Red flags to screen for vertebral fracture in people presenting with low back pain

Abstract

Background

Low back pain is a common presentation across different healthcare settings. Clinicians need to confidently be able to screen and identify people presenting with low back pain with a high suspicion of serious or specific pathology (e.g. vertebral fracture). Patients identified with an increased likelihood of having a serious pathology will likely require additional investigations and specific treatment. Guidelines recommend a thorough history and clinical assessment to screen for serious pathology as a cause of low back pain. However, the diagnostic accuracy of recommended red flags (e.g. older age, trauma, corticosteroid use) remains unclear, particularly those used to screen for vertebral fracture.

Objectives

To assess the diagnostic accuracy of red flags used to screen for vertebral fracture in people presenting with low back pain. Where possible, we reported results of red flags separately for different types of vertebral fracture (i.e. acute osteoporotic vertebral compression fracture, vertebral traumatic fracture, vertebral stress fracture, unspecified vertebral fracture).

Search methods

We used standard, extensive Cochrane search methods. The latest search date was 26 July 2022.

Selection criteria

We considered primary diagnostic studies if they compared results of history taking or physical examination (or both) findings (index test) with a reference standard test (e.g. X‐ray, magnetic resonance imaging (MRI), computed tomography (CT), single‐photon emission computerised tomography (SPECT)) for the identification of vertebral fracture in people presenting with low back pain. We included index tests that were presented individually or as part of a combination of tests.

Data collection and analysis

Two review authors independently extracted data for diagnostic two‐by‐two tables from the publications or reconstructed them using information from relevant parameters to calculate sensitivity, specificity, and positive (+LR) and negative (−LR) likelihood ratios with 95% confidence intervals (CIs). We extracted aspects of study design, characteristics of the population, index test, reference standard, and type of vertebral fracture. Meta‐analysis was not possible due to heterogeneity of studies and index tests, therefore the analysis was descriptive. We calculated sensitivity, specificity, and LRs for each test and used these as an indication of clinical usefulness. Two review authors independently conducted risk of bias and applicability assessment using the QUADAS‐2 tool.

Main results

This review is an update of a previous Cochrane Review of red flags to screen for vertebral fracture in people with low back pain. We included 14 studies in this review, six based in primary care, five in secondary care, and three in tertiary care. Four studies reported on 'osteoporotic vertebral fractures', two studies reported on 'vertebral compression fracture', one study reported on 'osteoporotic and traumatic vertebral fracture', two studies reported on 'vertebral stress fracture', and five studies reported on 'unspecified vertebral fracture'. Risk of bias was only rated as low in one study for the domains reference standard and flow and timing. The domain patient selection had three studies and the domain index test had six studies rated at low risk of bias. Meta‐analysis was not possible due to heterogeneity of the data. Results from single studies suggest only a small number of the red flags investigated may be informative.

In the primary healthcare setting, results from single studies suggest 'trauma' demonstrated informative +LRs (range: 1.93 to 12.85) for 'unspecified vertebral fracture' and 'osteoporotic vertebral fracture' (+LR: 6.42, 95% CI 2.94 to 14.02). Results from single studies suggest 'older age' demonstrated informative +LRs for studies in primary care for 'unspecified vertebral fracture' (older age greater than 70 years: 11.19, 95% CI 5.33 to 23.51). Results from single studies suggest 'corticosteroid use' may be an informative red flag in primary care for 'unspecified vertebral fracture' (+LR range: 3.97, 95% CI 0.20 to 79.15 to 48.50, 95% CI 11.48 to 204.98) and 'osteoporotic vertebral fracture' (+LR: 2.46, 95% CI 1.13 to 5.34); however, diagnostic values varied and CIs were imprecise. Results from a single study suggest red flags as part of a combination of index tests such as 'older age and female gender' in primary care demonstrated informative +LRs for 'unspecified vertebral fracture' (16.17, 95% CI 4.47 to 58.43).

In the secondary healthcare setting, results from a single study suggest 'trauma' demonstrated informative +LRs for 'unspecified vertebral fracture' (+LR: 2.18, 95% CI 1.86 to 2.54) and 'older age' demonstrated informative +LRs for 'osteoporotic vertebral fracture' (older age greater than 75 years: 2.51, 95% CI 1.48 to 4.27). Results from a single study suggest red flags as part of a combination of index tests such as 'older age and trauma' in secondary care demonstrated informative +LRs for 'unspecified vertebral fracture' (+LR: 4.35, 95% CI 2.92 to 6.48). Results from a single study suggest when '4 of 5 tests' were positive in secondary care, they demonstrated informative +LRs for 'osteoporotic vertebral fracture' (+LR: 9.62, 95% CI 5.88 to 15.73).

In the tertiary care setting, results from a single study suggest 'presence of contusion/abrasion' was informative for 'vertebral compression fracture' (+LR: 31.09, 95% CI 18.25 to 52.96).

Authors' conclusions

The available evidence suggests that only a few red flags are potentially useful in guiding clinical decisions to further investigate people suspected to have a vertebral fracture. Most red flags were not useful as screening tools to identify vertebral fracture in people with low back pain. In primary care, 'older age' was informative for 'unspecified vertebral fracture', and 'trauma' and 'corticosteroid use' were both informative for 'unspecified vertebral fracture' and 'osteoporotic vertebral fracture'. In secondary care, 'older age' was informative for 'osteoporotic vertebral fracture' and 'trauma' was informative for 'unspecified vertebral fracture'. In tertiary care, 'presence of contusion/abrasion' was informative for 'vertebral compression fracture'. Combinations of red flags were also informative and may be more useful than individual tests alone. Unfortunately, the challenge to provide clear guidance on which red flags should be used routinely in clinical practice remains. Further research with primary studies is needed to improve and consolidate our current recommendations for screening for vertebral fractures to guide clinical care.

Plain language summary

Use of red flags to screen for vertebral fractures in people with low back pain

Key messages

– The four best red flags for vertebral fractures in people with low back pain were corticosteroid use (e.g. medicines that can weaken bones), older age (e.g. aged above 70 years), trauma (e.g. a fall), and a contusion (bruising) or abrasion (cuts and grazes).

– More research is needed to identify the best red flags or combination of red flags to screen for spinal fractures.

Red flags to screen for spinal fractures

Red flags for spinal fractures are signs and symptoms found by a health professional (e.g. doctor, physiotherapist) during an examination that warn that something is wrong within the spine (backbone). The accuracy of red flags is important, as low‐quality tests can lead to incorrect diagnosis and treatment. On the one hand, if the tests are not accurate, people without a spinal fracture (break) may undergo unnecessary imaging (e.g. X‐ray, magnetic resonance imaging that uses radio waves to produce detailed images of the inside of the body). Some of these imaging methods lead to radiation exposure, extra costs, and added worry for the patient. On the other hand, missing a spinal fracture will result in a delay in receiving treatment and reduce quality of life. Therefore, identifying the most accurate red flags to screen for spinal fractures is needed.

What did we want to find out?

We wanted to assess how accurate the red flags used to screen for spinal fracture are in people presenting with low back pain. Where possible, we reported results of red flags separately for the different types of spinal fracture, such as osteoporotic vertebral compression fractures (e.g. fractures due to osteoporosis), vertebral traumatic fracture (e.g. due to falls), vertebral stress fracture (e.g. rapid increase in load on the spine), or unspecified vertebral fracture (e.g. no specific cause reported).

What did we do?

We updated a previous Cochrane Review. We searched for studies that investigated the accuracy of red flags across different healthcare settings. We included studies that compared results of history taking and physical examination (or both) (known as index tests or red flags) with different types of imaging (known as the reference standard) to identify spinal fractures in people with low back pain. We also included studies if they reported on the results of red flags separately for the different types of spinal fractures.

What did we find?

Fourteen studies investigated different red flags used to identify spinal fractures, and most of the red flags were not accurate or useful. Overall, the four best red flags found were corticosteroid use (e.g. medicines that can weaken bones), person's age (e.g. aged above 70 years), trauma (e.g. a fall), and a contusion (bruising) or abrasion (cuts and grazes).

In the primary healthcare setting (e.g. general practitioners), 'trauma' as a red flag was best to screen for 'unspecified spinal fracture' and 'osteoporotic spinal fracture'. 'Older age' as a red flag was best to screen for 'unspecified spinal fracture' in primary care. 'Corticosteroid use' may be useful as a red flag in primary care to screen for 'unspecified spinal fracture' and 'osteoporotic spinal fracture'. Red flags as part of a combination of index tests such as 'older age and female gender' as a red flag in primary care is best to screen for 'unspecified spinal fracture'.

In the secondary healthcare setting (e.g. specialists and consultants), 'trauma' as a red flag is best to screen for 'unspecified spinal fracture' and 'older age' for 'osteoporotic spinal fracture'. Red flags as part of a combination of index tests such as 'older age and trauma' in secondary care as a red flag is best to screen for 'unspecified spinal fracture'. When 'four of five tests' are positive in secondary care as a red flag, it may be used to screen for 'osteoporotic spinal fracture'.

In the tertiary care setting (e.g. specialised care in a hospital setting), the 'presence of contusion/abrasion' as a red flag was best to screen for 'spinal compression fracture'.

What are the limitations of the evidence?

A limitation of our review is that most of the included studies were different in terms of the healthcare setting they were performed in, used different study designs, or presented data for different types of spinal fracture, which made drawing conclusions difficult. Many of the red flags investigated were from single studies and few studies investigated the same index tests. There was also little uniform agreement on the definition of red flags (e.g. for corticosteroids, it was not clear how long and how much was used), which may explain why the accuracy of some red flags varied from study to study. Some red flags may also affect different types of spinal fractures differently; however, this was not clearly reported in most cases.

How up to date is this evidence?

The evidence is up to date to July 2022.

Summary of findings

Summary of findings 1. Results of potentially useful red flags.

Review question: what is the accuracy of red flags to screen for vertebral fracture in people presenting with low back pain or for lumbar examination? Patient population: people with low back pain or requiring examination of the lumbar spine when presenting to care in primary, secondary, or tertiary settings Index tests: all relevant features taken during a history or physical examination Target condition: vertebral fracture Reference standard: diagnostic imaging (MRI, CT, X‐ray, bone scan) Included studies: 4 prospective cohorts, 4 retrospective chart reviews Main limitations: small number of studies included; large heterogeneity between studies and index tests precluded pooling of results; descriptive analysis presented; inadequate reporting of methods
Population	Reference standard	Number of studies with high risk of bias per QUADAS‐2 domain: participant selection/index test/ reference standard/flow and timing	Sensitivity (95% CI) or range of estimatesa	Specificity (95% CI) or range of estimatesa	+LR (95% CI)	−LR (95% CI)
Results: single red flags
Primary care: 1 study No. of participants: 1172 Index test: age > 70 years	Follow‐up with diagnostic imaging	0/0/0/1	0.50 (0.16 to 0.84)	0.96 (0.94 to 0.97)	11.19 (5.33 to 23.51)	0.52 (0.26 to 1.05)
Primary care: 2 studies No. of participants: 3179 Index test: age > 74 years	X‐ray, follow‐up with diagnostic imaging	0/0/0/1	Range: 0.25–0.59	Range: 0.84–0.97	Range: 3.69–9.39	Range: 0.49–0.77
Primary care: 3 studies No. of participants: 2152 Index test: history of corticosteroid use	X‐ray, follow‐up with diagnostic imaging	2/0/2/2	Range: 0.00–0.25	Range: 0.93–0.99	Range: 2.46–48.50	Range: 0.75–0.98
Primary care: 4 studies No. of participants: 3023 Index test: trauma	X‐ray, follow‐up with diagnostic imaging	2/0/2/2	Range: 0.21–0.65	Range: 0.89–0.98	Range: 1.93–12.85	Range: 0.37–0.88
Secondary care: 1 study No. of participants: 9940 Index test: age > 70 years	Imagingb	0/0/0/0	0.31 (0.27 to 0.35)	0.80 (0.79 to 0.81)	1.55 (1.36 to 1.76)	0.86 (0.82 to 0.91)
Secondary care: 1 study No. of participants: 9940 Index test: trauma	X‐ray	0/0/0/0	0.25 (0.21 to 0.29)	0.89 (0.88 to 0.89)	2.18 (1.86 to 2.54)	0.85 (0.81 to 0.89)
Tertiary care: 3 studies No. of participants: 1142 Index test: trauma	X‐ray	0/1/0/0	Range: 0.07–1.00	Range: 0.51–0.60	Range: 0.18–1.93	Range: 0.12–1.54
Tertiary care: 1 study No. of participants: 552 Index test: contusion/abrasion	X‐ray	0/0/0/0	0.85 (0.70 to 0.94)	0.97 (0.95 to 0.98)	31.09 (18.25 to 52.96)	0.15 (0.07 to 0.32)
Results: combined red flags
Primary care: 2 studies No. of participants: 3182 Index test: female and age > 64 years	X‐ray, follow‐up with diagnostic imaging	0/0/0/1	Range: 0.59–0.63	Range: 0.79–0.96	Range: 2.75–14.58	Range: 0.39–0.52
Primary care: 2 studies No. of participants: 3280 Index test: female and age > 74 years	X‐ray, follow‐up with diagnostic imaging	0/0/0/1	Range: 0.25–0.45	Range: 0.90–0.98	Range: 4.36–16.17	Range: 0.62–0.76
Primary care: 1 study No. of participants: 1172 Index test: 2/4 red flags positive (Henschke)c	Follow‐up with diagnostic imaging	0/0/0/1	0.63 (0.24 to 0.91)	0.96 (0.95 to 0.97)	15.48 (8.45 to 28.36)	0.39 (0.16 to 0.96)
Secondary care: 1 study No. of participants: 9940 Index test: age > 70 and Trauma	Imagingb	0/0/0/0	0.05 (not reported)	0.99 (not reported)	4.35 (2.92 to 6.48)	0.96 (0.94 to 0.98)
Secondary care: 1 study No. of participants: 1448 Index test: 4/5 features positive (Roman)d	X‐ray or CT	0/0/0/1	0.37 (0.22 to 0.54)	0.96 (0.95 to 0.97)	9.62 (5.88 to 15.73)	0.66 (0.52 to 0.84)
Tertiary care: 1 study No. of participants: 108 Index test: trauma and neurological signs	X‐ray	0/0/0/0	0.29 (0.04 to 0.71)	0.98 (0.93 to 1.00)	14.43 (2.38 to 87.64)	0.73 (0.46 to 1.15)
CI: confidence interval; CT: computed tomography; LR: likelihood ratio; MRI: magnetic resonance imaging.

^a Sensitivity and specificity are the point estimate and 95% CI when only one study investigated an index test and the point estimate range when multiple studies investigated the same test.
^b Imaging type not reported.
^c Henschke combination: age > 70 years, significant trauma, prolonged corticosteroid use, altered sensation from the trunk down.
^d Roman combination: age > 52 years, no leg pain present, body mass index < 23, does not exercise regularly, female gender.

Background

Low back pain continues to be the leading cause for years lived with disability according to the Global Burden of Disease Study (Wu 2020). It is estimated that 577 million people experience low back pain globally, and in 2017 the estimated global years lived with disability was 64.9 million (Wu 2020). It is widely agreed that low back pain can be very disabling and impose an enormous social and economic burden on the community (Buchbinder 2018). Despite the disability burden of low back pain, it is rare for presentations to be due to a serious underlying condition (e.g. cauda equina syndrome, spinal tumours, and cancer) (Hartvigsen 2018). Most presentations can be managed in primary care without further diagnostic investigation (Oliveira 2018). However, there is an increasing number of low back pain presentations to emergency departments (Ferreira 2021).

Guidelines recommend that a thorough history and clinical assessment be used to identify people with suspicion of serious pathology who then require further diagnostic work‐up to confirm the diagnosis. In the primary care setting, between 1% and 5% of all people who present with low back pain will have a serious spinal pathology (Downie 2014; Hartvigsen 2018; Henschke 2009; Williams 2013). The prevalence of serious pathology in people presenting to the emergency department is thought to be somewhat higher, one review reporting figures between 2.5% and 5.1% (Galliker 2019). In both primary care and emergency department settings vertebral fractures are amongst the most common serious pathologies which a clinician needs to screen for.

The use of red flags with high false‐positive rates may contribute to unnecessary imaging and costs in low back pain management (Williams 2013). In contrast, red flags with high false‐negative rates can lead to failure to recognise serious pathology, resulting in delayed testing and treatment, and may increase patient morbidity (DePalma 2020).

Target condition being diagnosed

Lumbar vertebral fractures can include vertebral traumatic fractures, vertebral stress fractures, and acute osteoporotic vertebral compression fractures. These three types of vertebral fractures may have different distinguishing clinical histories and clinical features, and may require different diagnostic work‐up and management.

Lumbar vertebral stress fractures are thought to be secondary to chronic low‐grade trauma or repetitive loading, which may occur in sport or training of new military recruits (Alqarni 2015). Lumbar vertebral traumatic fracture generally results from high‐energy trauma (Lee 2016; Marongiu 2018). Osteoporotic vertebral compression fractures can occur from minor trauma (e.g. fall from a standing height or less) or without any definitive trauma (e.g. non‐acute fracture) in older people with osteoporosis (Lee 2016). Osteoporotic vertebral fractures are associated with a two‐ to eight‐fold risk of mortality (Schousboe 2016), and a five‐fold increased risk of sustaining a further vertebral fracture (Marongiu 2018), and future hip fracture (Warriner 2011). One systematic review of 14 primary care studies (14,860 adults aged 50 years and older) found the median prevalence of vertebral fractures in people presenting with low back pain was estimated at 3.6% in primary care and 6.5% in secondary and tertiary care (Downie 2014).

Index test(s)

The purpose of the index test is screening for lumbar vertebral fracture. In people presenting with low back pain, clinical practice guidelines generally recommend assessing for the following red flags to help identify people with a greater suspicion of vertebral fracture: major/significant trauma, use of corticosteroids, greater than 50 years of age, and obvious structural deformity (Verhagen 2016).

Despite the inclusion of several red flags in guidelines, the usefulness of red flags to screen for serious pathology, such as fracture, continues to be debated (Grunau 2017; Underwood 2009). Several red flags endorsed by previous guidelines have low or untested diagnostic accuracy with little information on their utility for clinical practice (e.g. dose and duration of corticosteroid use not reported) (Cook 2018), or are based on evidence from single studies only (Parreira 2019).

In 2013, our group published a Cochrane Review of eight studies which evaluated 29 different index tests used to screen people presenting with low back pain for vertebral fracture (Williams 2013). Results from individual studies suggested many of the endorsed red flags, such as thoracic pain and neurological deficit, were uninformative because these red flags had both positive (+LR) and negative (−LR) likelihood ratios very close to 1 (Williams 2013). A +LR or −LR close to 1 indicates that the test does not substantially increase or decrease the likelihood of a fracture being present (if positive) or absent (if negative) (Deeks 2004). There was also variation between studies in the estimates of the diagnostic accuracy of the index tests. For example, for significant trauma, the sensitivity ranged from 0.25 (95% confidence interval (CI) 0.03 to 0.65) to 0.65 (95% CI 0.44 to 0.83) and the specificity ranged from 0.90 (95% CI 0.86 to 0.93) to 0.98 (95% CI 0.96 to 0.98). For corticosteroid use, the sensitivity ranged from 0.00 (95% CI 0.00 to 0.23) to 0.25 (95% CI 0.03 to 0.65), however the test was highly specific (0.99, 95% CI 0.98 to 1.00). For older age (greater than 74 years), the sensitivity ranged from 0.25 to 0.59 and the specificity ranged from 0.84 to 0.97. The +LRs for the different red flags were: significant trauma (from 3.42 (95% CI 1.57 to 7.45) to 12.85 (95% CI 8.58 to 19.24)); older age (from 3.69 (95% CI 3.00 to 4.53) to 9.39 (95% CI 2.69 to 32.75)); and corticosteroid use (from 3.97 (95% CI 0.20 to 79.15) to 48.50 (95% CI 11.46 to 204.98)).

Red flags used in isolation had only modest diagnostic accuracy. Combinations of red flags seemed more informative than red flags used alone. For example, age greater than 64 years had a +LR of 7.13 which increased to 16.17 when combined with female gender (Williams 2013). However, when combining red flags, the imprecision of the estimates also increased. One red flag in a tertiary care setting appeared informative with a high +LR for contusion/abrasion (31.09, 95% CI 18.25 to 52.96). Overall, there was uncertainty about the value of most red flags, as estimates were imprecise, there was a lack of agreement for older age cut‐off points between studies and an overall lack of uniform agreement of definitions of index tests, and most red flags were only evaluated within a single study.

Clinical pathway

Most clinical practice guidelines recommend history and physical examination to identify people with a higher likelihood of low back pain due to vertebral fracture (Oliveira 2018), who require further diagnostic work‐up. Different types of vertebral fractures may require different imaging procedures.

The American College of Radiology (ACR) recommends standard radiography for initial imaging in people presenting with low back pain with a suspicion of osteoporotic vertebral compression fracture (Patel 2016). Genant's semi‐quantitative assessment and the algorithm‐based qualitative approach are methods used to classify and confirm the presence of a vertebral fracture on standard radiographs (Genant 1993; Jiang 2004). Computed tomography (CT) is recommended over a plain radiograph if there is suspicion of vertebral traumatic fracture and bone scan with single‐photon emission computerised tomography (SPECT) if there is suspicion of vertebral stress fracture (Patel 2016).

Due to concerns of radiation exposure from standard radiography and CT (Berrington de González 2009; Wall 2006), vertebral fracture assessment by dual‐energy X‐ray absorptiometry (DXA) is a potential alternative to assess for osteoporotic and traumatic vertebral fractures (Lee 2016). However, there is currently insufficient evidence to replace standard radiography with DXA or CT due to its reduced ability to detect lower‐grade vertebral fractures (Fuerst 2009; Lee 2016; Lems 2021), reduced image resolution (Bazzocchi 2012; Damiano 2006; Schousboe 2006), and numerous potential sources of error (Johansson 2020).

Magnetic resonance imaging (MRI) is now preferred over CT in evaluating people with suspicion of lumbar stress fracture. MRI can detect bone marrow oedema early, facilitating early diagnosis of lumbar stress reaction and fracture without radiation exposure (Ang 2016; Astur 2015). Given that, most people presenting with low back pain with suspicion of lumbar stress fracture tend to be from younger age groups, MRI may be beneficial as first‐line imaging given its high sensitivity and specificity, and non‐ionising radiation (Alqarni 2015; Astur 2015; Cheung 2018; Murthy 2012). However, the higher cost of MRI and potential increase in downstream expenditure need to be considered (Jarvik 2003).

Prior test(s)

In this review on index tests to screen for vertebral fracture, no prior tests are required as the index tests are used in the initial screening to increase suspicion of vertebral fracture in people presenting with low back pain.

Role of index test(s)

Index tests are used as screening tests, that is to rule in vertebral fracture in order to identify those who may require further diagnostic testing and specialised care. However, index tests are also used as screening tests to rule out vertebral fracture in order to reduce unnecessary imaging, treatments, and costs.

Alternative test(s)

Apart from a combination of history taking and physical examination, there are no alternative screening tests for vertebral fracture in people presenting with low back pain. Routinely imaging all people with low back pain would overburden healthcare systems and is discouraged in current guidelines (Oliveira 2018).

Rationale

In light of the publication of primary diagnostic studies (de Schepper 2016; Premkumar 2018), and due to the evolving guidance for the most appropriate methods to systematically review studies of diagnostic test accuracy (Deeks 2022; Whiting 2011), we aimed to update the previous Cochrane Review (Williams 2013). The results of this review will provide researchers and clinicians with more clarity in the use of red flags when screening for vertebral fracture in people presenting with low back pain in primary, secondary, and tertiary care settings.

Objectives

To assess the diagnostic accuracy of red flags used to screen for vertebral fracture in people presenting with low back pain. Where possible, we reported results of red flags separately for different types of vertebral fracture (i.e. acute osteoporotic vertebral compression fracture, vertebral traumatic fracture, vertebral stress fracture, unspecified vertebral fracture).

Secondary objectives

To assess the influence of sources of heterogeneity on the diagnostic accuracy of red flag tests, if appropriate. In particular, if possible, we explored heterogeneity within the different types of vertebral fracture (i.e. osteoporotic vertebral compression fracture, vertebral traumatic fracture, and vertebral stress fracture) due to factors such as healthcare setting (e.g. primary, emergency, or secondary care). We also aimed to assess the mean age of participants in each included study at different cut‐off thresholds (e.g. greater than 50 years of age versus greater than 70 years of age) across different studies, as this has previously been shown to influence results.

Methods

Criteria for considering studies for this review

Types of studies

We considered primary diagnostic studies if they compared the results of history taking and physical examination (or both) findings (index test) with a reference standard test for the identification of vertebral fracture in people presenting with low back pain. We included studies that reported results of red flags separately for the different types of vertebral fracture or studies that reported combined results for vertebral fractures. We included studies of cross‐sectional designs that presented sufficient data to allow estimates of diagnostic accuracy (such as sensitivity and specificity) to be derived. We contacted authors of abstracts and conference proceedings for the full text; if the full text was unavailable, we excluded such studies.

We excluded studies with participants diagnosed with pathological vertebral fractures (e.g. malignant fractures). We excluded case‐control studies (i.e. two‐gate studies) that had included some participants (cases) already diagnosed with vertebral fracture and some participants (controls) without vertebral fracture as these types of studies may overestimate both sensitivity and specificity (Whiting 2013).

We retrieved and included studies published in any language. If necessary, we organised appropriate translation of potentially eligible articles.

Participants

We included studies if they evaluated adolescents (from the age of 15 years and over) or adults presenting to primary, secondary, or tertiary care for management of low back pain. We excluded paediatric populations as presentation, risk factors, and fracture types would likely differ compared to adult populations.

Index tests

Studies evaluating any aspects of the history or physical examination (or both) for people presenting with low back pain were eligible for inclusion. This included all information regarding:

• demographic characteristics (e.g. age, gender);

• clinical and medical history (e.g. pain intensity, previous history of falls, previous or current medication use);

• physical examination results (e.g. tenderness on palpation, muscle strength testing, or lumbar range of motion).

We included studies if the diagnostic accuracy of individual red flags was evaluated in isolation, or as part of a combination (e.g. female sex, greater than 70 years of age, significant trauma, and prolonged corticosteroid use). We extracted or derived indices of diagnostic performance from data presented in each primary study for each red flag or combination of red flags. We excluded studies that used only a 'clinical diagnosis' or 'global clinician judgement' (without specifying which diagnostic features were considered) as the diagnostic accuracy of clinical judgement based on undefined parameters is difficult to interpret.

Target conditions

We included studies that investigated people presenting with low back pain or for lumbar examination and reported appropriate data to calculate an estimate of the diagnostic accuracy of tests for vertebral fracture. Where possible, we described results separately for different types of vertebral fracture (i.e. osteoporotic vertebral compression fracture, vertebral traumatic fracture, and vertebral stress fracture).

Where possible, we attempted to categorise the different types of vertebral fractures under the following definitions.

• Acute osteoporotic vertebral compression fracture: insufficiency fractures (i.e. a fracture due to normal stresses on an abnormal bone).

• Vertebral traumatic fracture: any vertebral fracture due to a trauma mechanism in normal bone (e.g. direct force to the lumbar spine).

• Vertebral stress fracture: fractures as a result of repetitive trauma over time (i.e. a fracture due to abnormal stresses on a normal bone).

• Unspecified vertebral fracture: if the type of fractures was not specified we referred to this as unspecified vertebral fracture.

If it was not possible to differentiate between the different types of vertebral fractures in the included studies, then we categorised these fractures as 'unspecified vertebral fractures'.

Reference standards

If studies reported sufficient data, we explored how the different reference standards impacted the diagnostic accuracy of the index test. For traumatic vertebral fractures, CT was used as the preferred reference test. If CT scan was not available, we used MRI as the reference test. If neither was used, we used whichever imaging modality was presented as the reference standard (e.g. X‐ray). For osteoporotic vertebral fractures and vertebral stress fractures, we used MRI as the preferred reference test. If MRI was not available, we used bone scintigraphy with SPECT as the reference test. If neither was used, we used whichever imaging modality was presented as the reference standard (e.g. X‐ray). However, we anticipated that most studies would not separate reference standards based on the different types of vertebral fracture.

We also considered clinical follow‐up after the initial consultation as a weaker reference standard if suspected vertebral fracture was subsequently confirmed by imaging.

Search methods for identification of studies

Electronic searches

The search strategy was based upon the Williams 2013 Cochrane Review and was updated in collaboration with a librarian. Williams 2013 search was run from inception to 7 March 2012. We searched relevant databases for eligible diagnostic studies from January 2012 to 26 July 2022. We searched the following databases.

• MEDLINE (Ovid, from 1946) (Appendix 1)

• Embase (Embase.com; Appendix 2)

• CINAHL (EBSCO; Appendix 3)

Searching other resources

We checked the reference lists of all included publications. We performed forward citation tracking of the included studies through Web of Science, Scopus, and Google Scholar. We sent the final list of included studies for review by experts in the field to minimise the risk of missing any relevant studies.

Data collection and analysis

Selection of studies

Two review authors (CSH and QC) independently applied the selection criteria to all studies (titles and abstracts) identified from the literature search. The two review authors (CSH and QC) independently excluded clearly irrelevant studies. Two review authors (CSH and QC) then retrieved and independently assessed the full‐text publications using the predefined inclusion criteria. We resolved all disagreements through consensus, and consulted other review authors where necessary. We contacted relevant study authors via email if any study information needed to be clarified. All excluded studies are presented with reasons for their exclusion in the Characteristics of excluded studies table and the study selection process is presented in the PRISMA‐DTA diagram (McInnes 2018) (Figure 1).

1.

Study flow diagram.

Data extraction and management

Two review authors (CSH and QC) independently extracted the relevant data from the included studies. This included data on the characteristics of the included studies such as:

• participant characteristics;

• type of vertebral fracture;

• types of index tests and reference standards, and their descriptors;

• relevant aspects of the study methods;

• data for the assessment of diagnostic accuracy.

Characteristics of participants (and studies) included details on:

• healthcare setting (e.g. primary care (medical centre, physiotherapist, chiropractor), emergency departments; secondary care (outpatient hospital service, medical hospital services, specialist services); tertiary care (higher specialised care in hospital setting));

• inclusion and exclusion criteria;

• enrolment procedures (consecutive or non‐consecutive);

• number of participants (including number eligible and enroled in the study);

• number of participants receiving the index test and reference standard;

• number of participants for whom results were reported in the two‐by‐two table;

• reasons for withdrawal;

• participant demographics (age, gender, duration, and history of low back pain).

Test characteristics included details on:

• type of index test;

• methods of execution of index tests;

• threshold for positive result on an index test;

• experience and expertise of the assessors;

• type of reference standard;

• cut‐off points for diagnosing vertebral fracture on a reference standard (e.g. quantitative radiographic measures) where relevant.

Thresholds for 'positive' results on an index test may vary across studies and some studies may present the diagnostic performance of an index test at several different cut‐off points (e.g. age greater than 50 years versus age greater than 70 years). This is presented on a Receiver Operating Characteristic (ROC) curve. We extracted diagnostic accuracy data for all cut‐off points from the included studies.

To assess diagnostic accuracy, we extracted data for diagnostic two‐by‐two tables (true positive, false positive, true negative, and false negative counts) from the publications or reconstructed them using information from other relevant parameters (sensitivity, specificity, or predictive values). If the reported sensitivity, specificity, and positive and negative predictive values did not match the reported two‐by‐two table data we reported results based on the raw two‐by‐two data. However, if there were large discrepancies in the values calculated from raw two‐by‐two data and the values presented in a study, we contacted the study authors to seek clarification of the data. If extracting raw data to produce a two‐by‐two table was not possible, we used the sensitivity and specificity values reported. We discussed any discrepancies in calculated two‐by two data and values presented in a study with the author team and reported them in the review. For eligible studies where we could not reconstruct the diagnostic two‐by‐two tables, we presented them in the review but did not include them in the quantitative analysis. Two review authors (CSH and QC) independently extracted the data to ensure adequate reliability of collected data. For each study, where possible, we presented in tables aspects of study design, characteristics of the population, index test, and reference standards, and type of vertebral fracture. If a review author was also an author of one of the primary diagnostic studies, they were not involved in the data extraction or quality rating of that study.

Assessment of methodological quality

Two review authors (CSH and QC) independently assessed the methodological quality of each included study using the QUADAS‐2 checklist (Whiting 2011), as recommended by the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy (Appendix 4; Deeks 2022). The QUADAS‐2 checklist assessed study quality for all studies included in the previous review and the additional studies included in this updated review. We tailored the QUADAS‐2 tool to our review and made one change by removing "was a case‐control design avoided?" as case control studies were excluded. The QUADAS‐2 checklist consists of four domains: patient selection, index test, reference standard, and flow and timing. We assessed all four domains for risk of bias and all three domains for concerns regarding applicability.

The same review authors classified each item as 'yes' (adequately addressed); 'no' (inadequately addressed); or 'unclear' (inadequate detail presented to allow a judgement to be made). We resolved disagreements through discussion and if necessary, consulted a third review author (CGM or MJH).

Statistical analysis and data synthesis

Statistical analysis followed the approaches outlined in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy (Deeks 2022). We extracted indices of diagnostic performance or derived them from data presented in each primary study for each red flag or combination of red flags. Sensitivity, specificity, and likelihood ratios (LR) of the index tests were the primary measures of diagnostic test accuracy, both in the text and Table 1. Table 1 includes the following variables, but was not limited to: the review question, patient population, index test, target condition, reference standard, study quality, included studies characteristics, sensitivity, specificity, and LRs. We generated diagnostic two‐by‐two tables, from which we calculated and presented sensitivities and specificities for each index test with 95% CIs and present these in forest plots. We calculated LRs for each test and used these as an indication of clinical informativeness. Likelihood ratios represent the increase (with a positive test) or decrease (with a negative test) in likelihood of a person having a vertebral fracture (Deeks 2004). A +LR greater than 1 indicates that a positive test is associated with an increase in the likelihood of a vertebral fracture being present. A negative likelihood ratio (−LR) less than 1 indicates that a negative test is associated with a decrease in the likelihood of a vertebral fracture (Deeks 2004). Criteria for clinical informativeness of LRs are presented in Table 2. In addition, we presented an ROC plot of sensitivity versus 1 − specificity to display the data. If possible, we intended to present results of how different pretest probability cut‐offs and positive or negative results impacted post‐test probabilities.

1. Criteria for study‐specific judgement of the informativeness of tests based on positive likelihood ratios.

Test result (Deeks 2004; Jaeschke 1994)	Interpretation
Upper bound of confidence interval < 2	Uninformative test
Lower bound of confidence interval spans 1 but upper bound > 2	Indeterminate result because we have an imprecise estimate of the test accuracy
Point estimate in the range 2–5 and lower bound of confidence interval is > 1.0	Small increase in likelihood of fracture
Point estimate in the range 5–10 and lower bound of confidence interval is > 1.0	Moderate increase in likelihood of fracture
Point estimate > 10 and lower bound of confidence interval is > 1.0	Large increase in likelihood of fracture

We intended to use the bivariate random‐effects model to compute summary estimates of sensitivity and specificity. The planned method accounts for both within‐study and between‐study variability, and allows for correlation that may exist between sensitivity and specificity across studies (Harbord 2007; Reitsma 2005). This approach would have assumed a binomial distribution within studies (Chu 2006), and we assumed the random effects to be normally distributed across studies. We intended to restrict these analyses to studies that showed sufficient clinical homogeneity (e.g. same index test and index test threshold, and a similar definition of vertebral fracture). Where possible, we intended to perform meta‐analysis on estimates of sensitivity and specificity to calculate the +LR and −LR for each index test.

For index tests where the cut‐off for test positivity varied across studies, we intended to use the Hierarchical Summary ROC (HSROC) method of Rutter and Gatsonis (Rutter 2001) to estimate a summary curve which we presented graphically if appropriate. Where a study had data for more than one cut‐off point, we intended to use the most commonly reported cut‐off point across studies. We initially intended to use SAS software for meta‐analysis, however meta‐analysis was not possible and instead used Review Manager 5 for all statistical analysis and data synthesis (Review Manager 2014).

Investigations of heterogeneity

Several factors can contribute to heterogeneity in the diagnostic accuracy of a test across multiple studies. Where possible, we investigated the potential influences of one study level variable at a time, such as healthcare setting (i.e. primary care, emergency department, secondary care), different types of vertebral fracture (i.e. osteoporotic vertebral compression fracture, vertebral traumatic fracture, and vertebral stress fracture), or different types of imaging. We chose healthcare setting as the prevalence of vertebral fracture may vary depending on the healthcare setting. We chose different types of vertebral fracture as the diagnostic accuracy of red flags may vary dependent on the type of fracture. We intended to perform meta‐analysis on results separately for different healthcare settings and for each of the different types of vertebral fracture if data were available. If results for an index test were not reported separately for different types of fractures we reported the results as the diagnostic accuracy for 'unspecified vertebral fracture'.

Where feasible, we intended to include study‐level covariates (e.g. different types of imaging) in the HSROC analysis models to investigate sources of heterogeneity. Where possible, we clearly acknowledged any unexplained heterogeneity between studies in Table 1, as per recommendations of the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy (Deeks 2022).

Sensitivity analyses

Where possible, we intended to perform a sensitivity analysis to investigate how individual QUADAS‐2 key domains (i.e. patient selection, index test, reference standard, and flow and timing) affected accuracy estimates. That is, we aimed to assess whether estimates of sensitivity and specificity were affected if studies at high risk of bias in any of the QUADAS‐2 domains were removed for the sensitivity analysis.

Assessment of reporting bias

Tests for publication bias in reviews of diagnostic test accuracy are not routinely used because study size may be related to test accuracy for reasons other than publication bias (Deeks 2005). Therefore, we did not report on this.

Results

Results of the search

Our search identified 5878 records, After removing 196 duplicates, 5682 records remained. Of these, 5629 were deemed clearly irrelevant based on screening by title and abstract. There were only two relevant systematic reviews found during the updated search (Downie 2014; Galliker 2019). We reviewed the reference lists of both systematic reviews and performed forward citation searches on the reference lists of all included studies with no additional studies meeting the inclusion criteria. Of the 53 articles included for full‐text review, we excluded 47 with reasons (five had no appropriate data available, three did not assess an index test, four had participants already being diagnosed with vertebral fracture, 21 investigated an inappropriate population, and 14 had an inappropriate study design) (see Characteristics of excluded studies table). Therefore, we included six new studies to add to the eight studies from the previous review (Williams 2013), resulting in 14 studies included for analysis.

Management of missing information

We contacted the authors of two studies for additional data or discrepancies (or both) (Premkumar 2018; Therriault 2020). Therriault 2020 provided some additional data but it was not suitable for our analyses. After contacting the authors again, they confirmed they no longer had access to the raw data we needed for our analysis to calculate the diagnostic accuracy for a combination of red flags (e.g. male sex, pain with lumbar extension, active pain worse than resting pain, or a combination of these). Therefore, we excluded this red flag from our analyses (Therriault 2020).

Description of studies

See Characteristics of included studies table.

Most studies were heterogeneous and did not present data based on the different vertebral fracture types. Four studies reported on 'osteoporotic vertebral fractures' (prevalence of fracture range; 3% to 68%) (Enthoven 2016; Jin 2020; Kilic 2021; Roman 2010), and two studies reported on 'vertebral compression fracture' (prevalence of fracture: Patrick 1983: 7.2%; Reinus 1998: 11%). One study reported on 'osteoporotic and traumatic vertebral fractures' (prevalence of fracture: 4.1%) but did not present data for the individual fracture types (van den Bosch 2004). Five studies did not differentiate between the different types of vertebral fracture and were considered to be reporting on 'unspecified vertebral fracture' (prevalence of fracture range: 0.68% to 6.5%) (Deyo 1986; Gibson 1992; Henschke 2009; Premkumar 2018; Scavone 1981). One study reported on 'early‐stage spondylolysis', and defined spondylolysis as 'the pathogenesis of lumbar spondylolysis is considered to be a stress fracture' (prevalence of fracture: 52.5%) (Sugiura 2021). Another study reported on 'active spondylolysis' and defined spondylolysis as a 'bone stress injury (stress reaction or fracture) of the pars interarticularis' (prevalence of fracture: 22%) (Therriault 2020). Based on each study's definition of spondylolysis, they were both categorised as 'vertebral stress fracture' (Sugiura 2021; Therriault 2020). The overall prevalence of vertebral fracture ranged from 0.68% to 68%; however, the patient sample in some studies was highly selective which likely explains the high prevalence of vertebral fracture. The high prevalence rate is explored further below in 'Findings based on healthcare setting'.

Six studies with 6365 participants were conducted in a primary healthcare setting (Deyo 1986; Enthoven 2016; Henschke 2009; Scavone 1981; Therriault 2020; van den Bosch 2004). The median sample size of the included studies in primary care was 948 (range: 621 to 2007). Of the six studies conducted in primary care, there were 86 potential red flags. One study presented data on a combination of red flags (Therriault 2020). Four studies used X‐ray as the reference standard (Deyo 1986; Enthoven 2016; Scavone 1981; van den Bosch 2004), one study used different types of imaging methods (e.g. X‐ray, MRI, CT, or a combination) based on the clinician's preference as the reference standard (Therriault 2020), and therefore not all participants received the same reference standard, and one study used clinical follow‐up as a reference standard (Henschke 2009).

Five studies with 12,135 participants were conducted in a secondary healthcare setting (Jin 2020; Kilic 2021; Premkumar 2018; Roman 2010; Sugiura 2021). The median sample size of the included studies in secondary care was 510 (range: 101 to 9940). The five studies in secondary care identified 29 red flags. Two studies presented data on a combination of red flags (Premkumar 2018; Roman 2010). Two studies used X‐ray as the reference standard (Kilic 2021; Premkumar 2018), two studies used MRI as the reference standard (Jin 2020; Sugiura 2021), and one study used X‐ray or SPECT as the reference standard (Roman 2010).

Three studies with 1259 participants were conducted in a tertiary healthcare setting (Gibson 1992; Patrick 1983; Reinus 1998). The median sample size of the studies in tertiary care was 482 (range: 225 to 552). The three studies in tertiary care identified 17 red flags. One study presented data on a combination of red flags (Gibson 1992). All three studies used X‐ray as the reference standard.

Methodological quality of included studies

Results of the quality assessment using the QUADAS‐2 scale are presented in Figure 2. Further details on how judgements were made are presented in the Characteristics of included studies table. No studies demonstrated low risk of bias across all domains; however, 10/14 studies demonstrated low concerns regarding applicability.

2.

Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study

Patient selection

For the domain of patient selection, only 3/14 studies demonstrated low risk of bias. Five of 14 studies demonstrated high risk of bias, due to participants being excluded for characteristics of their presentation that could be considered clinically relevant. For example, Jin 2020 and Sugiura 2021 excluded people with low back pain and pain radiating to the limbs which could be considered part of the clinical presentation of people with vertebral fracture (Kim 2015). However, both studies investigated different types of vertebral fracture, Jin 2020 included people with 'osteoporotic vertebral fractures' and Sugiura 2021 included people with 'early‐stage spondylolysis'. Enthoven 2016, Jin 2020, Kilic 2021, and Sugiura 2021 included specific age cut‐offs as part of their inclusion criteria, potentially excluding people presenting with the target condition. The main concern in this domain was unclear risk of bias due to insufficient information regarding inappropriate exclusions in 6/14 studies. Eleven of 14 studies demonstrated low concerns regarding applicability and 3/14 studies demonstrated unclear concerns regarding applicability due to insufficient information available regarding patient presentation.

Index tests

For the domain of index test, 6/14 studies demonstrated low risk of bias. Two of 14 studies demonstrated high risk of bias, due to no prespecified thresholds being used for the index test (Reinus 1998), and index test results being interpreted with the knowledge of the reference standard result (Therriault 2020). Six of 14 studies demonstrated unclear risk of bias, as it was unclear if the index test results were interpreted with or without the knowledge of the reference standard result. All the included studies demonstrated low concerns for applicability.

Reference standards

For the domain of reference standard, only 1/14 studies demonstrated low risk of bias (Jin 2020). Three of 14 studies demonstrated high risk of bias as the reference standard result was interpreted with the knowledge of the index test result. Ten of 14 studies demonstrated unclear risk of bias, as it was unclear if the reference standard results were interpreted with or without the knowledge of the index test result. Thirteen of 14 included studies demonstrated low concerns regarding applicability. One study demonstrated high concern regarding applicability as they used clinical follow‐up, and we considered this a weaker reference standard.

Flow and timing

For the domain of flow and timing, only 1/14 studies demonstrated low risk of bias (Patrick 1983). Five of 14 studies demonstrated high risk of bias, due to inappropriate interval time between index test and the reference standard (i.e. six weeks and greater) (Henschke 2009), not all participants included in the final analysis (i.e. 518/582 participants included for analysis) (Jin 2020), not all participants receiving the same reference test (Roman 2010; Therriault 2020), and one study for all three reasons listed (Deyo 1986). Eight of 14 included studies demonstrated unclear risk of bias, due to most studies not reporting clear interval times between index test and the reference standard.

Overall ratings

In summary, risk of bias was often rated unclear regarding patient selection, implementation of the index test and reference standard, and flow and timing. However, the applicability of the included studies did not demonstrate substantial concerns.

Findings

In total there were 130 potential index tests assessed across 14 studies. Studies used index tests assessing a similar construct but with slightly different wording/cut‐offs (e.g. trauma = significant trauma, history of trauma, major trauma, direct trauma, or fall; older age = older age with eight different age cut‐offs). There were 18 index tests which were presented as different combinations of red flags. Few index test descriptions were reported adequately for clinical use. We attempted to present index test results for the different types of vertebral fractures (i.e. osteoporotic vertebral compression fracture, vertebral traumatic fracture, vertebral stress fracture); however, this was not always possible as only some studies presented data based on the different types of fractures. Performing meta‐analysis on estimates of test accuracy for this updated review was not possible due to the heterogeneity of the index tests presented across the studies, study healthcare setting, and different vertebral fracture types. Therefore, the findings of this review are presented descriptively.

Results for all tests are presented in Table 3, Table 4, Table 5, Table 6, and Table 7. Table 1 describes the performance of red flags that were potentially useful.

2. Results of all red flags taken during a history in primary care.

Study	Index test	Reference standard	Sample size	Prevalence	+LR (95% CI)	−LR (95% CI)
Deyo 1986	Age > 50 years	X‐ray	621	0.045	2.16 (1.58 to 2.95)	0.34 (0.12 to 0.92)
Henschke 2009	Age > 50 years	Follow‐up/imaging	1178	0.007	1.84 (1.07 to 3.17)	0.57 (0.23 to 1.39)
van den Bosch 2004	Age > 54 years	X‐ray	2100	0.041	1.72 (1.54 to 1.91)	0.33 (0.20 to 0.53)
Henschke 2009	Age > 54 years	Follow‐up/imaging	1178	0.007	2.57 (1.49 to 4.44)	0.50 (0.2 to 1.21)
van den Bosch 2004	Age > 64 years	X‐ray	2100	0.041	2.46 (2.16 to 2.8)	0.32 (0.21 to 0.48)
Henschke 2009	Age > 64 years	Follow‐up/imaging	1178	0.007	7.13 (4.04 to 12.59)	0.41 (0.17 to 1.01)
Henschke 2009	Age > 70 years	Follow‐up/imaging	1178	0.007	11.19 (5.33 to 23.51)	0.52 (0.26 to 1.05)
van den Bosch 2004	Age > 74 years	X‐ray	2100	0.041	3.69 (3.00 to 4.53)	0.49 (0.38 to 0.63)
Henschke 2009	Age > 74 years	Follow‐up/imaging	1178	0.007	9.39 (2.69 to 32.75)	0.77 (0.52 to 1.15)
Enthoven 2016	Age ≥ 75 years	X‐ray	669	0.045	3.08 (2.03 to 4.67)	0.64 (0.47 to 0.88)
van den Bosch 2004	Female gender	X‐ray	2100	0.041	1.26 (1.10 to 1.45)	0.65 (0.46 to 0.92)
Enthoven 2016	Female gender	X‐ray	669	0.045	1.12 (0.87 to 1.44)	0.82 (0.50 to 1.34)
Therriault 2020	Female gender	X‐ray, MRI, or SPECT	1025	0.22	0.72 (0.61 to 0.84)	1.41 (1.23 to 1.63)
van den Bosch 2004	Female > 54 years	X‐ray	2100	0.041	2.01 (1.68 to 2.40)	0.54 (0.41 to 0.72)
Henschke 2009	Female > 54 years	Follow‐up/imaging	1178	0.007	5.39 (3.08 to 9.43)	0.42 (0.17 to 1.04)
van den Bosch 2004	Female > 64 years	X‐ray	2100	0.041	2.75 (2.26 to 3.35)	0.52 (0.40 to 0.68)
Henschke 2009	Female > 64 years	Follow‐up/imaging	1178	0.007	14.59 (8.00 to 26.61)	0.39 (0.16 to 0.96)
van den Bosch 2004	Female > 74 years	X‐ray	2100	0.041	4.14 (3.17 to 5.44)	0.62 (0.51 to 0.75)
Henschke 2009	Female > 74 years	Follow‐up/imaging	1178	0.007	16.17 (4.47 to 58.43)	0.76 (0.51 to 1.14)
Deyo 1986	Significant trauma	X‐ray	621	0.045	1.93 (0.67 to 5.53)	0.88 (0.67 to 1.17)
Henschke 2009	Significant trauma	Follow‐up/imaging	1178	0.007	10.03 (2.87 to 35.13)	0.77 (0.52 to 1.15)
Scavone 1981	Significant trauma	X‐ray	871	0.030	12.85 (8.58 to 19.24)	0.36 (0.22 to 0.62)
Enthoven 2016	Trauma	X‐ray	669	0.045	6.42 (2.94 to 14.02)	0.82 (0.68 to 0.97)
Deyo 1986	Corticosteroid use	X‐ray	621	0.045	3.97 (0.20 to 79.15)	0.98 (0.89 to 1.07)
Enthoven 2016	Prolonged corticosteroid use	X‐ray	669	0.045	2.46 (1.13 to 5.34)	0.88 (0.75 to 1.04)
Henschke 2009	Prolonged corticosteroid use	Follow‐up/imaging	1178	0.007	48.50 (11.46 to 204.98)	0.75 (0.51 to 1.13)
Enthoven 2016	Osteoporosis	X‐ray	669	0.045	3.13 (1.90 to 5.15)	0.72 (0.56 to 0.93)
Enthoven 2016	Acute onset of pain	X‐ray	669	0.045	0.87 (0.52 to 1.47)	1.07 (0.85 to 1.35)
Enthoven 2016	Back pain intensity score ≥ 7	X‐ray	669	0.045	1.82 (1.40 to 2.37)	0.53 (0.32 to 0.86)
Therriault 2020	Previous history of LBP	X‐ray, MRI, or SPECT	1025	0.22	0.97 (0.75 to 1.26)	1.01 (0.93 to 1.10)
Therriault 2020	Waking night pain	X‐ray, MRI, or SPECT	1025	0.22	0.96 (0.73 to 1.27)	1.01 (0.94 to 1.09)
Enthoven 2016	Osteoarthritis in hip/knee	X‐ray	669	0.045	0.50 (0.22 to 1.12)	1.22 (1.05 to 1.42)
Enthoven 2016	Thoracic back pain	X‐ray	669	0.045	1.96 (1.28 to 2.99)	0.74 (0.55 to 0.99)

+LR: positive likelihood ratio; −LR: negative likelihood ratio; CI: confidence interval; MRI: magnetic resonance imaging; SPECT: single‐photon emission computerised tomography.

3. Results of all red flags taken during a physical examination in primary care.

Study	Index test	Reference standard	Sample size	Prevalence	+LR (95% CI)	−LR (95% CI)
Henschke 2009	Altered sensation (from trunk down)	Follow‐up/imaging	1178	0.007	3.32 (0.22 to 50.86)	0.96 (0.82 to 1.13)
Scavone 1981	Sensation change	X‐ray	871	0.030	2.21 (1.14 to 4.27)	0.83 (0.66 to 1.05)
Scavone 1981	Motor deficit	X‐ray	871	0.030	2.19 (1.06 to 4.54)	0.86 (0.70 to 1.06)
Scavone 1981	DTR abnormality	X‐ray	871	0.030	1.08 (0.37 to 3.18)	0.99 (0.86 to 1.14)
Scavone 1981	Tenderness	X‐ray	871	0.030	0.70 (0.25 to 1.97)	1.11 (0.87 to 1.41)
Scavone 1981	Spasm	X‐ray	871	0.030	1.25 (0.42 to 3.70)	0.98 (0.85 to 1.12)
Scavone 1981	Sciatica	X‐ray	871	0.030	0.42 (0.06 to 2.91)	1.06 (0.98 to 1.15)
Scavone 1981	Hip/leg pain	X‐ray	871	0.030	0.21 (0.01 to 3.35)	1.08 (1.02 to 1.14)
Enthoven 2016	Severe disability	X‐ray	669	0.045	2.28 (1.26 to 4.14)	0.83 (0.67 to 1.02)
Enthoven 2016	Percussion tenderness of spine	X‐ray	669	0.045	1.12 (0.57 to 2.21)	0.97 (0.81 to 1.16)
Enthoven 2016	Sudden decrease in height	X‐ray	669	0.045	2.89 (0.90 to 9.24)	0.94 (0.84 to 1.05)
Therriault 2020	Radicular symptoms	X‐ray, MRI, or SPECT	1025	0.22	1.91 (0.71 to 5.10)	0.99 (0.97 to 1.01)
Therriault 2020	Numbness or tingling	X‐ray, MRI, or SPECT	1025	0.22	0.69 (0.47 to 1.01)	1.07 (1.01 to 1.13)
Therriault 2020	Lumbar extension	X‐ray, MRI, or SPECT	1025	0.22	1.27 (1.19 to 1.36)	0.40 (0.28 to 0.57)
Therriault 2020	Lumbar flexion	X‐ray, MRI, or SPECT	1025	0.22	1.12 (0.88 to 1.41)	0.96 (0.87 to 1.05)
Therriault 2020	Lumbar rotation	X‐ray, MRI, or SPECT	1025	0.22	0.96 (0.66 to 1.41)	1.01 (0.95 to 1.07)
Therriault 2020	Lumbar lateral flexion	X‐ray, MRI, or SPECT	1025	0.22	1.15 (0.85 to 1.55)	0.97 (0.90 to 1.04)
Therriault 2020	Single‐leg hyperextension test	X‐ray, MRI, or SPECT	1025	0.22	1.32 (1.25 to 1.40)	0.19 (0.11 to 0.33)
Therriault 2020	Vertebral tenderness	X‐ray, MRI, or SPECT	1025	0.22	0.84 (0.58 to 1.23)	1.03 (0.97 to 1.09)

+LR: positive likelihood ratio; −LR: negative likelihood ratio; CI: confidence interval; DTR: deep tendon reflex; MRI: magnetic resonance imaging; SPECT: single‐photon emission computerised tomography.

4. Results of all red flags in secondary care.

Study	Index test	Reference standard	Sample size	Prevalence	+LR (95% CI)	−LR (95% CI)
Roman 2010	Age > 52 years	X‐ray or CT	1448	0.026	1.52 (1.42 to 1.68)	0.14 (0.04 to 0.53)
Premkumar 2018	Age > 50 years	Imaginga	9940	0.056	1.10 (1.05 to 1.16)	0.79 (0.69 to 0.91)
Premkumar 2018	Age > 70 years	Imaginga	9940	0.056	1.55 (1.36 to 1.76)	0.86 (0.82 to 0.91)
Kilic 2021	Age 65–75 years	X‐ray	136	0.544	0.60 (0.46 to 0.79)	2.51 (1.48 to 4.27)
Kilic 2021	Age ≥ 75 years	X‐ray	136	0.544	2.51 (1.48 to 4.27)	0.60 (0.46 to 0.79)
Roman 2010	Female gender	X‐ray or CT	1448	0.026	1.51 (1.35 to 1.71)	0.26 (0.10 to 0.65)
Sugiura 2021	Female gender	MRI	101	0.525	0.53 (0.27 to 1.05)	1.26 (0.98 to 1.61)
Roman 2010	Concomitant osteoarthritis	X‐ray or CT	1448	0.026	1.05 (0.76 to 1.45)	0.95 (0.69 to 1.32)
Roman 2010	No regular exercise	X‐ray or CT	1448	0.026	1.47 (1.25 to 1.75)	0.42 (0.21 to 0.81)
Roman 2010	Absence of leg or buttock pain	X‐ray or CT	1448	0.026	2.24 (1.38 to 3.64)	0.80 (0.64 to 0.99)
Roman 2010	BMI < 23	X‐ray or CT	1448	0.026	2.22 (1.44 to 3.42)	0.76 (0.59 to 0.97)
Roman 2010	Decreased pain on sitting	X‐ray or CT	1448	0.026	1.56 (0.94 to 2.59)	0.87 (0.71 to 1.07)
Roman 2010	No gait abnormality	X‐ray or CT	1448	0.026	0.85 (0.68 to 1.08)	1.49 (0.95 to 2.34)
Jin 2020	BPIT	MRI	510	0.68	3.09 (2.47 to 3.85)	0.01 (0.00 to 0.04)
Premkumar 2018	Trauma	Imaginga	9940	0.056	2.18 (1.86 to 2.54)	0.85 (0.81 to 0.89)
Sugiura 2021	Participation in sports activities	MRI	101	0.525	1.05 (0.97 to 1.12)	0.18 (0.01 to 3.61)
Sugiura 2021	Lumbar hyperextension	MRI	101	0.525	0.86 (0.73 to 1.02)	2.49 (0.85 to 7.28)
Sugiura 2021	Lumbar hyperflexion	MRI	101	0.525	0.62 (0.45 to 0.86)	2.40 (1.30 to 4.42)
Sugiura 2021	Pain quality	MRI	101	0.525	1.38 (0.97 to 1.96)	0.58 (0.32 to 1.04)
Sugiura 2021	Pain extent	MRI	101	0.525	2.05 (1.22 to 3.44)	0.59 (0.40 to 0.85)
Sugiura 2021	Pain location	MRI	101	0.525	2.40 (1.63 to 3.54)	0.18 (0.08 to 0.39)

^a Imaging type not reported.
+LR: positive likelihood ratio; −LR: negative likelihood ratio; BMI: body mass index; BPIT: Back Pain‐Inducing Test; CI: confidence interval; CT: computed tomography; MRI: magnetic resonance imaging.

5. Results of red flags in tertiary care.

Study	Index test	Reference standard	Sample size	Prevalence	+LR	−LR
Gibson 1992	History of direct trauma	X‐ray	225	0.06	1.93 (1.48 to 2.52)	0.12 (0.01 to 1.79)
Patrick 1983	Presenting history of trauma	X‐ray	552	0.07	1.77 (1.48 to 2.13)	0.36 (0.20 to 0.68)
Reinus 1998	History of trauma (fall)	X‐ray	482	0.11	0.18 (0.07 to 0.48)	1.54 (1.38 to 1.71)
Patrick 1983	Sensory deficit	X‐ray	552	0.07	1.42 (0.19 to 10.95)	0.99 (0.94 to 1.04)
Patrick 1983	Motor deficit	X‐ray	552	0.07	1.39 (0.08 to 25.38)	0.98 (0.96 to 1.03)
Patrick 1983	DTR abnormality	X‐ray	552	0.07	1.54 (0.49 to 4.87)	0.97 (0.89 to 1.06)
Gibson 1992	Neurological signs or SLR < 40°, or both	X‐ray	225	0.06	2.40 (0.67 to 8.70)	0.81 (0.51 to 1.30)
Reinus 1998	Any motor or sensory dysfunctiona	X‐ray	482	0.11	0.69 (0.22 to 2.17)	1.03 (0.96 to 1.10)
Patrick 1983	Positive SLR	X‐ray	552	0.07	1.02 (0.51 to 2.05)	1.00 (0.86 to 1.16)
Patrick 1983	Tenderness	X‐ray	552	0.07	1.76 (1.42 to 2.19)	0.47 (0.28 to 0.78)
Patrick 1983	Spasm	X‐ray	552	0.07	1.47 (0.83 to 2.60)	0.90 (0.75 to 1.09)
Patrick 1983	Contusion/abrasion	X‐ray	552	0.07	31.09 (18.25 to 52.96)	0.15 (0.07 to 0.32)

^a Weakness, numbness, or paraesthesia present on testing or questioning
+LR: positive likelihood ratio; −LR: negative likelihood ratio; CI: confidence interval; SLR: straight leg raise.

6. Results of combined red flags in primary, secondary, and tertiary care.

Study	Index test	Reference standard	Sample size	Prevalence	+LR	−LR
Premkumar 2018	Age > 50 years and trauma	Imaginga	9940	0.056	2.54 (2.05 to 3.16)	0.90 (0.87 to 0.94)
Premkumar 2018	Age > 70 years and trauma	Imaginga	9940	0.056	4.35 (2.92 to 6.48)	0.96 (0.94 to 0.98)
Gibson 1992	Trauma and neurological signs	X‐ray	225	0.06	14.43 (2.38 to 87.64)	0.73 (0.46 to 1.15)
Patrick 1983	Multiple findings	X‐ray	552	0.07	2.01 (1.35 to 2.99)	0.73 (0.56 to 0.96)
Henschke 2009	1 of 4 red flags +ve	Follow‐up/imaging	1178	0.007	1.75 (1.34 to 2.29)	0.25 (0.04 to 1.57)
Henschke 2009	2 of 4 red flags +ve	Follow‐up/imaging	1178	0.007	15.48 (8.45 to 28.36)	0.39 (0.16 to 0.96)
Henschke 2009	3 of 4 red flags +ve	Follow‐up/imaging	1178	0.007	906.11 (50.37 to 16,299.11)	0.61 (0.36 to 1.03)
Roman 2010	1 of 5 red flags +ve	X‐ray or CT	1448	0.026	1.04 (0.98 to 1.10)	0.43 (0.06 to 2.98)
Roman 2010	2 of 5 red flags +ve	X‐ray or CT	1448	0.026	1.43 (1.32 to 1.56)	0.16 (0.04 to 0.60)
Roman 2010	3 of 5 red flags +ve	X‐ray or CT	1448	0.026	2.45 (2.02 to 2.97)	0.34 (0.19 to 0.61)
Roman 2010	4 of 5 red flags +ve	X‐ray or CT	1448	0.026	9.62 (5.88 to 15.73)	0.66 (0.52 to 0.84)
Roman 2010	5 of 5 red flags +ve	X‐ray or CT	1448	0.026	7.63 (0.91 to 63.67)	0.98 (0.93 to 1.03)

^aImaging type not reported.
+LR: positive likelihood ratio; −LR: negative likelihood ratio; CT: computed tomography.

Findings based on healthcare setting

Primary care

Six studies were based in a primary healthcare setting and investigated 86 different index tests as potential red flags (Deyo 1986; Enthoven 2016; Henschke 2009; Scavone 1981; Therriault 2020; van den Bosch 2004). The potential red flags investigated included components of the history (e.g. age, gender, mechanism of injury) and physical examination (e.g. palpation, lumbar range of motion) and are presented below descriptively. Two studies reported on screening for 'osteoporotic vertebral fracture' (Enthoven 2016; van den Bosch 2004), one study reported on 'vertebral stress fracture' (Therriault 2020), and three studies reported results for 'unspecified vertebral fracture' (Deyo 1986; Henschke 2009; Scavone 1981). Four of the six studies included participants across a wide range of ages (0 to 98 years), however, Therriault 2020 included participants between the ages of 10 and 19 years and Enthoven 2016 only included adults over the age of 55 years. Any index tests investigated in Enthoven 2016 was in addition to the age cut‐off used for inclusion to their study (i.e. over 55 years). For example, 'trauma' was investigated as a potential red flag, therefore the index test would be age over 55 years and 'trauma', not trauma alone.

Four studies investigated 'trauma' as a potential red flag (unspecified vertebral fracture: three studies, osteoporotic fracture: one study) (Deyo 1986; Enthoven 2016; Henschke 2009; Scavone 1981). The sensitivity for the index test 'trauma' for 'unspecified vertebral fracture' was generally low and varied greatly, ranging from 0.21 (95% CI 0.05 to 0.51) to 0.65 (95% CI 0.44 to 0.83) (Deyo 1986; Henschke 2009; Scavone 1981). However, 'trauma' was highly specific in all studies, ranging from 0.90 (95% CI 0.86 to 0.93) to 0.98 (95% CI 0.96 to 0.98). The +LRs ranged from 1.93 (95% CI 0.67 to 5.53) to 12.85 (95% CI 8.58 to 19.24) and the −LRs ranged from 0.36 (95% CI 0.22 to 0.62) to 0.88 (95% CI 0.67 to 1.17). Enthoven 2016 investigated 'trauma' as a red flag for 'osteoporotic vertebral fracture', but only included people over the age of 55 years; and demonstrated low sensitivity 0.21 (95% CI 0.09 to 0.39) but high specificity 0.97 (95% CI 0.95 to 0.98). The +LR 6.42 (95% CI 2.94 to 14.02) demonstrated informative diagnostic value, but uninformative −LR (0.82, 95% CI 0.68 to 0.97).

Four studies investigated 'older age' as a potential red flag (unspecified vertebral fracture: two studies, osteoporotic fracture: two studies), with eight different age cut‐offs (Deyo 1986; Enthoven 2016; Henschke 2009; van den Bosch 2004). Diagnostic accuracy values for the different age cut‐offs are presented in Table 3. The cut‐off age of 'age greater than 70 years' demonstrated the highest diagnostic accuracy for 'unspecified vertebral fracture' with a sensitivity of 0.50 (95% CI 0.16 to 0.84), specificity of 0.96 (95% CI 0.94 to 0.97), +LR of 11.19 (95% CI 5.33 to 23.51), and −LR of 0.52 (95% CI 0.26 to 1.05) (Henschke 2009). One study investigated the 'older age' and 'gender' as part of a combination of red flags, which demonstrated higher diagnostic accuracy compared to older age alone. For example, when 'age greater than 74 years' and 'female gender' were combined, the +LR of a suspected 'unspecified vertebral fracture' increased from 9.39 (95% CI 2.69 to 32.75) to 16.17 (95% CI 4.47 to 58.43) (Henschke 2009).

Three studies investigated 'corticosteroid use' as a potential red flag (unspecified vertebral fracture: two studies, osteoporotic fracture: one study) (Deyo 1986; Enthoven 2016; Henschke 2009). The sensitivity for 'corticosteroid use' for 'unspecified vertebral fracture' was low and ranged from 0.00 (95% CI 0.00 to 0.23) to 0.25 (95% CI 0.03, 0.65), however demonstrated high specificity, ranging from 0.99 (95% CI 0.98 to 1.00) to 0.99 (95% CI 0.99 to 1.00) (Deyo 1986; Henschke 2009). The +LR varied greatly, ranging from 3.97 (95% CI 0.20 to 79.15) to 48.50 (95% CI 11.46 to 204.98), and the −LR ranged from 0.75 (95% CI 0.51 to 1.13) to 0.98 (95% CI 0.89 to 1.07). One study investigated 'corticosteroid use' as a red flag for 'osteoporotic vertebral fracture', but only included people over the age of 55 years (Enthoven 2016). It demonstrated low sensitivity (0.18, 95% CI 0.07 to 0.35) but high specificity (0.93, 95% CI 0.90 to 0.95). The +LR demonstrated informative diagnostic value (2.46, 95% CI 1.13 to 5.34), but uninformative −LR (0.88, 95% CI 0.75 to 1.04).

Two studies investigated a combination of index tests as red flags (unspecified vertebral fracture: one study, stress fracture: one study) (Henschke 2009; Therriault 2020). In Henschke 2009, the presence of at least two of four red flags (older age, gender, corticosteroid use, and history of trauma) for the suspicion of a 'unspecified vertebral fracture' demonstrated moderate sensitivity (0.63, 95% CI 0.24 to 0.91) and high specificity (0.96, 95% CI 0.95 to 0.97). Although there were large +LRs (15.48, 95% CI 8.45 to 28.36) when two of the four red flags were present, and for three of the four red flags present (906.11, 95% CI 50.37 to 16,299.10), the CIs were very wide and, therefore, these results should be interpreted with caution.

Secondary care

Five studies were based in a secondary healthcare setting and investigated 29 different index tests as potential red flags (Jin 2020; Kilic 2021; Premkumar 2018; Roman 2010; Sugiura 2021). These included components of the history (e.g. age, gender, mechanism of injury) and physical examination (e.g. lumbar range of motion) and are presented below descriptively. Three studies reported on screening for 'osteoporotic vertebral fracture' (Jin 2020; Kilic 2021; Roman 2010), one study reported on 'vertebral stress fracture' (Sugiura 2021), and one study reported on 'unspecified vertebral fracture' (Premkumar 2018). Three of five studies included age as part of their inclusion criterion and as an index test. Kilic 2021 only included geriatric adults over the age of 65 years. Sugiura 2021 only included adolescents under the age of 18 years. Index tests investigated in Sugiura 2021 would only apply to the age cut‐off used for inclusion to their study. Jin 2020 only included postmenopausal women or women over 50 years old and men over 60 years old or people with a history of prolonged use of a glucocorticosteroid for three months or more. Participants with previous vertebral fractures were also included in this study (Jin 2020). Index tests investigated in Jin 2020 would apply to the participants who fit the 'age' cut‐offs and 'corticosteroid use' used for inclusion to their study.

Three studies investigated 'older age' as a potential red flag (osteoporotic fracture: two studies, unspecified vertebral fracture: one study), with five different age cut‐offs (Kilic 2021; Premkumar 2018; Roman 2010). Diagnostic accuracy values for the different age cut‐offs are presented in Table 5. The cut‐off age of 'age greater than 74 years' demonstrated the most useful diagnostic accuracy values for 'osteoporotic vertebral fracture' with a sensitivity of 0.53 (95% CI 0.41 to 0.64), specificity of 0.79 (95% CI 0.67 to 0.88), +LR of 2.51 (95% CI 1.48 to 4.27), and −LR of 0.60 (95% CI 0.46 to 0.79) (Kilic 2021). One study investigated 'trauma' as a potential red flag and demonstrated informative diagnostic accuracy values for 'unspecified vertebral fracture' with a sensitivity of 0.25 (95% CI 0.21 to 0.29), specificity of 0.89 (95% CI 0.88 to 0.89), +LR of 2.18 (95% CI 1.86 to 2.54), and −LR of 0.85 (95% CI 0.81 to 0.89) (Premkumar 2018). However, when 'older age' and 'trauma' were investigated as part of a combination of index tests for 'unspecified vertebral fracture', the diagnostic values increased compared to 'older age' or 'trauma' alone. For example, when 'older age greater than 70 years' and trauma were combined, the +LR of a suspected fracture increased from 1.55 (95% CI 1.36 to 1.76) to 4.35 (95% CI 2.92 to 6.48) (Premkumar 2018).

Two studies investigated individual components of the history (e.g. pain location, pain quality, etc.; Sugiura 2021) and physical examination (e.g. lumbar range of motion; Jin 2020; Sugiura 2021) as red flags (osteoporotic fracture; one study, stress fracture; one study). Jin 2020 investigated a series of dynamic movements (e.g. moving from lying to sitting) as a red flag for 'osteoporotic vertebral fracture'. This test demonstrated high sensitivity (0.99, 95% CI 0.97 to 1.00) and moderate specificity (0.68, 95% CI 0.60 to 0.75). The results of this index test should be interpreted with caution as the index test investigated dynamic movements is a combination with other 'index tests' used as part of their study inclusion criteria as mentioned above. The sample of participants in Jin 2020 was also highly selective which likely explains the high prevalence of vertebral fracture in this study (i.e. prevalence of vertebral fracture was 68%). In Sugiura 2021, the only index test with some diagnostic value was 'pain location' (i.e. pain on the left or right (or both) side or central pain) as a red flag for 'vertebral stress fracture'. This test demonstrated high sensitivity (0.89, 95% CI 0.77 to 0.96) and moderate specificity (0.63, 95% CI 0.48 to 0.77). The results of this index test should be interpreted with caution as the index test investigated (i.e. pain location) is in combination with 'age of 18 years or younger' which was used as part of their study inclusion criteria as mentioned above. The prevalence of a 'vertebral stress fracture' in Sugiura 2021 was also high (i.e. prevalence of spondylolysis was 53%), however previous research has shown this population sample is likely to have a high prevalence of spondylolysis due to their young age and participation in sporting activities (Selhorst 2019).

One study investigated eight different index tests as potential red flags for 'osteoporotic vertebral fracture' (Roman 2010); however, the diagnostic values of these tests varied greatly. Index tests with some diagnostic value included body mass index (BMI) less than 23, decreased pain on sitting, and no leg or buttock pain. The sensitivity of these three tests ranged from 0.29 (95% CI 0.15 to 0.46) to 0.37 (95% CI 0.22 to 0.54) and the specificity ranged from 0.81 (95% CI 0.79 to 0.83) to 0.86 (95% CI 0.84 to 0.88). These three index tests also demonstrated the largest +LRs; BMI less than 23 (2.22, 95% CI 1.44 to 3.42), decreased pain on sitting (1.56, 95% CI 0.94 to 2.59), and no leg or buttock pain (2.24, 95% CI 1.38 to 3.64). This study also investigated a combination of index tests (age greater than 52 years, no presence of leg pain, BMI less than 23, does not exercise regularly, and female gender) and found that having four of the five positive tests demonstrated a moderate increase in the likelihood of 'osteoporotic vertebral fracture' (+LR 9.62, 95% CI 5.88 to 15.73).

Tertiary care

Three studies were based in a tertiary healthcare setting and investigated 17 different index tests as potential red flags (Gibson 1992; Patrick 1983; Reinus 1998). The red flags investigated included components of the subjective history (e.g. mechanism of injury) and physical examination (e.g. palpation, neurological assessment) and are presented below descriptively. Two studies reported on screening for 'vertebral compression fracture' (Patrick 1983; Reinus 1998), and one study reported on 'unspecified vertebral fracture' (Gibson 1992).

Three studies investigated 'trauma' as a potential red flag (compression fracture: two studies, unspecified fracture: one study) (Gibson 1992; Patrick 1983; Reinus 1998). For 'vertebral compression fracture' (Patrick 1983; Reinus 1998) and 'unspecified vertebral fracture' (Gibson 1992), the diagnostic accuracy values varied greatly across the three studies and, therefore, meta‐analysis was not performed. The sensitivity for the index test 'trauma' varied greatly from 0.07 (95% CI 0.02 to 0.18) to 1.00 (95% CI 0.59 to 1.00) and the specificity ranged from 0.51 (95% CI 0.41 to 0.62) to 0.60 (95% CI 0.56 to 0.65). The +LRs were also inconsistent ranging from 0.18 (95% CI 0.07 to 0.48) to 1.93 (95% CI 1.48 to 2.52).

One study investigated a range of index tests including components of subjective history and physical examination as potential red flags for 'vertebral compression fracture' (Patrick 1983). This study demonstrated one index test with some diagnostic value. The 'presence of contusion or abrasion' in an emergency department setting demonstrated high sensitivity (0.85, 95% CI 0.70 to 0.94) and specificity (0.97, 95% CI 0.95 to 0.98) indicating some diagnostic value. The +LR also demonstrated high diagnostic value (31.09, 95% CI 18.25 to 52.96). However, these results should be interpreted with caution as they were from a single study.

Discussion

Summary of main results

This review is an update of a previous Cochrane Review of red flags to screen for vertebral fracture in people with low back pain (Williams 2013). Index tests included components of history taking or physical examination and included studies were in primary, secondary, and tertiary healthcare settings.

Our review only included six additional studies (total 14) investigating red flags to screen for vertebral fracture compared to the previous review. Meta‐analysis of studies was not possible due to heterogeneity of the data and results from single studies suggested only a few of the red flags investigated may be informative. In the primary healthcare setting, results from single studies suggest 'trauma' demonstrated informative +LRs for 'unspecified vertebral fracture' and 'osteoporotic vertebral fracture'. Results from single studies suggest 'older age' demonstrated informative +LRs for studies in primary care for 'unspecified vertebral fracture'. Results from single studies suggest 'corticosteroid use' may be an informative red flag in primary care for 'unspecified vertebral fracture' and 'osteoporotic vertebral fracture'; however, diagnostic values varied, and CIs were imprecise. Results from a single study suggest red flags as part of a combination of index tests such as 'older age and female gender' in primary care demonstrated informative +LRs for 'unspecified vertebral fracture'. In the secondary healthcare setting, results from a single study suggest 'trauma' and 'older age' demonstrated informative +LRs for 'unspecified vertebral fracture' and 'osteoporotic vertebral fracture', respectively. Results from a single study suggest red flags as part of a combination of index tests such as 'older age and trauma' in secondary care demonstrated informative +LRs for 'unspecified vertebral fracture'. When four of five tests were positive in secondary care, they demonstrated informative +LRs for 'osteoporotic vertebral fracture'. In the tertiary care setting, results from a single study suggest 'presence of contusion/abrasion' was informative for 'vertebral compression fracture'.

Studies included in our review presented data on the current red flags recommended in guidelines to screen for vertebral fracture (e.g. osteoporosis, history of trauma, corticosteroid use, older age, and female gender) (Verhagen 2016). However, only one study presented data on osteoporosis despite its recommendation in current guidelines. Trauma, older age, corticosteroid use, and presence of contusion/abrasion were suggestive of informative +LRs in our study; however, CIs for corticosteroid use varied greatly. Results for female gender did not demonstrate any informative diagnostic values in our study. Current guidelines do not report on any specific red flags for vertebral stress fracture and only two studies in our review presented data on this type of fracture (Sugiura 2021; Therriault 2020).

A combination of red flags was more informative than red flags used in isolation. For example, when 'older age' (i.e. age greater than 74 years) and 'female gender' were combined the +LR increased from 9.39 (95% CI 2.69 to 32.75) to 16.17 (95% CI 4.47 to 58.43). When 'older age' (i.e. age greater than 70) and 'trauma' were combined, the +LR increased from 1.55 (95% CI 1.36 to 1.76) to 4.35 (95% CI 2.92 to 6.48). Another combination of index tests (age greater than 52 years, no presence of leg pain, BMI less than 23, does not exercise regularly, and female gender) found large +LRs if four of the five tests were positive (9.62, 95% CI 5.88 to 15.73). However, these findings should be interpreted with caution as they were based on single studies and there was a noticeable lack of available data in this field. Of the available data in our review, often +LR had wide CIs and imprecise estimates. Uncertainty in the risk of bias across the included studies should also be considered when interpreting results.

Many red flags used to screen for vertebral fracture have high false‐positive rates and are currently based on limited data. Given the prevalence of serious pathology in primary care (1% to 5%) (Downie 2014; Hartvigsen 2018; Henschke 2009; Williams 2013), and in emergency departments (2.5% to 5.1%) (Galliker 2019) is quite low, the limited evidence makes it challenging for clinicians to identify vertebral fractures in people with low back pain. Clinicians also face the challenges of considering the consequences of unnecessary imaging and its associated impact on patient outcomes and patient costs.

Population and healthcare setting

The aim of our review was to identify red flags to assist clinicians to screen for vertebral fracture in people with low back pain. Although a few studies reported on the same or similar index test, many studies provided data based on different healthcare settings, different vertebral fracture types, or demonstrated large variability in diagnostic values, making meta‐analysis of included data inappropriate. We included studies from different healthcare settings; however, based on the results of our review, most red flags recommended to screen for vertebral fracture as a cause of low back pain are not well‐supported in any healthcare setting due to the lack of studies in this field and because most red flags identified have poor diagnostic value or have only been evaluated in single studies.

There was no clear pattern of differences between the studies conducted in primary and secondary care in the patient population and type of vertebral fracture investigated. The primary and secondary care studies included a mix of patient populations, with some studies only including older participants (e.g. Enthoven 2016; Jin 2020; Kilic 2021), and other studies only including younger populations (e.g. Sugiura 2021; Therriault 2020). Studies in primary and secondary care also investigated a variety of vertebral fractures such as 'osteoporotic vertebral fracture', 'stress fracture', and 'unspecified fracture'. Of the three studies conducted in tertiary care, two studies investigated 'vertebral compression fracture' and the patient populations were of similar age ranges across the three studies. Five of the 14 studies did not specify the type of vertebral fracture.

In the primary healthcare setting, the risk of bias of the included studies was not consistent in this domain (i.e. some studies had a risk of bias that was low, high, or unclear); however, the risk of bias was consistently unclear across most studies in the secondary and tertiary healthcare setting. It was evident that in four studies verification had occurred as in these studies there was a failure to compare the index tests and reference standards in all participants. This may have overestimated the accuracy of the index tests investigated.

Reference standard

Most studies used an adequate reference standard according to our methodological quality criteria (Appendix 4), except one study (Henschke 2009). This study demonstrated high concerns of applicability as clinical follow‐up of participants with suspected vertebral fracture was used as the reference standard. It is possible that some participants with vertebral fracture in this study may have been missed, which would affect the estimates of diagnostic accuracy. Another study used a combination of different types of imaging methods (e.g. X‐ray, MRI, CT, or a combination of these) based on the clinician's clinical preference as the reference standard (Therriault 2020). Although all participants received some form of imaging, not all participants received the same (form of imaging) reference standard.

Most studies used X‐ray as the reference standard, however one of the main limitations of X‐rays is that they are often unable to identify the acuity of a vertebral fracture (i.e. whether a fracture is acute or non‐acute), whilst MRI has the ability to detect bone marrow oedema and more accurately distinguish acute versus chronic fractures (Ang 2016; Astur 2015). However, MRI has its limitations due to contraindications (e.g. indwelling electrical devices), feasibility, access, and cost (Kim 2022). CT may be used for detection of vertebral fractures, but often fails to determine the acuity or chronicity of fractures because of variability in fracture morphology, inhomogeneous background bone marrow attenuation, and different degrees of bone marrow oedema (Kim 2022). Another issue is that vertebral fractures such as compression fractures may often be incidental findings on CT, with some people being either asymptomatic or having symptoms not severe enough to seek medical care (Urrutia 2019).

Most studies also did not report on the method of imaging (e.g. X‐ray views used, experience of the radiographer, thresholds for identifying vertebral fractures, sequences of imaging). Therefore, the influence of different imaging methods could not be investigated. All reference standards (e.g. X‐ray, MRI, CT scan) available would likely be similar across all healthcare settings as these tests could be ordered in primary, secondary, or tertiary care. Given that all the reference standards are available across all healthcare settings, this would have little to no impact on concerns for applicability (i.e. there is low concern for applicability).

Index tests

If red flags have sufficiently high diagnostic accuracy their presence would indicate that further investigation is necessary to confirm serious pathology. If red flags are absent, then this should reduce the clinician's suspicion of serious pathology and negate the need for patients to undergo further investigation. Unfortunately, the accuracy of most red flags is low or uncertain (or both) so this use of red flags to guide diagnostic work‐up is challenged.

The results of our review are based on a very small number of studies and currently do not support most of the red flags investigated. The sensitivity of most index tests was generally low, indicating that when vertebral fracture is present, there is a high number of people with vertebral fracture who will also go undiagnosed. Most tests also had small‐to‐modest +LRs and only a few index tests would increase the suspicion of having a vertebral fracture above the reported prevalence. In addition, the specificity of some of the informative index tests (e.g. older age, corticosteroid use, trauma, contusion/abrasion, a combination of some index tests) was high suggesting that if these tests are positive, it is likely that a fracture is present. However, the −LRs of most index tests in our review were not informative and are unlikely to be useful in clinical practice. Given the low prevalence of vertebral fracture, the consequences of being unable to rule out vertebral fractures using red flags may not be as detrimental as it is unlikely a negative test will meaningfully lower the probability of fracture any further or shift the provability of vertebral fracture.

The results of our review suggest some index tests were informative; however, these results were based on single studies. Trauma, older age, and corticosteroid use were suggestive of informative +LRs; however, the estimates were quite imprecise (e.g. +LRs 48.50, 95% CI 11.46 to 204.98 for corticosteroid use) and so should be interpreted with caution. The imprecision is likely due to the low number of vertebral fractures identified in people presenting with low back pain. Recently, The International Federation of Orthopaedic Manipulative Physical Therapists (IFOMPT) developed an international framework for red flags for potential serious spinal pathologies (Finucane 2020). The recommendations of this framework were in line with the results of our review that trauma, older age, and corticosteroid use should increase clinical suspicion of vertebral fracture. The framework also recommended that history of osteoporosis, female sex, and previous spinal fracture should increase suspicion of vertebral fracture. In our review, history of osteoporosis demonstrated some informativeness (+LR 3.13, 95% CI 1.90 to 5.15) (Enthoven 2016); however, we did not find evidence to support the use of female sex and no studies in our review reported on previous spinal fracture. Clinicians may have to rely on frameworks such as the guidelines developed by IFOMPT as a guide for screening for vertebral fractures as there are currently very little primary data in this field. This creates challenges for researchers and clinicians to utilise red flags confidently in clinical practice.

Strengths and weaknesses of the review

A strength of this review is that we used the same search strategy used by Williams 2013, which was a sensitive and systematic search performed to include all possible studies. Our search identified previous systematic reviews on red flags for people presenting with low back pain (Downie 2014; Galliker 2019), and our search identified all relevant studies in those reviews.

A limitation of our review is the heterogeneous nature of the included studies and the paucity of diagnostic research evaluating red flags to screen for vertebral fracture. There were many index tests investigated by single studies, but few studies investigated the same index tests. When multiple studies did investigate the same index tests there was large variability in the diagnostic values between studies, which limited our ability to perform meta‐analysis of diagnostic values. There was also little uniform agreement on the definition of the index tests, which may explain the diagnostic accuracy of such tests. For example, dose and duration of corticosteroid use is often not reported, yet this may influence the risk of certain vertebral fractures (e.g. osteoporotic vertebral fracture) (Vestergaard 2008). The acuity of osteoporotic vertebral compression fracture needs to be considered as prevalence estimates may differ in those with acute versus non‐acute osteoporotic compression fracture. The studies investigating osteoporotic compression fractures did not report acuity and identifying the acuity of this type of fracture is often difficult due to factors such as the imaging modality used, scoring method used, and population investigated (Lentle 2019; Oei 2013). Many single studies investigated the same index tests across different healthcare settings, used different study designs, or presented diagnostic values for different types of vertebral fracture, which meant performing meta‐analysis for diagnostic accuracy was not appropriate. Due to the limitations of the available evidence in our review, it was not possible to investigate the diagnostic accuracy of index tests or investigate potential sources of heterogeneity on the diagnostic accuracy of potential red flags. Some studies also included a highly selective sample and used an index test as part of both their inclusion criteria and as a red flag. For example, Enthoven 2016 only included adults over the age of 55 years and investigated trauma as an index test, therefore the index test investigated would be more akin to age over 55 years and 'trauma', not trauma alone. Studies that included a highly selective sample inflate prevalence estimates and this likely explains the high prevalence of vertebral fracture in those studies and may also be further influenced by the healthcare setting the study is performed in and the different types of vertebral fracture.

Two review authors independently performed the risk of bias assessment for this review using clear guidelines set a priori to minimise differences in scoring (Appendix 4). Most studies did not provide enough detail or information to make a clear decision on bias and were therefore scored at unclear risk of bias (Figure 2). Future research in this field needs to focus on higher quality and reporting in areas such as clarifying aspects of patient selection, standardising reference standards used across studies, and clarifying the interval time between the index test performed and the reference standard.

Applicability of findings to the review question

The current literature recommends that routine imaging for low back pain in clinical practice is not appropriate and should only be performed in patients where there is suspicion of serious pathology (Foster 2018). Unfortunately, due to the paucity of new studies and primary data in index tests to screen for vertebral fracture, it remains unclear which red flags are useful when screening for vertebral fracture in people presenting with low back pain. The limited research in this field may be a factor in the current high use of imaging in primary care. In our review, there was a total of 130 individual red flags that were investigated across 14 studies with very few studies investigating the same index test. Given the large number of index tests available to clinicians, it is understandable that many clinicians may refer for further imaging or investigations. Inappropriate or excessive imaging leads to excessive costs, potentially unnecessary radiation exposure, and can also lead to negative downstream consequences (i.e. more invasive procedures, poorer outcomes, etc.) (Jacobs 2020). However, there are consequences to also not using red flags in clinical practice with the potential for misdiagnosis of vertebral fracture, which is further complicated by the low prevalence of vertebral fracture in primary care. Without robust evidence clinicians will continue to face the challenge of being selective when referring patients for further investigations. The results from our review may guide clinicians when screening for vertebral fracture, as we found trauma, older age, corticosteroid use, history of osteoporosis, or a combination of red flags demonstrated some diagnostic value.

The results from our review support the view that referral for further investigations in people with suspected vertebral fracture should be based on a select number of red flags and using certain combinations of red flags may provide more precise diagnostic values. Given the overall quality of the current available evidence and that no meta‐analysis of studies was possible, the recommendations of this review should be interpreted with caution. Although results from single studies do demonstrate some red flags may be useful and have the potential to increase clinical suspicion of vertebral fracture in people presenting with low back pain, more robust research is needed.

Authors' conclusions

Implications for practice.

The available evidence suggests that only a few red flags are potentially useful in guiding clinical decisions regarding the need to further investigate people suspected to have a vertebral fracture. Most red flags are not useful in increasing the likelihood of vertebral fracture when present, nor are they useful in decreasing the likelihood of vertebral fracture when absent. Our review found a few red flags demonstrated some promise. In primary care, 'older age' was informative for 'unspecified vertebral fracture' and 'trauma' and 'corticosteroid use' were both informative for 'unspecified vertebral fracture' and 'osteoporotic vertebral fracture'. In secondary care, 'older age' was informative for 'osteoporotic vertebral fracture' and 'trauma' was informative for 'unspecified vertebral fracture'. In tertiary care, 'presence of contusion/abrasion' was informative for 'vertebral compression fracture'. Combinations of red flags were also informative and may be more useful in assisting clinicians with decisions than individual tests alone. Given the limitations of our review, it is still challenging to provide clear guidance as to which red flags should be used routinely in clinical practice. Further research with studies published in peer‐reviewed journals is needed to improve and consolidate our current recommendations for screening for vertebral fractures to guide clinical care.

Implications for research.

It is clear there is a paucity of primary diagnostic research in this field as we only found six additional studies since the previous version of the review (Williams 2013). Future research should consider which index tests may be clinically relevant based on the different types of vertebral fracture. There is very little consistency across studies in the index tests investigated and most studies do not provide clear details or thresholds for a positive or negative test. Most red flags investigated in the literature are individual tests; however, red flags (like many other tests) are rarely applied individually in clinical practice, yet these tests continue to be investigated in the research this way. Future studies should consider investigating combinations of index tests that seem most clinically appropriate or relevant for the different types of vertebral fracture. The results from our review highlight that a few promising index tests may be relevant in clinical practice, hence future studies should investigate these index tests to validate their use in a clinical care. Given the prevalence of vertebral fracture presenting as low back pain is low, primary diagnostic research requires large sample sizes to produce robust diagnostic accuracy estimates. These types of studies are challenging to conduct; however, they have the potential to reduce unnecessary imaging and its associated costs in a musculoskeletal condition (i.e. low back pain) that has been challenged with overimaging and its associated downstream consequences (Jacobs 2020).

History

Protocol first published: Issue 7, 2022

Appendices

Appendix 1. MEDLINE search strategy

I. Index test: clinical red flags

1. exp Medical History Taking/

2. exp Physical Examination/

3. exp "Wounds and Injuries"/

4. exp Accidental Falls/

5. exp Accidents/

6. physical examination*.tw,kf

7. history.tw,kf

8. red flag*.tw,kf

9. physical test*.tw,kf

10. trauma*.tw,kf

11. injur*.tw,kf

12. (clinical* adj (diagnosis or sign* or significance or symptom* or parameter* or assessment* or finding* or evaluat* or indication* or examination*)).tw,kf

13. OR/1–12

II. Population: low back pain

14. exp Back Pain/

15. exp Low Back Pain/

16. exp Sciatica/

17. exp Sciatic Neuropathy/

18. dorsalgia*.tw,kf

19. backache*.tw,kf

20. lumbago.tw,kf

21. back disorder*.tw,kf

22. (lumb* adj pain).tw,kf

23. coccyx.tw,kf

24. coccydynia.tw,kf

25. sciatica.tw,kf

26. spondylosis.tw,kf

27. OR/14–26

III. Target condition: vertebral fracture

28. Fractures, Bone/

29. Fractures, Stress/

30. Fractures, Spontaneous/

31. Fractures, Compression/

32. Fractures, Closed/

33. Spinal Injuries/

34. Spinal Diseases/

35. fracture*.tw,kf

36. OR/28–35

IV. Limits and search combination

37. (20120401‐20220726).ed

38. (2012‐2022).yr

39. OR/38–39

40. 13 AND 27 AND 36 AND 40

Appendix 2. Embase search strategy

1. exp Anamnesis/

2. exp Physical Examination/

3. exp Physical Performance/

4. exp Accident/

5. physical examination*.tw,kw

6. history.tw,kw

7. red flag*.tw,kw

8. physical test*.tw,kw

9. trauma*.tw,kw

10. injur*.tw,kw

11. (clinical* adj (diagnosis or sign* or significance or symptom* or parameter* or assessment* or finding* or evaluat* or indication* or examination*)).tw,kw

12. OR/1–11

II. Population: low back pain

13. exp Low back Pain/

14. exp Backache/

15. exp Spondylosis/

16. exp Sciatica/

17. exp Ischialgia/

18. dorsalgia.tw,kw

19. back pain.tw,kw

20. backache.tw,kw

21. lumbago.tw,kw

22. back disorder*.tw,kw

23. (lumb* adj pain).tw,kw

24. spondylosis.tw,kw

25. coccyx.tw,kw

26. coccydynia.tw,kw

27. sciatica.tw,kw

28. OR/13–27

III. Target condition: vertebral fracture

29. exp Fracture/

30. exp Spine Fracture/

31. exp Bone Injury/

32. exp Spine Injury/

33. exp Spine Disease/

34. fracture*.tw,kw

35. OR/29–34

IV. Limits and search combination

36. (20120401‐20220726).dd

37. (2012‐2022).yr

38. OR/36–37

39. 12 AND 28 AND 35 AND 38

40. limit 39 to embase

Appendix 3. CINAHL search strategy

I. Index test: clinical red flags

1. (MH "Patient History Taking")

2. (MH "Diagnostic Tests, Routine")

3. (MM "Physical Examination")

4. (TX "history" or "red flag*" or "physical examination" or "physical test" or "clinical*" or "diagnosis" or TX "sign" or TX "signs" or TX "significance" or TX "symptom*" or TX "parameter*" or TX "assessment" or TX "finding*" or TX "evaluat*" or TX "indication*" or TX "examination*")

5. OR/1–4

II. Population: low back pain

6. (MH "Back Pain+")

7. (MH "Low Back Pain")

8. (MH "Coccyx")

9. (MH "Sciatica")

10. (MH "Lumbar Vertebrae")

11. (MH "Spondylolisthesis")

12. (MH "Spondylolysis")

13. (TX "back pain")

14. (TX "backache")

15. (TX "dorsalgia")

16. (TX "lumbago")

17. (TX "sciatica")

18. (TX "coccyx")

19. (TX "coccydynia")

20. (TX "back disorder*")

21. (TX "lumbago")

22. lumb* w1 pain

23. lumb* n5 pain

24. lumb* n2 vertebra*

25. OR/11–29

III. Target condition: vertebral fracture

26. (MH "Fractures, Closed") or (MH "Fractures, Compression+") or (MH "Fractures, Spontaneous") or (MH "Fractures, Stress+")

27. (MH "Spinal Fractures+")

28. (MH "Spinal Injuries")

29. (MH "Spinal Diseases")

30. (MH "Wounds and Injuries")

31. (TX "fracture*")

32. (TX "trauma")

33. (TX "injury")

34. OR/26–33

IV. Limits and search combination

35. EM 201204‐

36. DT 2012‐

37. OR/35‐36

38. 5 AND 25 AND 34 AND 37

Appendix 4. QUADAS‐2 tool: risk of bias and applicability judgements

DOMAIN 1: PATIENT SELECTION
A. Risk of bias
Describe methods of patient selection
Was a consecutive or random sample of patients enrolled? Score 'yes' if a consecutive series of patients or a random sample has been selected. Details regarding setting, in inclusion and exclusion criteria, and preferably number of patients eligible and excluded should be given. If a mixed population (e.g. primary and secondary care) is used, the number of participants from each setting will be presented. Score 'no' if healthy controls are used. If non‐response is high and selective, or there is clear evidence of selective sampling, this will be scored 'no'. Score 'unclear' if insufficient information is given on the setting, selection criteria, or selection procedure to make a judgement.	Yes/No/Unclear
Did the study avoid inappropriate exclusions? Score 'yes' if participants were excluded based on the prespecified exclusion criteria or if the exclusion criteria was clearly appropriate for the study population. Score 'no' if inappropriate exclusion criteria are used such as excluding based on gender, or inappropriate selection bias occurs. Score 'unclear' if insufficient information is provided to assess this item.	Yes/No/Unclear
Could the selection of patients have introduced bias? Score 'low' if all questions are rated as 'yes'. Score 'high' if any of the questions is rated as 'no'. Score 'unclear if any of the questions are reported as 'unclear', then there is an unclear risk of bias and a judgement will be made on whether or not there is sufficient information to make a decision about the risk of bias.	Risk: Low/High/Unclear
B. Concerns regarding applicability
Describe included patients (prior testing, presentation, intended use of index test and setting) We expect that most of the diagnostic studies will include participations appropriate for the review question and would have used the index tests described in our inclusion criteria. Many of the index tests are applicable across the different healthcare settings.
Is there concern that the included patients do not match the review question? Score 'low concern' if participants meet all inclusion criteria or meet 3 out of the 4 criteria. Score 'high concern' if participants meet 1 or less of the 4 criteria. Score 'unclear' if insufficient information is provided to assess this item.	Concern: Low/High/Unclear
DOMAIN 2: INDEX TEST(S) (if more than 1 index test was used, please complete for each test)
A. Risk of bias
Describe the index test and how it was conducted and interpreted:
Were the index test results interpreted without knowledge of the results of the reference standard? Score 'yes' if the results of the index test are interpreted blind to the results of the reference test or if the sequence of testing is always the same (i.e. the index test is always performed first, followed by the reference standard), and consequently the index test is interpreted blind of the reference standard. Score 'no' if the assessor is aware of the results of the reference standard. Score 'unclear' if insufficient information is given on independent or blind assessment of the reference standard.	Yes/No/Unclear
If a threshold was used, was it prespecified? Score 'yes' if it is clear that a prespecified threshold was used e.g. the authors follow a study protocol. Score 'no' if a prespecified threshold was not used or different thresholds were used and not predefined. Score 'unclear' if insufficient information is provided to assess this item.	Yes/No/Unclear
Could the conduct or interpretation of the index test have introduced bias? Score 'low' if all questions are rated as 'yes'. Score 'high' if any of the questions is rated as 'no'. Score 'unclear' if any of the questions are reported as 'unclear', then there is an unclear risk of bias and a judgement will be made on whether or not there is enough information to make a decision about the risk of bias.	Risk: Low/High/Unclear
B. Concerns regarding applicability
Is there concern that the index test, its conduct, or interpretation differ from the review question? Score 'low concerns' if the index text performed is the same as specified in the review question and was intended to assess vertebral fracture. Score 'high concerns' if the index test performed does not follow what is specified in the review question and was not intended to assess vertebral fracture. Score 'unclear' if insufficient information is provided to assess this item.	Concern: Low/High/Unclear
DOMAIN 3: REFERENCE STANDARD
A. Risk of bias
Describe the reference standard and how it was conducted and interpreted:
Is the reference standard likely to correctly classify the target condition? Score 'yes' if the following criteria is followed. For traumatic vertebral fractures, CT is preferentially used as the reference test. If CT scan is not available, MRI is used as the reference test. For osteoporotic vertebral fractures and vertebral stress fractures, MRI is preferential used as the reference test. If MRI is not available, bone scintigraphy with SPECT is used as the reference test. Score 'no' if consensus among observers, or an unknown combination of the clinical assessment ('clinical judgement') is used as reference standard. Score as 'unclear' if insufficient information is given on the reference standard to make an adequate assessment.	Yes/No/Unclear
Were the reference standard results interpreted without knowledge of the results of the index test? Score 'yes' if the results of the reference standard are interpreted blind to the results of the index tests or if the sequence of testing is always the same (i.e. the reference standard is always performed first, followed by the index test) and consequently, the reference standard is interpreted blind of the index test. Score 'no' if the assessor is aware of the results of the index test. Score 'unclear' if insufficient information is given on independent or blind assessment of the index test.	Yes/No/Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias? Score 'low' if all questions are rated as 'yes'. Score 'high' if any of the questions is rated as 'no'. Score 'unclear' if any of the questions are reported as 'unclear', then there is an unclear risk of bias and a judgement will be made on whether or not there is sufficient information to make a decision about the risk of bias.	Risk: Low/High/Unclear
B. Concerns regarding applicability
Is there concern that the target condition as defined by the reference standard does not match the review question? Score 'low' concern if the target condition defined by the reference standard matches the target condition specified in the review. Score 'high' concern if the target condition defined by the reference standard does not match the target condition specified in the review. Score 'unclear' if insufficient information is provided to assess this item.	Concern: Low/High/Unclear
DOMAIN 4: FLOW AND TIMING
A. Risk of bias
Describe any patients who did not receive the index test(s) and/or reference standard or who were excluded from the 2×2 table (refer to flow diagram)
Describe the time interval and any interventions between index test(s) and reference standard:
Was there an appropriate interval between index test(s) and reference standard? Score 'yes' if the time period between clinical assessment and the reference standard ≤ 1 week. Score 'no' if the time period between clinical assessment and the reference standard is > 1 week. Score 'unclear' if there is insufficient information on the time period between index tests and reference standard.	Yes/No/Unclear
Did all patients receive the same reference standard? Score 'yes' if it is clear that all patients receiving the index test are subjected to the same reference standard. Score 'no' if different reference standards are used or if only some patients are subjected to the reference standard. Score 'unclear' if insufficient information is provided to assess this item.	Yes/No/Unclear
Were all patients included in the analysis? Score 'yes' if it is clear that all recruited patients (> 95%) were included in the analysis. Score 'no' if the number of patients recruited differs (> 5% difference) from the number of patients included for analysis. Score 'unclear' if insufficient information is provided to assess this item.	Yes/No/Unclear
Could the patient flow have introduced bias? Score 'low' if all questions are rated as 'yes'. Score 'high' if any of the questions is rated as 'no'. Score 'unclear' if any of the questions are reported as 'unclear', then there is an unclear risk of bias and a judgement will be made on whether or not there is sufficient information to make a decision about the risk of bias.	Risk: Low/High/Unclear

CT: computed tomography; MRI: magnetic resonance imaging; SPECT: single‐photon emission computed tomography.

Data

Presented below are all the data for all of the tests entered into the review.

Tests. Data tables by test.

Test	No. of studies	No. of participants
1 Age > 50 years	3	11423
2 Age > 54 years	2	3179
3 Age > 52 years	1	1449
4 Age > 64 years	2	3179
5 Age > 70 years	2	11112
6 Age > 74 years	2	3179
7 Female gender	5	5250
8 Female > 54 years	2	3179
9 Female > 64 years	2	3182
10 Female > 74 years	2	3280
11 Trauma	8	14105
12 Corticosteroid use	3	2152
13 OA	2	2117
14 No regular exercise	1	1448
15 Sensation change	4	3620
16 Motor deficit	2	1424
17 DTR abnormality	2	1423
18 Neurological signs	2	590
19 SLR	1	552
20 Tenderness	3	2427
21 Spasm	2	1423
22 Contusion/abrasion	1	552
23 Sciatica	2	1896
24 Hip/leg pain	1	871
25 Absence of buttock/leg pain	1	1448
26 BMI < 23	1	1448
27 Decreased pain on sitting	1	1448
28 No gait abnormality	1	1448
29 Trauma and neurological signs	1	108
30 Multiple findings	1	552
31 Henschke rule 1 sign positive	1	1172
32 Henschke rule 2 positive signs	1	1172
33 Henschke rule 3 positive signs	1	1172
34 1 of 5 positive ‐ Roman Cluster	1	1449
35 2 of 5 positive ‐ Roman Cluster	1	1449
36 3 of 5 positive ‐ Roman Cluster	1	1448
37 4 of 5 positive ‐ Roman Cluster	1	1448
38 5 of 5 positive ‐ Roman Cluster	1	1448
39 Age ≥ 75 years	2	805
40 Osteoporosis	1	669
41 Percussion tenderness of spine	1	669
42 Sudden decrease in height	1	669
43 Severe disability	1	669
44 Acute onset of pain	1	669
45 Back pain score ≥ 7	1	669
46 Thoracic back pain	1	669
47 BPIT	1	510
48 Age 65–75 years	1	136
49 Participation in sports	1	100
50 Hyperextension	2	1111
51 Hyperflexion	2	1109
52 Sharp pain	1	82
53 Pain extent (fingertip size)	1	96
54 Pain location	1	99
55 Previous history of LBP	1	1025
56 Waking night pain	1	1025
57 Lumbar rotation	1	1010
58 Lumbar lateral flexion	1	1012
59 Single leg hyperextension	1	989
60 2 of 3 negative ‐ Therriault cluster	0	0
61 2 of 3 positive ‐ Therriault cluster	0	0
62 3 of 3 negative ‐ Therriault cluster	0	0

1. Test.

Age > 50 years

2. Test.

Age > 54 years

3. Test.

Age > 52 years

4. Test.

Age > 64 years

5. Test.

Age > 70 years

6. Test.

Age > 74 years

7. Test.

Female gender

8. Test.

Female > 54 years

9. Test.

Female > 64 years

10. Test.

Female > 74 years

11. Test.

Trauma

12. Test.

Corticosteroid use

13. Test.

OA

14. Test.

No regular exercise

15. Test.

Sensation change

16. Test.

Motor deficit

17. Test.

DTR abnormality

18. Test.

Neurological signs

19. Test.

SLR

20. Test.

Tenderness

21. Test.

Spasm

22. Test.

Contusion/abrasion

23. Test.

Sciatica

24. Test.

Hip/leg pain

25. Test.

Absence of buttock/leg pain

26. Test.

BMI < 23

27. Test.

Decreased pain on sitting

28. Test.

No gait abnormality

29. Test.

Trauma and neurological signs

30. Test.

Multiple findings

31. Test.

Henschke rule 1 sign positive

32. Test.

Henschke rule 2 positive signs

33. Test.

Henschke rule 3 positive signs

34. Test.

1 of 5 positive ‐ Roman Cluster

35. Test.

2 of 5 positive ‐ Roman Cluster

36. Test.

3 of 5 positive ‐ Roman Cluster

37. Test.

4 of 5 positive ‐ Roman Cluster

38. Test.

5 of 5 positive ‐ Roman Cluster

39. Test.

Age ≥ 75 years

40. Test.

Osteoporosis

41. Test.

Percussion tenderness of spine

42. Test.

Sudden decrease in height

43. Test.

Severe disability

44. Test.

Acute onset of pain

45. Test.

Back pain score ≥ 7

46. Test.

Thoracic back pain

47. Test.

BPIT

48. Test.

Age 65–75 years

49. Test.

Participation in sports

50. Test.

Hyperextension

51. Test.

Hyperflexion

52. Test.

Sharp pain

53. Test.

Pain extent (fingertip size)

54. Test.

Pain location

55. Test.

Previous history of LBP

56. Test.

Waking night pain

57. Test.

Lumbar rotation

58. Test.

Lumbar lateral flexion

59. Test.

Single leg hyperextension

60. Test.

2 of 3 negative ‐ Therriault cluster

61. Test.

2 of 3 positive ‐ Therriault cluster

62. Test.

3 of 3 negative ‐ Therriault cluster

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Deyo 1986.

Study characteristics
Patient Sampling	Study design: prospective cohort Enrolment procedures: consecutive sample Sample size: 621; 311 received imaging Inclusion criteria: people presenting with LBP Exclusion criteria: maximal pain above T12, evidence of urinary tract disease, women aged < 45 years who did not use contraception and whose last menstrual period occurred > 10 days previous, and participants of a concurrent study
Patient characteristics and setting	Age group: adolescents and adults (mean 40.5 years; range 15–86) years Gender: 293 male, 328 female Duration of pain: < 1 month: 447, ≥ 1 month: 146, unknown: 28 History of LBP: 289/621 Healthcare setting: walk‐in university clinic Country: USA Type of vertebral fracture: vertebral fracture; type unclear from text Reasons for withdrawal: unclear from text
Index tests	Index tests Significant trauma Age > 50 years Corticosteroid use Methods of execution: checklist Threshold for positive result on index test: presence or absence of red flags from checklist Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture or malignancy on: X‐ray – anteroposterior and lateral lumbar views
Flow and timing	Timing of index test and reference test: 84% of imaging was obtained on the day of the index visit or within 6 days thereafter.
Comparative
Notes	Prevalence of vertebral fracture: 4.5%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	No
Could the selection of patients have introduced bias?		High risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Low risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	No
Could the reference standard, its conduct, or its interpretation have introduced bias?		High risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	No
Did all patients receive the same reference standard?	No
Were all patients included in the analysis?	No
Could the patient flow have introduced bias?		High risk

Enthoven 2016.

Study characteristics
Patient Sampling	Study design: prospective observational cohort Enrolment procedures: consecutive sample Sample size: 669 Inclusion criteria: people consulting a GP with a new episode of back pain Exclusion criteria: unable to fill out the questionnaires due to cognitive impairment, were unable to read and write in Dutch, unable to undergo physical examination (e.g. people using wheelchairs)
Patient characteristics and setting	Age group: adults aged > 55 years Gender: 269 men, 400 women Duration of pain: > 3 months: 154/669 History of LBP: 87/669 Healthcare setting: primary care: general practice Country: Netherlands Type of vertebral fracture: osteoporotic vertebral fracture (30/33); unclear from text type of fracture for 3/30 participants Reasons for withdrawal: 6 (0.9%) participants were excluded from the analyses because they moved to another city during the 1‐year follow‐up and, due to a change of GP practice, or diagnosis could not be retrieved.
Index tests	Index tests Aged ≥ 75 years Female sex Prolonged corticosteroid use Trauma Osteoporosis Sudden decrease in height Percussion tenderness of spine Severe disability Acute onset of pain Back pain intensity score ≥ 7 Osteoarthritis in hip/knee Thoracic back pain Methods of execution: questionnaire and a structured physical examination of the back Threshold for positive result on index test: presence or absence of red flags from questionnaire and physical examination Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: lumbar or thoracic X‐ray
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of osteoporotic vertebral fracture: 4.5%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	No
Could the selection of patients have introduced bias?		High risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Low risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	No
Could the reference standard, its conduct, or its interpretation have introduced bias?		High risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Unclear risk

Gibson 1992.

Study characteristics
Patient Sampling	Study design: prospective cohort Enrolment procedures: consecutive sample Sample size: 225, 108 received imaging Inclusion criteria: people presenting with pain in the lumbar region of < 48 hours' duration Exclusion criteria: not reported
Patient characteristics and setting	Age group: adolescents and adults (range 16–65 years) Gender: not reported Duration of pain: < 48 hours History of LBP: not reported Healthcare setting: accident and emergency department Country: UK Type of vertebral fracture: vertebral fracture; type unclear from text Reasons for withdrawal: not reported
Index tests	Index tests Trauma Trauma and neurological signs Neurological signs or straight leg raise < 40° (or both) Methods of execution: questionnaire Threshold for positive result on index test: presence or absence of red flags from questionnaire Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: X‐ray; no further details
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of vertebral fracture: 6.5%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	Unclear
Could the selection of patients have introduced bias?		Unclear risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Low risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	No
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Unclear risk

Henschke 2009.

Study characteristics
Patient Sampling	Study design: prospective cohort Enrolment procedures: consecutive sample Sample size: 1172 Inclusion criteria: people presenting with acute LBP, with < 1 week since onset Exclusion criteria: serious pathology diagnosed prior to consultation, and the serious pathology was considered to be the cause of the current episode of LBP
Patient characteristics and setting	Age group: adults (mean age 43.97 years) Gender: 626 men, 546 women Duration of pain: < 1 week: 696, 1–2 weeks: 145, 2–3 weeks: 174, 3–4 weeks: 73, 4–5 weeks: 30, 5–6 weeks 54 History of LBP: 888/1172 Healthcare setting: primary care Country: Australia Type of vertebral fracture: vertebral fracture; type unclear from text Reasons for withdrawal: all participants completed follow‐up
Index tests	Index tests: 25 red flags total – 5 specific to vertebral fracture Age > 70 years Significant trauma Prolonged use of corticosteroids Altered sensation Clinician diagnosis of fracture Methods of execution: checklist Threshold for positive result on index test: presence or absence of red flags from checklist Experience and expertise of assessors: unclear from text
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: clinical follow‐up at 12 months with suspected cases confirmed by imaging studies and specialist review
Flow and timing	Timing of index test and reference test: time between index test and reference standard was 12 months
Comparative
Notes	Prevalence of vertebral fracture: 0.68%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	Yes
Could the selection of patients have introduced bias?		Low risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Low risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	No
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			High
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	No
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		High risk

Jin 2020.

Study characteristics
Patient Sampling	Study design: prospective observational cohort Enrolment procedures: consecutive sample Sample size: 510 Inclusion criteria: presented to the spine clinic with chief complaints of LBP who were postmenopausal women or women over 50 years old and men over 60 years old or people with a history of prolonged use of a glucocorticosteroid for ≥ 3 months at any age Exclusion criteria: back pain with pain radiating to the limbs or night/resting pain; definite recent high‐energy trauma; an established diagnosis of infection, tumour, or fractures in other medical facilities before presenting to clinic; back pain with dysfunction of the central or peripheral nervous system; inability to complete the index test due to a cognitive deficit, mental impairment, or communication dysfunction (e.g. inability to hear or speak); or failure to complete an MRI scan due to any conditions, such as claustrophobia or artificial pacemaker implantation
Patient characteristics and setting	Age group: women > 50 years or men > 60 years (mean age 70.6 years) Gender: 113 men, 397 women Duration of pain: < 2 weeks: 219, 2–8 weeks: 138, ≥ 8 weeks: 153 History of LBP: not reported Healthcare setting: hospital spine clinic Country: China Type of vertebral fracture: osteoporotic vertebral fracture Reasons for withdrawal 8 participants refused to attempt the first movement because of cardiac insufficiency (2) and severe back pain (6) MRI not performed in 56 participants for individual reasons 8 participants with inconclusive or inconsistent MRI results were excluded from the analysis because of a suspicious malignancy/infection based on MRI
Index tests	Index tests Physical examination test (Back Pain‐Inducing Test) Motion of lying to supine, rolling over, sitting up Methods of execution: physical examination Threshold for positive result on index test Defined as positive: participants reporting back pain occurring during any movement or participants not being able to perform any of the 3 movements due to back pain Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: MRI scan with short‐tau inversion recovery (STIR) sequencing of the whole spine using a 1.5‐Tesla system. Diagnoses were made by combining T1‐weighted images (T1WIs), T2‐weighted images (T2WIs), and STIR images in the sagittal plane.
Flow and timing	Timing of index test and reference test: index test and reference standard both completed within 3 days.
Comparative
Notes	Prevalence of vertebral fracture: 68%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	No
Could the selection of patients have introduced bias?		High risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Low risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Yes
Could the reference standard, its conduct, or its interpretation have introduced bias?		Low risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Yes
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	No
Could the patient flow have introduced bias?		High risk

Kilic 2021.

Study characteristics
Patient Sampling	Study design: retrospective cohort Enrolment procedures: unclear from text Sample size: 136 Inclusion criteria: records of the people admitted to physical therapy and rehabilitation polyclinic with complaints of back pain Exclusion criteria: those whose data could not be accessed; and history of rheumatological disease, malignancy, and trauma
Patient characteristics and setting	Age group: adults (mean age 73.9 years) Gender: 17 men, 119 women Duration of pain: not reported History of LBP: not reported Healthcare setting: hospital (physiotherapy and rehabilitation polyclinic) Country: Turkey Type of vertebral fracture: spontaneous osteoporosis vertebral fracture Reasons for withdrawal: all participants included in analysis
Index tests	Index tests Aged ≥ 75 years Aged 65–75 years Methods of execution: hospital records system Threshold for positive result on index test: aged ≥ 75 years and aged 65–75 years Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: radiograph; presence of fractures in participants; T4–L5 in lateral spinal radiographs anterior, middle, or posterior heights in intervertebrae by semi‐quantitative method or lateral height in ≥ 1 vertebra between T4 and L5
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of vertebral fracture: 54.4%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Unclear
Did the study avoid inappropriate exclusions?	No
Could the selection of patients have introduced bias?		High risk
Are there concerns that the included patients and setting do not match the review question?			Unclear
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Unclear
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Unclear risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Unclear risk

Patrick 1983.

Study characteristics
Patient Sampling	Study design: retrospective chart review Enrolment procedures: consecutive sample, reason for patient presentation not clear Sample size: 552 Inclusion criteria: lumbar spine series ordered in the emergency departments were enroled Exclusion criteria: unclear from text
Patient characteristics and setting	Age group: children to adults (range 6–95 years) Gender: 296 male, 256 female Duration of pain: unclear from text History of LBP: unclear from text Healthcare setting: accident and emergency department and university hospital medical centre Country: USA Type of vertebral fracture: vertebral compression fracture Reasons for withdrawal: all participants included in analysis
Index tests	Index tests Trauma Tenderness LBP with radiation or hip pain, or both Spasm Sensory deficit Motor deficit Tendon reflex abnormality Positive straight leg raise Contusion/abrasion Multiple findings Methods of execution: hospital records system Threshold for positive result on index test: presence or absence of red flags from checklist Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: X‐ray; not further defined
Flow and timing	Timing of index test and reference standard: index test and reference standard performed at time of admission
Comparative
Notes	Prevalence of vertebral fracture: 7.2%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Unclear
Did the study avoid inappropriate exclusions?	Unclear
Could the selection of patients have introduced bias?		Unclear risk
Are there concerns that the included patients and setting do not match the review question?			Unclear
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Unclear
If a threshold was used, was it pre‐specified?	Unclear
Could the conduct or interpretation of the index test have introduced bias?		Unclear risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Yes
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Low risk

Premkumar 2018.

Study characteristics
Patient Sampling	Study design: retrospective cohort Enrolment procedures: consecutive sample Sample size: 9940 Inclusion criteria: people who presented as a new patient with a chief complaint of LBP with or without leg pain Exclusion criteria: chief complaint of deformity rather than pain, chief complaint of neck pain
Patient characteristics and setting	Age group: adults (mean age 56.8 years) Gender: 5102 men, 4838 women Duration of pain: not reported History of LBP: not reported Healthcare setting: academic multidisciplinary spine centre Country: USA Type of vertebral fracture: vertebral fracture; type unclear from text Reasons for withdrawal: all participants included in analysis
Index tests	Index tests Age > 50 years Age > 70 years Trauma Combination 1: trauma and age > 50 years Combination 2: trauma and age > 70 years Methods of execution: questionnaire Threshold for positive result on index test: presence or absence of red flags from questionnaire Experience and expertise of assessors: fellowship trained spine surgeons
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: imaging; type of imaging not specified. Diagnostic information was drawn directly from the physician entry and was corroborated by imaging reports.
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of vertebral fracture: 5.6%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	Yes
Could the selection of patients have introduced bias?		Low risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Unclear
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Unclear risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	Unclear
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Unclear risk

Reinus 1998.

Study characteristics
Patient Sampling	Study design: prospective cohort Enrolment procedures: unclear from text Sample size: 482 Inclusion criteria: all patients receiving lumbosacral radiographs in a 14‐month period Exclusion criteria: unclear from text
Patient characteristics and setting	Age group: adolescents and adults (range 17–98 years) Gender: 168 male, 314 female Duration of pain: unclear from text History of LBP: unclear from text Healthcare setting: accident and emergency department Country: USA Type of vertebral fracture: vertebral compression fracture Reasons for withdrawal: unclear from text
Index tests	Index test(s) Trauma Neurological deficit Trauma or neoplasm (indeterminate age fracture) Methods of execution: questionnaire Threshold for positive result on index test: presence or absence of red flags from questionnaire Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: X‐ray – lumbosacral anteroposterior, lateral, bilateral posterior oblique, and coned down radiological views lumbar views
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of fracture: 11%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Unclear
Did the study avoid inappropriate exclusions?	Unclear
Could the selection of patients have introduced bias?		Unclear risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	No
Could the conduct or interpretation of the index test have introduced bias?		High risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Unclear risk

Roman 2010.

Study characteristics
Patient Sampling	Study design: retrospective chart review Enrolment procedures: consecutive sample Sample size: 1448 Inclusion criteria: all patients with a lumbar‐related disorder Exclusion criteria: unclear from text
Patient characteristics and setting	Age group: adults Gender: 587 men, 861 women Duration of pain: not reported History of LBP: not reported Healthcare setting: university; Department of Surgery Country: USA Type of vertebral fracture: osteoporotic vertebral compression fracture Reasons for withdrawal: all participants included in analysis
Index tests	Index tests Gender Age BMI Gait abnormality Regular exercise Sitting pain Osteoarthritis Leg or buttock pain Multiple signs Methods of execution: clinical examination Threshold for positive result on index test: presence or absence of red flags from during physical examination Experience and expertise of assessors: 2 assessors (4 and 22 years' experience)
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: standard radiograph or CT; assessing sagittal alignment, vertebral body compression, and spinal canal dimensions
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of vertebral fracture: 3%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	Unclear
Could the selection of patients have introduced bias?		Unclear risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Unclear
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Unclear risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	No
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		High risk

Scavone 1981.

Study characteristics
Patient Sampling	Study design: retrospective chart review Enrolment procedures: unclear from text Sample size: 871 Inclusion criteria: people with lumbar spine X‐ray Exclusion criteria: unclear from text
Patient characteristics and setting	Age group: unclear from text Gender: unclear from text Duration of pain: unclear from text History of LBP: unclear from text Healthcare setting: university teaching hospital medical centre Country: USA Type of vertebral fracture: vertebral fracture; unclear from text Reasons for withdrawal: unclear from text; history for 57 patients could not be obtained
Index tests	Index tests Major trauma Minor trauma Tenderness LBP with radiation Hip/leg pain Muscle spasm Neurological deficits Sciatica Abnormal physical examination Methods of execution: chart Threshold for positive result on index test: presence or absence of red flags from chart Experience and expertise of assessors: unclear from text
Target condition and reference standard(s)	Reference test(s): vertebral fracture on AP and lateral X‐ray views
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of vertebral fracture: 3%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Unclear
Did the study avoid inappropriate exclusions?	Unclear
Could the selection of patients have introduced bias?		Unclear risk
Are there concerns that the included patients and setting do not match the review question?			Unclear
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	Unclear
Could the conduct or interpretation of the index test have introduced bias?		Unclear risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	Unclear
Could the patient flow have introduced bias?		Unclear risk

Sugiura 2021.

Study characteristics
Patient Sampling	Study design: retrospective comparative cohort study Enrolment procedures: consecutive sample Sample size: 101 Inclusion criteria: adolescent presenting for rehabilitation within 1 month of acute LBP Exclusion criteria: participants with lower extremity symptoms (to exclude the possibility of radicular back pain), clear spondylolysis or spondylolisthesis based on plain radiography findings, and other spinal disorders based on MRI findings
Patient characteristics and setting	Age group: adolescents aged < 18 years (mean age 14.4 years; range 10–18 years) Gender: 74 male, 27 female Duration of pain: < 1 month History of LBP: not reported Healthcare setting: unclear from text Country: Japan Type of vertebral fracture: early‐stage spondylolysis; vertebral stress fracture Reasons for withdrawal: unclear from text; participants missing from analysis
Index tests	Index tests Gender Participation in sports activities Presence or absence of LBP during lumbar spine extension in a standing position (the hyperextension test) Presence or absence of LBP during lumbar spine flexion in a standing position (the hyperflexion test) Pain quality (sharp or dull) Pain extent (categorised as a fingertip‐sized or palm‐sized area) Pain location (pain on the left or right side (or both) or central pain centre) Methods of execution: physical examination Threshold for positive result on index test: presence or absence of red flags during physical examination Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference tests: vertebral fracture on: MRI; the diagnosis was made when the lumbar spine pedicle showed high signal intensity on fat saturation T2‐weighted images and low signal intensity on T1‐weighted images
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of vertebral fracture: 52.5%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	No
Could the selection of patients have introduced bias?		High risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Unclear
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Unclear risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Unclear risk

Therriault 2020.

Study characteristics
Patient Sampling	Study design: retrospective case series Enrolment procedures: consecutive sample Sample size: 1025 Inclusion criteria: people with a primary complaint of LBP Exclusion criteria: person did not participate in an organised sport, defined as physical activity with formal practice or competition ≥ 2 times per week, or diagnosis of spondylolysis was not substantiated by imaging. Any patient without a report of LBP was removed from the potential patient list.
Patient characteristics and setting	Age group: adolescents (mean age 15 years) Gender: 455 male, 570 female Duration of pain: mean duration 19.5 weeks History of LBP: 253/1025 participants Healthcare setting: sports medicine department at a large children's hospital Country: USA Type of vertebral fracture: active spondylolysis (vertebral stress fracture) Reasons for withdrawal: all participants included in analysis
Index tests	Index tests Physical examination and history Sex Previous history of LBP Pain at rest Pain reported with lumbar extension Pain reported with lumbar flexion Pain reported with lumbar rotation Pain reported with lumbar lateral flexion Radicular symptoms reported Numbness or tingling reported Single‐leg Hyperextension Test Vertebral tenderness Methods of execution: physical examination and history Threshold for positive result on index test: presence or absence of red flags during physical examination and history Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: anteroposterior and standing lateral radiographs; if radiographs did not identify a spondylolytic injury and the physician suspected this diagnosis further imaging was performed; MRI; or SPECT.
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of vertebral fracture: 22%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	Yes
Could the selection of patients have introduced bias?		Low risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	No
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		High risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	No
Could the reference standard, its conduct, or its interpretation have introduced bias?		High risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	No
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		High risk

van den Bosch 2004.

Study characteristics
Patient Sampling	Study design: retrospective chart review Enrolment procedures: random sample Sample size: 2007 Inclusion criteria: people with full radiographic and demographic details Exclusion criteria: unclear from text
Patient characteristics and setting	Age group: adolescents and adults (range 0 to > 75 years) (age range as reported by study authors) Gender: 845 male, 1162 female Duration of pain: not reported History of LBP: not reported Healthcare setting: hospital setting Country: UK Type of vertebral fracture: osteoporotic and traumatic fracture; data not differentiated for different type of fracture Reasons for withdrawal: not reported
Index tests	Index tests Age Gender Methods of execution: hospital records system Threshold for positive result on index test: presence or absence of red flags from hospital record system Experience and expertise of assessors: not reported
Target condition and reference standard(s)	Reference test(s): vertebral fracture on: X‐ray; not further defined
Flow and timing	Timing of index test and reference test: unclear from text
Comparative
Notes	Prevalence of vertebral fracture: 4.1%
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient Selection
Was a consecutive or random sample of patients enrolled?	Yes
Did the study avoid inappropriate exclusions?	Unclear
Could the selection of patients have introduced bias?		Unclear risk
Are there concerns that the included patients and setting do not match the review question?			Low concern
DOMAIN 2: Index Test (Index Test)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	Yes
Could the conduct or interpretation of the index test have introduced bias?		Low risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference Standard
Is the reference standards likely to correctly classify the target condition?	Yes
Were the reference standard results interpreted without knowledge of the results of the index tests?	Unclear
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and Timing
Was there an appropriate interval between index test and reference standard?	Unclear
Did all patients receive the same reference standard?	Yes
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Unclear risk

AP: anterior‐posterior; BMI: body mass index; CT: computed tomography; GP: general practitioner; LBP: low back pain; MRI: magnetic resonance imaging; SPECT: single‐photon emission computerised tomography.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Althoff 2017	Inappropriate population sample investigated
Baleanu 2020	Inappropriate population sample investigated
Bian 2022	People already diagnosed with vertebral fracture
Borgen 2020	Inappropriate population sample investigated
Bottai 2016	People already diagnosed with vertebral fracture
Cheung 2018	Inappropriate study design
Clark 2017	Inappropriate study design
Della‐Giustina 2013	Inappropriate study design
Della‐Giustina 2015	Inappropriate study design
Durham 1995	Inappropriate population sample investigated
Enoki 2022	Inappropriate population sample investigated
Finucane 2020	Inappropriate study design
Frankel 1994	Inappropriate population sample investigated
Gaddikeri 2018	No index test assessed
Galliker 2019	Wrong study design
Gazzola 2015	Inappropriate population sample investigated
Gestring 2002	Inappropriate population sample investigated
Gill 2013	Inappropriate population sample investigated
Hiett 2014	No appropriate data available
Holmes 2003	Inappropriate population sample investigated
Hsu 2003	Inappropriate population sample investigated
Johannesdottir 2021	Inappropriate study design
Johansson 2018	Inappropriate population sample investigated
Kendler 2016	Inappropriate study design
Khera 2022	No appropriate data available
Kherad 2015	Inappropriate population sample investigated
Korner 2014	Inappropriate study design
Kuru 2014	Inappropriate population sample investigated
Manji 2022	People already diagnosed with vertebral fracture
Mazziotti 2016	Inappropriate population sample investigated
Min 2017	Inappropriate study design
Mizukami 2014	No index test assessed
Moretti 2019	Inappropriate population sample investigated
Padlina 2017	Inappropriate population sample investigated
Pressney 2014	Inappropriate study design
Rajakulasingam 2022	Inappropriate population sample investigated
Rapala 2012	No appropriate data available
Reito 2018	No appropriate data available
Rudman 2019	Inappropriate population sample investigated
Sakai 2016	Inappropriate population sample investigated
Samuels 1993	Inappropriate population sample investigated
Scaturro 2020	No appropriate data available
Shah 2021	Inappropriate population sample investigated
Soto‐Subiabre 2020	People already diagnosed with vertebral fracture
St Jeor 2020	Inappropriate population sample investigated
Sucato 2012	Inappropriate study design
Terakado 2017	No index test assessed
Terregino 1995	Inappropriate population sample investigated
Trout 2015	Inappropriate study design
Upadhye 2016	Inappropriate study design
van der Jagt‐Willems 2012	Inappropriate population sample investigated
Yoshimura 2017	Inappropriate population sample investigated
Zong 2016	Inappropriate population sample investigated
Zorin 2020	Inappropriate population sample investigated

LBP: low back pain.

Differences between protocol and review

In our protocol, we planned to investigate study level covariates in the HSROC analysis to explore heterogeneity but were unable due to a lack of data.

We planned to present results based on the different types of vertebral fracture; however, due to a lack of data this was not feasible. We instead presented our results based on the different healthcare settings and presented the results separately for different vertebral fracture types where available, within each healthcare setting.

We also planned to perform a sensitivity analysis to assess whether diagnostic accuracy differs based on study quality (risk of bias) or the different types of vertebral fracture on study quality but were unable to due to a lack of data. We were unable to perform a sensitivity analysis due to a lack of data.

We planned to use SAS software for meta‐analysis; however, meta‐analysis was not possible and instead used Review Manager 5 for all statistical analyses and data syntheses (Review Manager 2014).

We intended to present results of how different pretest probability cut‐offs and positive or negative results impacted post‐test probabilities; however, we were unable to due to a lack of data.

Contributions of authors

CSH wrote the subsequent and final (published) versions of the protocol and participated in the conception of the review; design of the review; co‐ordination of the review; search and selection of studies for inclusion in the review; collection of data for the review; assessment of the risk of bias in the included studies; analysis of data; interpretation of data; and writing of the review.

MJH provided critical intellectual input and participated in the conception of the review; design of the review; co‐ordination of the review; analysis of data; interpretation of data; and writing of the review.

AD provided critical intellectual input and participated in the conception of the review; design of the review; and writing of the review.

JGJ provided critical intellectual input and participated in the conception of the review; design of the review; and writing of the review.

BWK provided critical intellectual input and participated in the conception of the review; design of the review; and writing of the review.

GCM wrote the first draft of the protocol with help from the other authors, provided critical intellectual input, and participated in the conception of the review; design of the review; and writing of the review.

APV provided critical intellectual input and participated in the conception of the review; design of the review; and writing of the review.

CMW provided critical intellectual input and participated in the conception of the review; design of the review; and writing of the review.

QC participated in the conception of the review; design of the review; search and selection of studies for inclusion in the review; collection of data for the review; assessment of the risk of bias in the included studies; and writing of the review.

CGM provided critical intellectual input and participated in the conception of the review; design of the review; co‐ordination of the review; analysis of data; interpretation of data, and; writing of the review.

Sources of support

Internal sources

• Vrije Universiteit, EMGO+ Institute for Health and Care Research, Netherlands

• University of Sydney, Australia

• Erasmus MC, University Medical Centre, Netherlands

• The George Institute for Global Health, Australia

• National Institute of Arthritis Musculoskeletal and Skin Diseases (NIAMS)/National Institutes of Health (NIH) P30AR072572, USA

  The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH

External sources

• Dutch Health Care Insurance Board, Netherlands

• National Health and Medical Research Council, Australia

Declarations of interest

CSH: none.

MJH: none.

AD: none.

JGJ declares the following: Association of University Radiologists (Fiduciary Officer), Wolters Klewer Health, Inc. (Independent Contractor – Consultant), Springer Publishing (Independent Contractor – Other), GE Healthcare (travel)

BWK: none.

GCM: none.

APV: none.

CMW: none.

QC: none.

CGM is a Senior Editor for Cochrane Musculoskeletal. Editors are required to conduct at least one Cochrane Review, which ensures that they are aware of the processes and commitment needed to conduct reviews.

New