (46.4)	24 (17.1)
Zakharov IS et al.35	2015	Russian	Caucasian	73	50–59	0/210	15 (20.5)	11 (15.1)
				58	60–69		26 (44.8)	14 (24.1)
				53	70–79		31 (58.8)	22 (41.5)
				26	≥80		20 (76.9)	12 (46.2)
Li K et al.36	2015	China	Mongolian	314	80.0 ± 7.0	314/0	141 (44.9)	35 (11.2)
Li K et al.37	2017	China	Mongolian	614	76.3 ± 26.0	314/300	352 (57.3)	71 (11.6)
Kim K et al.38	2018	Korea	Mongolian	81	58.0 (57–60)	0/81	25 (30.9)	17 (21.0)
Xu X et al.39	2019	China	Mongolian	313	79.6 ± 7.2	313/0	141 (45.0)	8 (2.6)
Xu D et al.40	2020	China	Mongolian	192	76.8 ± 8.8	0/192	132 (68.9)	72 (37.5)
Milisic L et al.41	2020	Bosnia-Herzegovina	Caucasian	44	59.6 ± 0.6	0/44	27 (61.4)	21 (47.7)
Kim K et al.42	2021	Korea	Mongolian	117	65.0 (63–68)	66/51	39 (33.3)	8 (6.8)
Yoon H et al.43	2021	Korea	Mongolian	59	70.76 ± 9.7	20/39	43 (72.9)	15 (25.4)
Yuan Y et al.44	2021	China	Mongolian	357	74.7 ± 16.1	357/0	144 (40.3)	12 (3.4)
Zhang P et al.45	2022	China	Mongolian	138	70.7 ± 12.3	138/0	27 (19.6)	4 (2.9)
Miao H et al.46	2022	China	Mongolian	148	63.7 ± 10.4	0/148	68 (45.9)	39 (26.4)
Lin W et al.29	2023	China	Mongolian	501	67.6 ± 10.4	106/0	47 (44.3)	27 (25.5)
					71.3 ± 7.2	0/395	257 (65.1)	227 (57.5)
Uppal R et al.47	2023	Pakistan	Caucasian	100	unreported	31/69	57 (57.0)	11 (11.0)
Hammood AH et al.48	2023	Iraq	Caucasian	116	45–64	0/164	55 (47.4)	48 (41.4)
				48	≥65		40 (83.3)	35 (72.9)
Boehm E et al.49	2024	Germany	Caucasian	304	71.5 ± 7.64	62/242	238 (78.3)	84 (27.6)
Total	NA	NA	NA	3939	NA	2106/1833	2044 (51.89)	833 (21.15)

M/F = Male/Female; OP = osteoporosis; QCT = quantitative computed tomography; DXA = dual-energy X-ray absorptiometry; NA = not applicable.

Main data analysis

Statistical analysis of the 19 included articles indicated significantly large heterogeneity between studies (I² = 93%), which was assessed by meta-analysis using REM (Fig. 2). The analysis results indicate that QCT has a significantly higher screening rate for OP compared to DXA (OR: 5.12, 95% CI: 3.30–7.96; p < 0.0001).

Fig. 2.

Forest plot for the odds ratio of the number of OP patients identified by QCT and DXA in the same population.

To investigate the sources of heterogeneity, we performed a meta-regression analysis on factors that may influence OP screening rates, including race (Caucasian and Mongolian), age, and sex. Since most studies only reported the average age or median age, we divided the studies into two categories based on the average or median age: <65 years and ≥65 years. Two studies reported OP screening rates for different age groups and were grouped according to the actual classifications.^35,48 One study was excluded due to not reporting age.⁴⁷ Since some studies did not report data separately for males and females, we categorized age as a three-level variable (female, male, and mixed) in the meta-regression analysis. The regression results revealed that race, and mixed gender (yes or not) did not have a significant impact on the pooled OR for identifying OP using QCT and DXA, whereas age, male (yes or not) and female (yes or not) had a significant effect (Appendix 4). Therefore, we conducted further subgroup analysis by age and gender.

Begg’s test and Egger’s test showed that the potential publication bias was negative because p = 0.97 on Begg’s test and p = 0.21 on Egger’s test (Appendix 5). The sensitivity analyses indicated that serial exclusion of studies has little impact on the pooled OR for identifying OP using QCT and DXA. Pooled ORs (range = 4.44–5.27) remained statistically significant irrespective of excluded studies (Appendix 6).

Subgroup data analysis by sex

Based on the extracted data, a total of 8 studies and 6 studies were included in the subgroup analysis for the female and male groups, respectively. Subgroup meta-analysis was conducted using REM. Although there is still high heterogeneity in both the female (I² = 72%) and male (I² = 88%) groups (Fig. 3), it has been reduced compared to the meta-analysis of all studies, indicating that sex is one of the factors contributing to the high heterogeneity.

Fig. 3.

Forest plot for the odds ratio of the number of OP patients identified by QCT and DXA in the male and female.

Among 1374 female subjects, QCT screened 761 (55.39%) and DXA identified 542 (39.45%) OP cases. In the male group, a total of 1287 participants were included, with QCT and DXA identifying 518 (40.2%) and 91 (7.10%) OP patients, respectively. Compared to DXA, QCT identified significantly more OP cases in both females (OR: 2.11, 95% CI: 1.53–2.90; p < 0.0001) and males (OR: 8.45, 95% CI: 3.80–18.77; p < 0.0001). Notably, compared to females, QCT demonstrated a significantly stronger capability than DXA in identifying OP in males (p < 0.05).

Subgroup data analysis by age

A subgroup analysis based on age was performed using a random effects model (REM). The results showed that the heterogeneity among studies with a population age ≥65 years (I² = 94%) was similar to that of all studies, while the heterogeneity among studies with a population age <65 years (I² = 59%) significantly decreased (Fig. 4), indicating that population age is one of the sources of heterogeneity.

Fig. 4.

Forest plot for the odds ratio of the number of OP patients identified by QCT and DXA among participants aged <65 and ≥65.

As shown in Fig. 4, a total of 686 cases with a mean or median age <65 years were included. Among them, QCT identified 293 (42.71%) OP patients, significantly more than the 171 (24.93%) identified by DXA (OR: 2.27, 95% CI: 1.55–3.33; p < 0.0001). In 3095 cases with a mean or median age of ≥65 years, QCT also screened more OP cases (n = 1663, 53.73%) than DXA (n = 637, 20.58%), and the difference was significant (OR: 6.01, 95% CI: 3.45–10.47; p < 0.0001). Furthermore, the difference in OP identification between QCT and DXA in the population with age ≥65 was significantly greater than that in the population with age <65 (p < 0.05).

Discussion

This meta-analysis revealed that QCT had a significantly higher OP detection rate than DXA. The subgroup analysis revealed that the difference in the identification of OP by QCT vs. DXA was notably greater in males than in females (OR: 8.45 vs. 2.11), which is similar to the results of two multicenter and large-sample size cohort studies of a Chinese population.^50,51 These two studies revealed that in people over 50 years of age, QCT screened 29.0% of women with OP, comparable to the 29.13% detected by DXA; for men, QCT detected 13.5% of patients with OP, obviously higher than the 6.46% screened by DXA. Herein, our meta-analysis indicated that QCT had a significantly higher detection rate for OP compared to DXA in women (55.39% vs. 39.45%) and men (40.20% vs. 7.10%). It is evident that our review showed a higher proportion of OP patients identified by QCT compared to DXA, and there are also greater differences in the diagnosis of OP prevalence between the two imaging techniques, regardless of gender. The substantial discrepancy between our findings and previous studies may be attributed to the fact that the average ages of the populations studied by Cheng et al.⁵⁰ and Zeng et al.⁵¹ were 49.7 and 51.8, respectively, while in our analysis, the proportion of the population aged ≥65 exceeds 70%. It is well known that age is a significant risk factor for OP. Importantly, our analysis indicated that the differences in the identification of OP patients by QCT and DXA were significantly greater in the older age group (≥65) compared to the younger age group (<65).

Appropriate OP diagnosis can prevent unnecessary non-OP treatments and costs, or it can avoid underestimating OP, allowing patients to start OP treatment early to prevent bone deterioration and the subsequent risk of fragile fractures. However, our meta-analysis indicated that the prevalence of OP diagnosed by QCT was significantly higher than that diagnosed by DXA in the same population. The inconsistency in OP prevalence rates derived from these two methods will inevitably confuse clinical and research institutions globally when screening for OP patients. The results obtained from the use of different radiographic techniques should be carefully considered. If QCT results are considered, the OP population may be overestimated, resulting in unnecessary medical treatment and costs. If DXA is considered, OP patients who should receive treatment may be overlooked or treated with delay, which increases the risk of fractures and disabilities. As highlighted in the Rotterdam study,⁵² many fragility fractures occur in individuals classified as osteopenic by DXA, underscoring the need for diagnostic methods with higher sensitivity. While QCT’s lower specificity may raise concerns about overdiagnosis, its superior fracture prediction AUC suggests clinical utility in high-risk populations.⁴⁹

There are several potential reasons why QCT has a higher ability to identify OP compared to DXA. First, conditions such as osteophytes, ligament calcification, spinal degeneration, or abdominal aortic calcification in the lumbar region can lead to an overestimation of BMD values, which in turn results in an underestimation of OP by lumbar DXA.^{29,34,38,39,42,43} Secondly, DXA measures 2D projections of cortical and cancellous bone, whereas QCT measures 3D BMD of cancellous bone, and since cancellous bone has a faster age-related loss than cortical bone, this may decrease the sensitivity of DXA in recognizing OP.^16,39 Third, many elderly individuals have central obesity, which increases the distance from the X-ray source to the fan beam detector, thereby increasing the projected bone area and leading to an overestimation of lumbar BMD by DXA.^41,53,54 However, these variabilities are insufficient to support the differences in OP detection rates between the two measures. We propose that the primary reason for the differences between QCT and DXA is the QCT diagnostic criteria for OP (lumbar BMD <80 mg/cm³), which does not equate to a T-scores < −2.5 measured by DXA. If they are equivalent, then the same criteria applied to the same population should yield the same prevalence rate of OP. Our meta-analysis indicated inconsistent results, suggesting that the currently recommended QCT criteria are not equivalent to the DXA criteria. Indeed, opinion from the 2007 ISCD Official Positions that the DXA T-score is −2.5 when the QCT T-score is −3.4 according to the reference data published by the manufacturer (Siemens).¹⁴ Furthermore, they also revealed that a QCT-measured spinal trabecular BMD of 80 mg/cm³ represents a T-score of −2.9 according to the German reference population. Therefore, the actual DXA T-score for a spinal BMD of 80 mg/cm³ measured by QCT is < −2.5, which likely resulted in an overestimation of OP.

The WHO recommends DXA as the diagnostic criteria for OP, and it has clear origins and is accepted worldwide.⁵⁵ The QCT diagnostic criteria for OP are recommended by ISCD and ACR.^14,15 However, we reviewed previous publications and found that the origin of the QCT diagnostic criteria remains unclear. A diagnostic cut point of 80 mg/cm³ for QCT to identify OP was first reported by a German study in 1999⁵⁶; however, the authors did not state how this reference value was obtained and only explained that it was proposed based on material-related fracture risk. The 2007 ISCD official position subsequently referred to this value, but it remains unclear how it was determined or calculated, only that the above German-language article is cited.¹⁴ In 2008, the ACR practice guideline explicitly recommended 80 mg/cm³ as the cutoff point of QCT for diagnosing OP, but they also did not clarify the source of this reference value.^15,57 Although the origin of 80 mg/cm³ threshold is ambiguous, it has been proven to be useful and stronger than DXA in fracture prediction.^58,59 However, its origin in Western populations may limit generalizability to other groups. Recent studies suggested that lower thresholds (e.g., 50 mg/cm³) may improve specificity for vertebral fractures in East Asian populations.^60,61 Thus, future efforts should focus on establishing QCT-specific thresholds that better align with fracture risk predictions across diverse populations.

We suggest that caution should be exercised when using a QCT value of 80 mg/cm³ for the diagnosis of osteoporosis (OP), or at least that this value may not be suitable for East Asian elderly women, as it is based on Western populations. Recently, Wáng et al.^60,61 indicated that for Chinese and Japanese women, if QCT-measured lumbar BMD of 80 mg/cm³ is used for diagnosing OP, the specificity of identifying vertebral fragility fractures related to the OP is too low, whereas setting the threshold at 45–50 mg/cm³ can identify women with and without vertebral fragility fractures as reported in Caucasian women. Therefore, the authors concluded that the QCT cut-off point for OP diagnosis in older East Asian women should be close to and not exceed 50 mg/cm³ lumbar BMD. In older Chinese men (≥50 years of age), the QCT cut-off point for lumbar OP diagnosis was also recommended to be 45–50 mg/cm³, similar to older Chinese women.⁶² Therefore, we propose that for certain populations (such as East Asian women) and clinical scenarios requiring high specificity (such as surgical decision-making), a lower QCT threshold (such as<50 mg/cm³) can be defined as “severe osteoporosis”. This approach balances sensitivity and specificity while addressing population-specific needs. However, our meta-analysis found that race did not affect the ability of QCT and DXA to identify OP. Although this conclusion contradicts the findings of Wáng et al., it is in fact reasonable. Wáng et al.60, 61, 62 assessed the differences between QCT and DXA in OP diagnosis based on fracture risk assessment, whereas our analysis was based on the diagnosis of spinal BMD. In other words, there may be no race differences in evaluating the diagnostic discrepancies between QCT and DXA for OP through spinal BMD assessment.

Although our meta-regression found no significant racial differences in diagnostic discordance, it should be noted that 13/19 included studies were conducted in East Asian populations (10 Chinese, 3 Korean), potentially limiting generalizability to other ethnic groups. Regional variations in body composition, skeletal geometry, and fracture risk patterns may influence the clinical implications of QCT-DXA discordance. For instance, lower BMI distributions in Asian populations compared to Western counterparts could amplify QCT's sensitivity to detect osteoporosis.⁴¹ Future multinational studies should prioritize collecting stratified data on ethnicity, geographic location, and population-specific fracture risk profiles to establish whether regionally adapted diagnostic thresholds are warranted. Additionally, comparative effectiveness research across diverse healthcare systems could elucidate how diagnostic discrepancies impact treatment patterns and fracture outcomes in different geographical contexts.

Beyond diagnostic accuracy, practical considerations such as cost and radiation dose influence clinical adoption. DXA remains the most cost-effective modality for OP screening compared to other techniques. However, higher sensitivity of QCT in identifying OP patients may offset costs by preventing fractures in undiagnosed individuals.⁶³ Radiation exposure also differs substantially: DXA delivers a negligible effective dose of <10 μSv,⁶⁴ whereas QCT exposes patients to 100–300 μSv—approximately 10–30 times higher than DXA but still lower than routine CT scans (3–5 mSv).¹⁴ Clinicians must weigh QCT’s diagnostic advantages against its higher radiation burden, particularly in younger patients or those requiring serial monitoring. Advances in low-dose QCT protocols and opportunistic screening (e.g., using abdominal CT scans performed for other indications) could mitigate these concerns.⁶⁵

Present review has some limitations: (1) the diagnosis of OP was based solely on lumbar BMD, and the detection rate does not represent the actual prevalence of OP in the population; (2) the subgroup analyses conducted based on the average or median age of populations from different studies do not truly reveal the actual impact of age on the differences in identifying OP between the two radiographic methods; (3) different QCT or DXA models were not categorized in this analysis, which may affect BMD measurements and increase heterogeneity between studies.

In conclusion, our meta-analysis highlighted that the prevalence of OP detected by QCT was significantly higher than that detected by DXA. This is a global phenomenon which substantially complicates the clinical diagnosis and treatment of OP. Moreover, we discussed the possible reasons for this discrepancy and found that the source of the diagnostic criteria for QCT in diagnosing OP is unclear and is not equivalent to the DXA criteria for diagnosing OP. Therefore, these two technologies are not comparable in their identification of OP. In the future, QCT criteria should be reassessed and validated with emphasis on fracture outcomes in diverse populations.

Contributors

JY: conceptualization, data curation, formal analysis, investigation, methodology, supervision, visualization, writing—original draft.

YZ: conceptualization, data curation, investigation, methodology, project administration, supervision, writing-review & editing.

WY: conceptualization, data curation, investigation, project administration, resources, supervision, writing-review & editing.

All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication. All authors had directly accessed and verified the underlying data reported in the manuscript.

Data sharing statement

Data will be made available upon request made to the corresponding author.

Declaration of interests

We declare no competing interests.