Risks of incident major osteoporotic fractures following SARS-CoV-2 infection among older individuals: a population-based cohort study in Hong Kong

Abstract

Population-based epidemiological studies on post-acute phase coronavirus 2019 (COVID-19)-related fractures in older adults are lacking. This study aims to examine the risk of incident major osteoporotic fractures following SARS-CoV-2 infection among individuals aged ≥50, compared to individuals without COVID-19. It was a retrospective, propensity-score matched, population-based cohort study of COVID-19 patients and non-COVID individuals identified from the electronic database of the Hong Kong Hospital Authority from January 2020 to March 2022. The primary outcome was a composite of major osteoporotic fractures (hip, clinical vertebral, and upper limb). COVID-19 patients were 1:1 matched to controls using propensity-score according to age, sex, vaccination status, medical comorbidities and baseline medications. Hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated using Cox proportional hazards regression models. A total of 429 459 COVID-19 patients were included, 1:1 matched to non-COVID individuals. Upon median follow-up of 11 months, COVID-19 patients had higher risks of major osteoporotic fractures (5.08 vs 3.95 per 1000 persons; HR 1.22 95%CI [1.15–1.31]), hip fractures (2.71 vs 1.94; 1.33 [1.22–1.46]), clinical vertebral fractures (0.42 vs 0.31; 1.29 [1.03–1.62]), and falls (13.83 vs 10.36; 1.28 [1.23–1.33]). Subgroup analyses revealed no significant interaction. In acute (within 30 days) and post-acute phases (beyond 30 days) following severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, we consistently observed a significant increase in fractures and falls risks. Our study demonstrated increased risk of major osteoporotic fractures after SARS-CoV-2 infection in both acute and post-acute phases in older adults, partly due to increased fall risk. Clinicians should be aware of musculoskeletal health of COVID-19 survivors.

Lay Summary

Our study showed that older individuals with coronavirus 2019 (COVID-19) infection are at a higher risk of suffering from major osteoporotic fractures, ie serious bone fractures related to osteoporosis, compared to those not infected. The study analyzed the health records of 429 459 patients aged 50 and older in Hong Kong who had been diagnosed with COVID-19 between January 2020 and March 2022. These patients were compared with a matched group without COVID-19, considering age, sex, vaccination status, medical comorbidities, and concomitant medications. Findings indicated that individuals who had contracted COVID-19 experienced a higher risk of major osteoporotic fractures, hip fractures, and clinical vertebral fractures. The risk of falls, a common cause of these fractures, was also higher in the COVID-19 group. This increased risk of major osteoporotic fractures and falls persists both shortly after infection and in the following months, underscoring the lasting impact of COVID-19 on the bone health of older adults. These results support the recommendations for the assessment of bone health and fall risks, and an urgent review of the requirement for interventions to reduce the risk of fragility fractures in older adult COVID-19 survivors.

Graphical Abstract

Graphical Abstract.

Introduction

As of 8 November 2023, the coronavirus 2019 (COVID-19) pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has infected more than 771 million people worldwide and caused close to 7 million deaths.¹ As the number of COVID-19 survivors keeps growing, it is necessary to delineate the epidemiology of post-acute endocrine and metabolic sequelae in COVID-19 survivors, which is crucial for health policymakers to plan for appropriate surveillance and preventive strategies.

COVID-19 survivors may be predisposed to increased fracture risk through the common link of advanced age and medical comorbidities. Individuals with older age and pre-existing medical comorbidities are prone to SARS-CoV-2 infection.² Concurrently, advancing age and multimorbidity are established risk factors of fragility fractures.^3-5 Furthermore, individuals infected with COVID-19 may require hospitalization, which has been shown to contribute to fracture risk.⁴ Specifically among COVID-19 survivors, there were pre-clinical and clinical studies raising concerns about impaired bone health. A preclinical study used the golden Syrian hamster model, the main model for mild to moderate human SARS-CoV-2 infection, to show that SARS-CoV-2 infected hamsters experienced progressive loss in bone trabeculae and bone density as early as 4 days post-infection.⁶ The bone density was not restored at 60 days post-infection.⁶ The involvement was systemic, including long bones and axial skeleton.⁶ The underlying pathophysiology was increased osteoclastogenesis via a pro-inflammatory cascade such as interleukin-6 and tumour necrosis factor-alpha.⁶ Limited data on the trajectory of bone mineral density (BMD) among COVID-19 survivors were reported in a Turkish single-centre retrospective study. That study evaluated 58 COVID-19 survivors with severe COVID-19 in the acute phase. At a mean interval of 3 months, there was a mean of 8.6% decrease in lumbar spine BMD, calculated from the quantitative CT method using CT thorax images.⁷ Although that study raised concern about accelerated bone loss in the post-acute phase of COVID-19, most subjects had been treated with steroids which had known deleterious effects on the bone, confounding the results and limiting their generalisability.

Fragility fractures are the most clinically relevant outcomes resulting from impaired bone health and are burdensome to the individual and society, involving long hospitalization and high financial costs. To date, there has been a lack of population-based epidemiological studies on the risk of fractures in the post-acute phase of SARS-CoV-2 infection among older adults. A population-based cohort helps to clarify the fracture risk following SARS-CoV-2 infection and determine whether this risk persists beyond the acute phase. Hence, we performed a retrospective population-based propensity-score matched cohort study in Hong Kong to examine the risk of incident major osteoporotic fractures following SARS-CoV-2 infection among individuals aged ≥50, compared to individuals without COVID-19.

Materials and methods

The study was conducted using COVID-19-confirmed case records and COVID-19 vaccination records from the Department of Health (DH), and electronic medical records (EMR) from the Hong Kong Hospital Authority (HA). HA compiles all individual data entered into each EMR system across 43 public hospitals, 49 speciality outpatient clinics, and 74 general outpatient clinics.⁸ Over 90% of all primary, secondary, and tertiary care services in Hong Kong are provided by HA, a statutory administrative organization that serves more than 7.3 million Hong Kong residents.⁹

Study population

Those who had at least one positive COVID-19 polymerase chain reaction test or rapid antigen test result between January 2020 and March 2022 were chosen for the COVID-19 group. The inclusion period was selected to capture all COVID-19 cases from the date of the first confirmed cases to the peak of the Omicron wave in Hong Kong. The non-COVID control group comprised individuals who had never been infected as of 31 March 2022. The index date was defined as the first date of SARS-CoV-2 infection for COVID-19 patients. To ensure that the distribution of the follow-up time was similar between COVID-19 and non-COVID groups, we used maximum ratio matching to match age and sex of these two groups and then assigned a pseudo-index date from a corresponding COVID-19 survivor.^10-13 Exclusion criteria were (1) people aged <50 years, (2) COVID-19 patients who died on the index date/non-COVID individuals who died on or before the pseudo-index date, and (3) COVID-19 patients who were admitted to the hospital for fractures on the index date/non-COVID individuals who were admitted to the hospital for fractures on the pseudo-index date (Figure 1). In total, we included 2 074 275 non-COVID individuals and 429 460 COVID-19 patients. This cohort was followed up until the occurrence of outcomes, death, or 31 January 2023, whichever came first. In addition, individuals in the non-COVID group were censored at the date of SARS-CoV-2 infection.

Figure 1.

Flowchart of record linkage and selection procedure. Notes: HA = Hospital Authority; PCR = Polymerase chain reaction; RAT = Rapid antigen test.

The study protocol was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (UW 20-556, UW 21-149 and UW 21-138); and the Department of Health Ethics Committee (LM 21/2021 and LM 175/2022). Informed patient consent was not required as the data used in this study were anonymized.

Data source

Data used in this study were collected from the HA and DH. These databases are linked through unique anonymized identifiers. DH provided records of SARS-CoV-2 infection, which included the type of COVID-19 tests and the date of cases reported. DH also provided the vaccination records, including the date of administration and the types of vaccines. Centralized medical records were gathered from inpatient and outpatient encounters under the HA, including demographics, date of registered death, drug dispensing records, diagnoses, procedures, and laboratory tests. The HA electronic database has been extensively utilized to conduct population-based studies regarding fracture risk in Hong Kong.^14,15

Definitions of outcomes

The primary outcome was a composite of major osteoporotic fractures, which were defined as hip, clinical vertebral, and upper limb fractures (consisting of proximal humerus and wrist fractures). Secondary outcomes were the individual outcomes of (1) hip fractures, (2) clinical vertebral fractures, (3) upper limb fractures, and (4) falls. We considered only the first episode as the outcome of interest in our analysis.

The fracture outcomes were identified based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes,^14,15 which had been validated using the centralized EMR of the HA.¹⁶ Episodes of fall were also identified based on ICD-9-CM codes.¹⁴ (Supplementary Table 1).

Propensity-score matching

We used propensity-score models conditional on demographic characteristics, vaccination status, comorbidities (with a look-back period back to 1 January 2018; including congestive heart failure, ischaemic stroke, transient ischaemic attack, chronic obstructive pulmonary disease, liver disease, osteoporosis, history of fractures, rheumatoid arthritis and other inflammatory polyarthropathies, history of falls, hyperparathyroidism, dementia, diabetes, obesity, chronic kidney disease, end-stage kidney disease, cancer, hyperthyroidism and malabsorption), and use of medications in the 6 months before index date (including anti-diabetic medications, anti-hypertensive medications, anticoagulants, lipid-lowering agents, proton pump inhibitors, antidepressants, systemic glucocorticoids, bisphosphonates, other anti-osteoporosis therapies, calcium and/or vitamin D supplements, hormonal replacement therapy, medications for Parkinson’s disease, aromatase inhibitors, and anti-androgen) in a logistic regression model. The definitions of comorbidities are shown in Supplementary Table 1. The COVID-19 patients and non-COVID individuals were matched 1:1 using the nearest neighbour algorithm with a caliper of 0.05.¹⁷ The difference in each covariate between these two groups was measured using standardised mean difference (SMD) with values <0.1 being considered acceptable.¹⁸

Statistical analyses

Baseline characteristics were expressed as means with standard deviations (SD) for continuous variables, and numbers with percentages for categorical variables. The associations between COVID-19 and incident fractures were examined through Cox proportional hazards regression. Crude incidence rates per 1000 persons and cumulative incidence rates per 1000 person-years were reported for the COVID-19 and control groups. We then estimated the hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) of incident outcomes. To account of the pair-matched structure in the data, we used a cluster-robust sandwich variance–covariance estimator in all Cox regression models.

Sensitivity analyses

We considered the potential effect of reverse causality on the evaluation of fracture risks following SARS-CoV-2 infection as it is possible for increased diagnosis of SARS-CoV-2 infection when patients with outcome events seek healthcare service in hospitals. To address this issue, two sensitivity analyses were performed: (i) after excluding individuals with fracture events before the index date, and (ii) after excluding individuals diagnosed to have COVID-19 during hospitalization – 411 879 community-diagnosed COVID-19 patients were identified (i.e., individuals who were not hospitalized on the date of SARS-CoV-2 infection).

Furthermore, we employed the Fine-Gray model to account for mortality as a competing risk.¹⁹ To enhance the robustness of our results, we assessed the association between COVID-19 and a negative control outcome (diseases of sebaceous glands), for which there is no prior information suggesting a relationship between SARS-CoV-2 infection and diseases of sebaceous glands based on the current knowledge (Supplementary Table 1).²⁰ To evaluate the risk of fall-related fractures after SARS-CoV-2 infection, individuals who had records of both a fall and a fracture on the same day were assumed to have sustained a fall-related fracture.

Subgroup analyses

We stratified the cohort according to (i) age (<75 and ≥ 75 years), (ii) sex, (iii) presence of diabetes, (iv) vaccination status, and (v) the periods of the COVID-19 pandemic (divided by the day of 1 January 2022 when Omicron variants started to predominate). For each subgroup analysis of risk of major osteoporotic fractures and falls, a separate propensity-score model was built. P values for interaction terms were calculated for each stratifying variable.

To better understand the risk of fractures within the COVID-19 group, we compared the risk of incident major osteoporotic fractures following SARS-CoV-2 infection between COVID-19 patients who required hospitalization and those who did not. Since dexamethasone can lead to accelerated bone loss and increased fracture risk,²¹ we categorized COVID-19 patients into those who received dexamethasone²² for SARS-CoV-2 infection treatment and those who did not. Patients were considered dexamethasone-treated if dexamethasone was initiated within the first 14 days of SARS-CoV-2 infection. We then assessed the risk of major osteoporotic fractures and falls between (i) COVID-19 patients who received dexamethasone and those who did not; and (ii) COVID-19 patients who did not receive dexamethasone and non-COVID individuals, after excluding individuals who used systemic glucocorticoids six months prior to baseline.

Finally, we separately analyzed the associations between COVID-19 and incident fractures in the acute and post-acute phase. To examine the associations between COVID-19 and incident fractures during the acute phase, analyses were repeated to estimate the risks of incident fractures within 30 days after SARS-CoV-2 infection. To examine the associations between COVID-19 and incident fractures during the post-acute phase, we selected COVID-19 patients who survived beyond 30 days after the positive test result in the COVID-19 group and non-COVID individuals who survived beyond 30 days after the pseudo-index date in the non-COVID control group.²³ After exclusion of people admitted to hospitals for fractures within 30 days after the index date or pseudo-index date, 2 071 893 non-COVID individuals and 420 972 COVID-19 patients were included. This cohort was followed up from 30 days after the index date until the occurrence of outcomes, death, or 31 January 2023, whichever came first.

A two-tailed significance level of p < 0.05 was considered statistically significant. All statistical analyses were performed using the Stata Version 16.0 (StataCorp LP, College Station, TX). The analyses were conducted by XX and analyzed independently by MC for quality assurance.

Results

Baseline characteristics of the cohort

Figure 1 shows the study flow diagram. The propensity-score matched study cohort comprised 429 459 individuals in the COVID-19 group and 429 459 in the matched control group. The propensity-score distributions of the COVID-19 and matched control groups were highly overlapping, and the baseline characteristics were balanced as indicated by SMD <0.1 (Table 1 and Supplementary Figure 1). The mean age of the cohort was 66.3 years (SD: 11.0). 45.9% were men. 83.8% were vaccinated. Diabetes was a common comorbidity accounting for 19.3% of the cohort. In addition, 47.1% of the cohort were on anti-hypertensive medications.

Table 1.

Baseline characteristics of COVID-19 patients and matched non-COVID individuals after 1:1 propensity-score matching.

Baseline characteristics	COVID-19 Patients (N = 429 459)	Non-COVID individuals (N = 429 459)	SMD
	Mean ± SD /N (%)	Mean ± SD /N (%)
Sociodemographic characteristics
Age, years	66.2 ± 11.2	66.4 ± 10.9	0.02
Sex			0.01
Male	198 379 (46.2%)	195 627 (45.6%)
Female	231 080 (53.8%)	233 832 (54.4%)
Vaccination status
Unvaccinated	70 148 (16.3%)	69 402 (16.2%)	0.00
Vaccinated	359 311 (83.7%)	360 057 (83.8%)
Health care utilizations in the past year
Hospital admission	94 982 (22.1%)	84 182 (19.6%)	0.06
Emergency department visit	125 909 (29.3%)	108 592 (25.3%)	0.09
Pre-existing comorbidities
Congestive heart failure	8571 (2.0%)	6980 (1.6%)	0.03
Ischaemic stroke	11 151 (2.6%)	10 000 (2.3%)	0.02
Transient ischemic attack	2475 (0.6%)	2382 (0.6%)	0.00
Chronic obstructive pulmonary disease	12 927 (3.0%)	11 236 (2.6%)	0.02
Liver disease	1413 (0.3%)	1187 (0.3%)	0.01
Osteoporosis	3138 (0.7%)	3221 (0.8%)	0.00
History of fractures	8986 (2.1%)	8092 (1.9%)	0.01
Rheumatoid arthritis and other inflammatory polyarthropathies	1942 (0.5%)	1910 (0.4%)	0.00
History of falls	22 562 (5.3%)	19 504 (4.5%)	0.03
Hyperparathyroidism	145 (0.0%)	134 (0.0%)	0.00
Dementia	4429 (1.0%)	3341 (0.8%)	0.03
Diabetes	84 678 (19.7%)	81 089 (18.9%)	0.02
Obesity	29 560 (6.9%)	27 615 (6.4%)	0.02
Chronic kidney disease	64 453 (15.0%)	60 398 (14.1%)	0.03
End-stage kidney disease	5334 (1.2%)	4279 (1.0%)	0.02
Cancer	16 672 (3.9%)	15 511 (3.6%)	0.01
Hyperthyroidism	3612 (0.8%)	3256 (0.8%)	0.01
Malabsorption	16 (0.0%)	12 (0.0%)	0.00
Use of medications (6 months prior to baseline)
Anti-diabetic medications	83 434 (19.4%)	79 182 (18.4%)	0.03
Anti-hypertensive medications	205 731 (47.9%)	198 390 (46.2%)	0.03
Anticoagulants	11 984 (2.8%)	10 612 (2.5%)	0.02
Lipid-lowering agents	146 685 (34.2%)	144 973 (33.8%)	0.01
Proton-pump inhibitors	71 301 (16.6%)	64 340 (15.0%)	0.04
Anti-depressants	24 467 (5.7%)	23 497 (5.5%)	0.01
Systemic glucocorticoids	18 213 (4.2%)	15 741 (3.7%)	0.03
Bisphosphonates	3143 (0.7%)	2993 (0.7%)	0.00
Calcium supplements	25 701 (6.0%)	23 733 (5.5%)	0.02
Vitamin D supplements	23 625 (5.5%)	21 576 (5.0%)	0.02
Hormone replacement therapy	5793 (1.3%)	5368 (1.2%)	0.01
Other anti-osteoporosis therapies	1550 (0.4%)	1545 (0.4%)	0.00
Medications for Parkinson’s disease	3182 (0.7%)	2752 (0.6%)	0.01
Aromatase inhibitors	1552 (0.4%)	1765 (0.4%)	0.01
Anti-androgen	419 (0.1%)	360 (0.1%)	0.00

Notes: SD = Standard deviation; SMD = Standardizsed mean difference

Risks of major osteoporotic fractures following SARS-CoV-2 infection

Upon a median follow-up of 333 days (interquartile range [IQR]: 324–337), 2183 COVID-19 patients sustained major osteoporotic fractures at a median of 155 days (IQR: 61–241) while 1695 non-COVID individuals sustained major osteoporotic fractures at a median of 162 days (IQR: 83–247). COVID-19 patients had an increased risk of major osteoporotic fractures (HR 1.22, 95% CI 1.15–1.31, P < .001) (Table 2). Regarding the secondary outcomes, COVID-19 survivors had significantly higher risks of hip (HR 1.33, 95% CI 1.22–1.46, P < .001) and clinical vertebral fractures (HR 1.29, 95% CI 1.03–1.62, P < .001). On the other hand, COVID-19 patients had numerically higher risks of upper limb fractures (HR 1.10, P = .064) although not reaching statistical significance. Fall episodes, being one of the contributors to fracture risks, were notably higher among COVID-19 patients (HR 1.28, 95% CI 1.23–1.33, P < .001).

Table 2.

Hazard ratios of fracture outcomes for COVID-19 patients versus matched non-COVID individuals.

Events	COVID-19 patients (N = 429 459)				Non-COVID individuals (N = 429 459)				HRc	95% CI	P-value
	Cases with event	Cumulative incidence ratea	Crude incidence rateb	95% CI	Cases with event	Cumulative incidence ratea	Crude incidence rateb	95% CI
Major osteoporotic	2183	5.08	5.69	(5.45, 5.94)	1695	3.95	4.66	(4.44, 4.89)	1.22	(1.15, 1.31)	<0.001
Hip	1162	2.71	3.03	(2.86, 3.21)	833	1.94	2.29	(2.14, 2.45)	1.33	(1.22, 1.46)	<0.001
Clinical vertebral	179	0.42	0.47	(0.40, 0.54)	132	0.31	0.36	(0.30, 0.43)	1.29	(1.03, 1.62)	0.026
Upper limb	891	2.07	2.32	(2.17, 2.48)	769	1.79	2.11	(1.97, 2.27)	1.10	(0.99, 1.21)	0.064
Fall	5941	13.83	15.57	(15.17, 15.97)	4448	10.36	12.27	(11.91, 12.64)	1.28	(1.23, 1.33)	<0.001

Notes: HR = Hazard ratio; CI = Confidence interval

^aThe unit of cumulative incidence rate: events per 1000 persons.

^bThe unit of crude incidence rate: events per 1000 person-years.

^cHR > 1 (or < 1) indicates COVID-19 patients had a higher risk (or lower risk) of outcome compared with the matched controls.

Sensitivity analyses

The three sensitivity analyses yielded results largely consistent with the main analysis: (1) after exclusion of individuals with fractures before the index date, (2) after exclusion of COVID-19 patients who had SARS-CoV-2 infection diagnosed during hospitalization, and (3) when considering death as a competing risk (Supplementary Tables S3-S4). In line with a priori expectation, there was no significant relationship between COVID-19 and the risk of diseases of the sebaceous glands (negative control outcome) (Supplementary Table S5). Over 90% of the individuals experienced fall-related fractures, regardless of their SARS-CoV-2 infection status. The risk of fall-related fractures increased significantly after SARS-CoV-2 infection, except for the clinical vertebral fractures, which might be contributed by the relatively small number of events in this subcategory (Supplementary Table S6).

Subgroup analyses

Subgroup analyses (stratified by age, sex, presence of diabetes, vaccination status, and the periods of the COVID-19 pandemic) yielded results largely consistent with the main analyses (Supplementary Table S7). There was no significant interaction in all subgroups.

In the subgroup analysis involving exclusively COVID-19 survivors, we observed a consistent trend of increased fall risk following SARS-CoV-2 infection among hospitalized patients, and patients who received dexamethasone within 14 days of infection (Supplementary Table S8). Among those who took dexamethasone within the 14 days following SARS-CoV-2 infection, the average duration was 10.9 ± 4.1 days. We observed a significantly higher risk of major osteoporotic and hip fractures in the dexamethasone-treated group. As for the COVID-19 patients who were not treated with dexamethasone, they also had higher risks of fractures and falls compared with non-COVID individuals, in line with the results from the main analysis (Supplementary Table S9).

Risks of major osteoporotic fractures in the acute and post-acute phase following SARS-CoV-2 infection

In the acute phase (within 30 days) following SARS-CoV-2 infection, 336 individuals had major osteoporotic fractures in the COVID-19 group at a median of 10 days (IQR: 3–19), while 134 individuals had major osteoporotic fractures in the non-COVID group at a median of 16 days (IQR: 8–23). COVID-19 patients had higher risks of major osteoporotic fractures (HR 2.53, 95% CI 2.07–3.09, P < .001), hip fractures (HR 2.69, 95% CI 2.08–3.47, P < .001), upper limb fractures (HR 2.24, 95% CI 1.58–3.17, P < .001) and fall (HR 3.83, 95% CI 3.39–4.34, P < .001) (Table 3).

Table 3.

Hazard ratios of fracture outcomes for COVID-19 patients versus matched non-COVID individuals in a) acute and b) post-acute phase of SARS-CoV-2 infection.

a)
Events	COVID-19 patients (N = 429 459)				Non-COVID individuals (N = 429 459)				HRc	95% CI	P-value
	Cases with event	Cumulative incidence ratea	Crude incidence rateb	95% CI	Cases with event	Cumulative incidence ratea	Crude incidence rateb	95% CI
Major osteoporotic	336	0.78	9.65	(8.64, 10.74)	134	0.31	3.81	(3.19, 4.51)	2.53	(2.07, 3.09)	<0.001
Hip	216	0.50	6.20	(5.40, 7.09)	81	0.19	2.30	(1.83, 2.86)	2.69	(2.08, 3.47)	<0.001
Clinical vertebral	20	0.05	0.57	(0.35, 0.89)	10	0.02	0.28	(0.14, 0.52)	2.02	(0.95, 4.32)	0.069
Upper limb	102	0.24	2.93	(2.39, 3.55)	46	0.11	1.31	(0.96, 1.74)	2.24	(1.58, 3.17)	<0.001
Fall	1217	2.83	34.99	(33.05, 37.01)	320	0.75	9.09	(8.12, 10.14)	3.83	(3.39, 4.34)	<0.001
b)
Events	COVID-19 patients (N = 419 530)				Non-COVID individuals (N = 419 530)				HRc	95% CI	P-value
	Cases with event	Cumulative incidence ratea	Crude incidence rateb	95% CI	Cases with event	Cumulative incidence ratea	Crude incidence rateb	95% CI
Major osteoporotic	1836	4.38	5.27	(5.03, 5.51)	1541	3.67	4.75	(4.51, 4.99)	1.11	(1.04, 1.19)	0.002
Hip	934	2.23	2.68	(2.51, 2.85)	751	1.79	2.31	(2.15, 2.48)	1.16	(1.06, 1.28)	0.002
Clinical vertebral	158	0.38	0.45	(0.38, 0.53)	131	0.31	0.40	(0.34, 0.48)	1.13	(0.90, 1.43)	0.299
Upper limb	783	1.87	2.24	(2.09, 2.41)	690	1.64	2.12	(1.97, 2.29)	1.06	(0.95, 1.17)	0.297
Fall	4816	11.48	13.86	(13.47, 14.26)	4299	10.25	13.28	(12.89, 13.69)	1.05	(1.01, 1.09)	0.027

Notes: HR = Hazard ratio; CI = Confidence interval

^aThe unit of cumulative incidence rate: events per 1000 persons.

^bThe unit of crude incidence rate: events per 1000 person-years.

^cHR > 1 (or < 1) indicates COVID-19 patients had a higher risk (or lower risk) of outcome compared with the matched controls.

For the evaluation of the fracture risk in the post-acute phase (beyond 30 days) following SARS-CoV-2 infection, Supplementary Table S2 shows the characteristics 30 days after the index date. Upon a median follow-up of 303 days (IQR: 295–307), 1836 individuals had major osteoporotic fractures in the COVID-19 group at a median of 149 days (IQR: 74–224), while 1541 individuals had major osteoporotic fractures in the non-COVID group at a median of 147 days (IQR: 70–221). COVID-19 survivors had higher risks of major osteoporotic fractures (HR 1.11, 95% CI 1.04–1.19, P = .002), hip fractures (HR 1.16, 95% CI 1.06–1.28, P = .002), and fall (HR 1.05, 95% CI 1.01–1.09, P = .027) (Table 3).

Discussion

Our study delineated the risk and burden of incident fractures following SARS-CoV-2 infection. In this population-based study of Hong Kong with a median follow-up of 11 months, COVID-19 survivors had a higher risk of incident major osteoporotic fractures following SARS-CoV-2 infection. Furthermore, the increased risk extended beyond the acute phase and remained elevated in the post-acute phase. The increased risk of fall following SARS-CoV-2 infection observed in this cohort could contribute to the increase in fracture risk. Our results would call for clinicians’ awareness of the musculoskeletal health of COVID-19 survivors, including fall prevention.

Previous epidemiological studies focused on how infection control policies related to COVID-19 pandemic affected the incidence of major osteoporotic fractures.^24,25 Mobility restrictions in the COVID-19 pandemic were associated with reduced rate of major osteoporotic fractures. Indeed, we also observed a lower incidence of fractures during the pre-Omicron period of the pandemic. Nonetheless, mobility restrictions related to confinement were probably more relevant in the early phase of the COVID-19 pandemic when isolation policies were strict back then. With the relaxation of the confinement requirement especially during the Omicron wave, the impact of mobility restrictions on the fracture risk has attenuated. Our study extended the current understanding by studying the fracture risk of COVID-19 survivors with reference to counterparts who did not have SARS-CoV-2 infection in the same phase of the COVID-19 pandemic.

Our study revealed elevated risks of incident major osteoporotic fractures among COVID-19 survivors compared to individuals without positive testing for COVID-19. Interplay of multiple factors may account for this. Firstly, fall is an important mechanism leading to fragility fractures.²⁶ We saw an increased fall risk among COVID-19 survivors, which could be secondary to the musculoskeletal involvements in COVID-19.²⁷ We further observed that most of the fractures were related to fall. Moreover, the increase in fall risk was exaggerated among hospitalized patients and patients with more severe SARS-CoV-2 infection. Myalgia and muscle weakness are common presentations of acute SARS-CoV-2 infection,²⁸ which can persist in the post-acute phase of COVID-19.²⁹ Sarcopenia may also contribute, especially among those with more severe illness.³⁰ Dexamethasone-treated COVID-19 survivors had higher fall risk, which could be related to the effect of exogenous steroid in decreasing the muscle mass.²¹ Secondly, SARS-CoV-2 infection may be directly (via potential direct viral effect) and indirectly (via the associated inflammation) associated with an increased fracture risk. With more severe COVID-19, there is a higher chance of development of significant inflammation. The pro-inflammatory cytokine cascade could lead to subsequent increased bone loss.³¹ Patients are also more prone to muscle wasting with more severe COVID-19 illness,³² which in turn is associated with increased fall risk. Several studies suggested that SARS-CoV-2 infection may be associated with subsequent bone loss. A pre-clinical study using the golden Syrian hamster model showed progressive bone loss following SARS-CoV-2 infection.⁶ A clinical study in Turkey showed the higher-than-expected rate of bone density deterioration following COVID-19 at a mean interval of 3 months.⁷ High dose of steroid for treating SARS 20 years ago was found to be associated with bone loss.³³ Our study showed that COVID-19 survivors who had been treated with dexamethasone had higher fracture risks than COVID-19 survivors who had not been treated with dexamethasone, suggesting that even a lower dose and shorter duration of steroid used in SARS-CoV-2 infection might be detrimental to bone health. In fact, even those who were not treated with dexamethasone were shown to have higher risks of fractures than non-COVID controls in our study. Similar phenomena of increased bone loss and fracture risk in the acute phase following an illness have been described among individuals suffering from acute respiratory distress syndrome³⁴ and burns injury.³⁵ Further work by including a non-COVID respiratory infection group may help characterize the specific effects of SARS-CoV-2 infection on the bone health. Thirdly, COVID-19 survivors may have long COVID, which may manifest as ongoing pulmonary and cardiac dysfunction or fatigue.³⁶ These may contribute to the increased fracture risk in the post-acute phase of illness. Our results support the need for clinicians to consider proactive measures, such as rehabilitation programme,³⁷ to minimize the musculoskeletal sequelae of COVID-19. This also applies to steroid-treated COVID-19 patients, as steroid therapy is part of the COVID-19 treatment regime especially for the more severe cases.²²

The increase in risks of incident major osteoporotic fractures among COVID-19 survivors was mainly driven by the increased risk of hip and clinical vertebral fractures. We also observed consistent trend of increased risk of upper limb fractures, although not reaching statistical significance. From the pre-clinical of the golden Syrian hamster model, it is expected that the bone loss is systemic.⁶ Given the availability of data in this population-based dataset, only clinical vertebral fractures could be identified and silent vertebral fractures could be missed.³⁸ Moreover, in the COVID-19 pandemic, patients with milder forms of wrist fractures may not present to the health care system.³⁹ These may lead to the underestimation of the true differential risk of upper limb fractures among COVID-19 survivors. To date, clinical studies of potential effect of COVID-19 on bone health have been limited to a short period following SARS-CoV-2 infection. The only clinical study evaluating bone density changes after COVID-19 so far was performed in Turkey, which reported a higher-than-expected rate of bone density deterioration following COVID-19 over three months,⁷ but confounded by steroid treatment. Furthermore, the evaluation of bone density was limited to the lumbar spine. The bone density changes at other sites, such as the hip, among COVID-19 survivors remain to be determined. Nonetheless, there are several case reports of incident osteoporosis or fracture temporally related to SARS-CoV-2 infection, both in vertebral and non-vertebral sites. One of them described a 61-year-old man who had transient osteoporosis of the hip in close temporal correlation with SARS-CoV-2 infection.⁴⁰ The other report described the occurrence of vertebral fracture, likely osteoporotic in nature, in a 53-year-old man within a few days after the diagnosis of COVID-19.⁴¹ Studies in larger populations may help to elucidate the impact of SARS-CoV-2 infection on the risk of upper limb fractures.

In the subgroup analyses, there was no significant heterogeneity across different clinical factors of osteoporosis, including age, sex, diabetes status, vaccination status, and fracture history. From the perspective of SARS-CoV-2, we observed no significant interaction when stratified by the prevalent SARS-CoV-2 strains, suggesting that the results from our study would apply to the current COVID-19 survivors.

Our results should be interpreted bearing certain limitations. First, asymptomatic or mild COVID-19 patients who were neither tested for SARS-CoV-2, nor reported to the health authority were classified as control individuals, leading to potential misclassification bias. Second, information on BMD measurements and bone turnover markers was not available in this electronic health database. Follow-up studies on the BMD trajectory and changes in bone turnover markers of COVID-19 survivors could help to elucidate the underlying pathophysiology of increased fracture risk following SARS-CoV-2 infection. Third, information on body mass index was not available. Nonetheless, obesity, defined by ICD-9-CM diagnostic code, was included as a categorical variable in our study in the propensity-score matching. Fourth, it is possible that silent vertebral fractures remained undetected, as this study did not assess vertebral fractures systematically by quantitative vertebral morphometry. Indeed, vertebral fractures are frequently asymptomatic,³⁸ so morphometric vertebral fractures could go undetected. This could lead to an underestimation of the number of vertebral fractures. Fifth, the analyses among individuals treated with dexamethasone may be subject to confounding by indication, as treatment decisions were likely made based on clinical need. Last but not least, inherent to the observational nature of the study, there may still be unmeasured confounders. However, we have quantitively evaluated the residual confounding by conducting a negative control analysis, which showed that the extent of residual confounding was likely small.

In conclusion, our study with a median follow-up of 11 months demonstrated an increased risk of major osteoporotic fractures following SARS-CoV-2 infection among older adults. The increased risk of fall following SARS-CoV-2 infection observed in this cohort could contribute to the increase in fracture risk. Our results support the recommendations for the assessment of bone health and fall risks, and an urgent review of the requirement for interventions to reduce the risk of fragility fractures in older adult COVID-19 survivors.

Author contributions

David Lui (Conceptualization, Investigation, Methodology, Writing – original draft, Writing – review & editing), Xi Xiong (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft), Ching-Lung Cheung (Investigation, Writing – review & editing), Francisco Lai (Funding acquisition, Investigation, Writing – review & editing), Xue Li (Investigation, Writing – review & editing), Eric Wan (Investigation, Writing – review & editing), Celine Chui (Investigation, Writing – review & editing), Esther Chan (Funding acquisition, Investigation, Writing – review & editing), Franco Cheng (Investigation, Writing – review & editing), Matthew Chung (Data curation, Formal analysis, Visualization), Ivan Au (Data curation, Formal analysis, Visualization ), Chi Ho Lee (Investigation, Writing – review & editing), Tai Pang Ip (Investigation, Writing – review & editing), Yu Cho Woo (Investigation, Writing – review & editing), Kathryn Tan (Investigation, Writing – review & editing), Carlos Wong (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing – original draft, Writing – review & editing), and Ian Wong (Funding acquisition, Investigation, Project administration, Resources, Software, Supervision, Writing – review & editing)

Funding

This work was supported by Collaborative Research Fund, University Grants Committee, The Hong Kong Special Administrative Region (HKSAR) Government (principal investigator, I.C.K.W; reference no. C7154-20GF); and the Health Bureau, HMRF Research on COVID-19, the HKSAR Government (principal investigator [work package 2], E.W.Y.C; reference no. COVID1903011). I.C.K.W, C.K.H.W, and F.T.T.L are partially supported by the Laboratory of Data Discovery for Health (D²4H) funded by the AIR@InnoHK administered by the Innovation and Technology Commission. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflicts of interest

F.T.T.L has been supported by the RGC Postdoctoral Fellowship under the Hong Kong Research Grants Council. X.L has received research grants from the Health Bureau of the Government of the Hong Kong SAR, research and educational grants from Janssen and Pfizer; internal funding from University of Hong Kong; consultancy fee from Merck Sharp & Dohme, unrelated to this work. E.Y.F.W has received research grants from the Health Bureau of the Government of the Hong Kong SAR, and the Hong Kong Research Grants Council, outside the submitted work. C.S.LC has received grants from the Health Bureau of the Hong Kong Government, Hong Kong Research Grant Council, Hong Kong Innovation and Technology Commission, Pfizer, IQVIA, and Amgen; personal fee from Primevigilance Ltd.; outside the submitted work. E.W.Y.C reports honorarium from Hospital Authority, grants from Research Grants Council (RGC, Hong Kong), grants from Research Fund Secretariat of the Health Bureau, grants from National Natural Science Fund of China, grants from Wellcome Trust, grants from Bayer, grants from Bristol-Myers Squibb, grants from Pfizer, grants from Janssen, grants from Amgen, grants from Takeda, grants from Narcotics Division of the Security Bureau of HKSAR, outside the submitted work. C.K.H.W reports receipt of research funding from the EuroQoL Group Research Foundation, the Hong Kong Research Grants Council, the Hong Kong Health and Medical Research Fund; AstraZeneca and Boehringer Ingelheim, unrelated to this work. I.C.K.W reports research funding outside the submitted work from Amgen, Bristol-Myers Squibb, Pfizer, Janssen, Bayer, GSK, Novartis, the Hong Kong RGC, and the Hong Kong Health and Medical Research Fund, National Institute for Health Research in England, European Commission, National Health and Medical Research Council in Australia, and also received speaker fees from Janssen and Medice in the previous 3 years. All other authors declare no competing interests.

Data availability

The data that support the findings of this study were extracted from the Hospital Authority database in Hong Kong. Restrictions apply to the availability of these data, which were used under license for this study. Data sharing is prohibited by the Hospital Authority.