Red flags for the early detection of spinal infection in back pain patients

Studies investigating red flags in SI were included. More specifically, the eligible studies were study designs and research articles that had primary data, this included diagnostic accuracy studies, cohort studies, case-control studies, and case-series studies. Two reviewers (MY & LF), independently, screened the eligible papers against the full article eligibility criteria, any disagreements between the two reviewers were resolved by a third reviewer (JS). Included studies were studies in the English language that studied adults over the age of 18 who had a spinal infection, this included Bacterial, Viral, Fungal, Prionic or Parasitic infections. Further, studies were included if they contained primary data such as clinical tests, diagnostic tests, history taking and/or physical examination and provided red flags or clinical features for Spinal Infection. Studies were excluded if they were a systematic review or a narrative review.

Once the full eligible papers were selected, the quality of the studies was assessed using the National Heart, Lung and blood institute (NHLBI) Quality Assessment Tools. These widely used assessment tools determine the quality of case-series, case-control and cohort studies.

One reviewer (MY) extracted the data from all the eligible papers. The following data were extracted from each study: study characteristics, participant characteristics, setting and diagnostic methods, red flags (exposures) and spinal infection (outcomes). After the data was extracted, 25% of the data was independently checked by reviewer two (LF), as recommended by the 2009 Updated Method Guidelines for Systematic Reviews in the Cochrane Back Review Group [25].

Patient data from the included studies were analysed descriptively using frequencies and percentages.

There were no patient or public involvement in any phase of this study, this included the development of the research question, the analysis and the conclusions.

After reviewing the 52 eligible papers, 40 papers met the full eligibility criteria as agreed by the two reviewers (see additional file 1). One paper was excluded from the review because it was a narrative review [26], two papers did not include a description of the testing performed [27, 28] and nine papers were excluded because they did not provide data for red flags/clinical features for SI [11, 12, 28‐35].

The kappa statistics for the agreement between the two reviewers was k = 0.80, this is considered as ‘Very good’ [36, 37]. Reviewer three (JS) resolved the disagreement between the first two reviewers, the three discordant papers were included in the study [38‐40]. All disagreements were due to criterion three, whether the papers provided clinical features for SI.

For a summary of the main characteristics in each paper, see additional file 1. Table 1 presents a summary of the baseline characteristics of the patients in the 40 papers reviewed. A total of 2224 patients were diagnosed, including 1385 male patients (62%) and 773 female patients (38%) with a mean age of 55 years (SD = 8). In total, 14 out of 40 papers did not describe or adequately describe the radiological tests performed. This includes plain film radiography, ultrasound, magnetic resonance imaging scan (MRI), gallium scan, bone scan, computed tomography (CT) scan and myelogram [7, 13, 15, 39, 41‐51]. MRI was the most used radiological test with 22 papers using it for investigation. For individual investigations, Plain radiographs and MRI were the most used radiological tests. Bone scans and Gallium scans were the most accurate tests with a positive result of 93 and 90%, respectively. However, ultrasound was the least sensitive test with a positive result of 21%. Out of all studies, 34 (85%) were carried out in High-Income countries and 6 (15%) were carried out in Upper middle-income countries, none were carried out in Lower Middle-Income and Lower Income countries.

The list of red flags is presented in two groups, determinants and clinical features (Table 2). This decision was made because of the way in which the papers reported this information. Determinants were defined as factors that may play a part in the occurrence of SI, this included comorbidities, social factors and sources of infection that may lead to SI. Within the Dictionary of Epdimeiology, Last et al. define determinant as:

“a collective or individual risk factor (or set of factors) that is causally related to a health condition, outcome, or other defined characteristic. In human health —and, specifically, in diseases of complex etiology—determinants often act jointly in relatively complex and long-term processes. They commonly operate at aggregate (e.g., social, regional, global) levels, as well as at the individual, personal level..” [52].

Adapted from the Myriam-webster dictionary, clinical features were characterised as observable and diagnosable symptoms [53] that may occur as a consequence of SI. Here we can consider determinants as a priori red flags leading to the development of SI, and the a posteriori as the the signs symptoms and clinical features that are present after the onset of SI.

In total there were 46 items, 23 determinants and 23 clinical features. For convenience and ease of use from a clinical perspective, the data were aggregated into relevant groups. The most frequently reported determinants were Diabetes (18%), Intravenous drug use (9%) and Surgery (6%) and the most frequently reported clinical features were Spinal Pain (72%), fever (55%) and Neurological Dysfunction (33%). Other salient red flags were immunosuppression (3%), invasive procedures (2%), corticosteriod use (2%), and history of TB (1%).

The pathogens reported were Bacterial (n = 1665), Fungal (n = 12), mixed organism (n = 27) and unknown growth (n = 82), with the most common microorganisms being Staphylococcus aureus (27%; n = 598), Mycobacterium tuberculosis (12%; n = 262) and Brucella (9%; n = 210) (Table 3). There were no reports found on the other potential pathogenic causes of SI i.e. Viral, Prionic or Parasitic

Red flags currently used in clinical practice that are considered specific to infection are; the use of corticosteroids, or immunosuppressant therapy, Intravenous drug abuse, past history of TB and fever [58, 59]. These, with the exception of fever, are all related to determinants rather than clinical signs of infection. The results of this paper concur and all of these determinants are reported, however, the most reported determinant was diabetes which featured in 30 papers. Although diabetes appeared as a determinant in a large number of papers it is interesting that it was only reported in 18% of patients. The review highlights that a number of morbidities such as diabetes, cancer, and HIV are among those conditions associated with immunosuppression which can consequently result in a suscebtibility to infection. Determinants such as corticosteroid use and alcohol abuse can also lead to a risk of immunodeficiency [60‐63]. Also of note is that diseases such as rheumatoid arthritis and cancer sufferers treated with medications known to cause immunodeficiency (e.g. Disease-modifying anti-rheumatic drugs (DMARDS), steroids) are also at risk of SI. Determinants for SI (including intravenous drug use, diabetes and cancer) have high sensitivity (98%) and negative predictive value (99%) making them a better predictor of SI than clinical features such as the ‘classic triad’ (spinal pain, fever and neurological dysfunction) [54].

Spinal surgery and invasive procedures are regarded as having a high risk of infection with a rate of 1–4% reported elsewhere [64]. However, in this review, surgery and invasive procedures were reported as 6 and 2%, respectively. With procedural-induced SI, there are modifiable and non-modifiable determinants, such as sterility and immunological state [65‐67]. Furthermore, post-procedural infection rate is highly linked to the invasiveness and complexity of the procedure carried out and the instrumentation used [68].

The overall mean age for those presenting with SI in this review was 55 years; previous research supports the notion that SI is a disease of older people [69, 70]. Amadoru et al. specifically looked at differences in presentation and outcomes between younger and older patients with SI [13]. Their findings suggested that older patients with SI present with fewer typical clinical features including fever and rigour and tend to seek medical advice later than younger patients. This could explain why age is a predictor for medical treatment failure [71]. However, the older population are also more likely to suffer from multi-morbidities which can suppress the immune system making them more susceptible to SI. A past history of TB is a question routinely asked by clinicians but it is only mentioned in one paper in this review [72]. Whilst Spinal TB is rare it is on the rise. The incidence of TB is dependent on birthplace and environmental factors, for example, born in countries with a high burden of TB, social conditions such as living in urban overcrowded conditions and homelessness. However, in this review, the social conditions reported related to rural living and occupational exposure [73].

The most commonly reported features of SI reported in this review were back pain, fever and neurological dysfunction which has been described elsewhere as the classic triad [54]. The classic triad is the hallmark of the SI but reliance on patients presenting with these features is likely to result in missed cases or diagnosis in the late stages as not all patients will present with all three features. The most common reported symptom in this review is that of back pain (71%), however, back pain is usually a benign condition with a prevalence of 80% in the general population [74], which can present a diagnostic challenge for clinicians and in isolation it does not aid in identifying SI. In primary care, it is often back pain symptoms that prompt patients to seek help in the first place. Equally neurological symptoms are also reported but are also prevalent in the general population. SI is a progressively worsening disease with neurological deficit occurring in the later stages, whereas in the general population the expectation would be that after a period of time these neurological symptoms would resolve. Fever was the second most reported clinical feature of infection in this review. However, it was only prevalent in 55% of patients suggesting that a lack of fever cannot rule out SI and clinicians should not necessarily be reassured by its absence. These findings are consistent with other literature [70, 75‐77], with Davis et al. reporting the ‘classic triad’ to be 8% sensitive and 99% specific [54].

Overall, Staphylococcus aureus was the most reported cause of SI. Whereas Viral, Prionic and Parasitic causative pathogens were not reported at all. This is comparable with the wider literature as other systematic reviews also report bacterial SI, in particular, Staphylococcus aureus, to be the dominant causative SI agent [70, 76]. Though rare and not reported in this review, Viral and Parasitic SI do exist [78, 79]. In addition to this, Prionic diseases are even rarer, they are diseases of the nervous system affecting both humans and animals, there is no evidence of them affecting the musculoskeletal system, in particular, the spine [80].

Despite their low sensitivity (between 43 and 75%) [81], plain radiographs have traditionally been considered as the first step for assessing vertebral destruction [82, 83]. CT and MRI scans are highly sensitive, the results in this review show that CT and MRI scans were the most accurate radiological tests for diagnosing SI. Though, it is suggested that CT scans fail to properly detect epidural abscesses and spinal cord lesions, as such they are the procedure of choice only when MRI cannot be performed [82]. Finally, due to its diagnostic accuracy and non-invasive nature, MRI is the imaging of choice when investigating suspected SI, in particular, the early stages. MRI sensitivity, specificity, and accuracy for detecting SI are reported as 96, 92, and 94%, respectively [83‐86]. Nevertheless, advanced diagnostic tools such as MRI and CT scans are costly and less available in primary care and low-income settings [87, 88], using them as the first diagnostic step for back pain patients is inefficient and wasteful. Therefore, it is important and cost-effective for clinicians to be equipped with the red flags to screen and decide whether further diagnostic tests are needed [87]. In one of the studies reviewed, Ferrari looked at the outcome of patients with spinal pain who were not referred to advanced imaging unless they presented with red flags, he found that using red flags as the threshold before ordering advanced imaging has a low threshold for missing serious spinal pathology [31].

To the author’s knowledge, this is the first review to investigate the use of red flags to identify SI. This review has identified gaps in research and knowledge and recommends future areas of research [89]. Our findings should be understood in the context of some limitations. Firstly, as we were not able to identify studies presenting diagnostic accuracy of red flags for spinal infection, as a result, we presented a descriptive review of the characteristics of people with spinal infection. As there was no control group to facilitate comparison of different determinants and clinical features, we do not know whether many of these features were more common in people with spinal infection compared to those without. As these results only tell us what features are common in people with spinal infection, they cannot be relied upon for diagnosis. Secondly, most of the studies included in this review were from high income (HI) countries, yet the global burden of infections such as Spinal TB falls mostly on low income and low middle-income countries (LMI); countries where TB, Brucellosis and HIV are endemic [90]. This might suggest selection bias (Table 1) but may also be as a result of underdiagnoses and underreporting of SI in LMI countries. LMI countries tend to have weak public health infrastructure and poor health-care research capacity compared to HI countries [91]. Furthermore, the disparity in the burden of disease in LMI countries and HI means efforts and resources are prioritised to different diseases [92]. Though the potential for publication bias must not be ruled out, there is also some research alluding to the possibility of an editorial bias against the academic output from low-income countries [93, 94].

The current evidence surrounding red flags for SI remains of low quality and clinical features alone should not be relied upon to identify SI. When a patient arrives in clinic understanding determinants of possible SI should initially be considered. These risks include immunosuppression due to co-morbidities or drug use and environmental factors (surgery and social circumstances). The prevalence of these determinants combined with the presentation of clinical features of spinal pain with possible neurological dysfunction and fever should lead to a lowered threshold (or heightened index of suspicion) to further investigate and diagnostic tests should be performed to rule out SI. MRI is the imaging technique of choice when investigating suspected SI. The authors, therefore, suggest grouping red flags into a priori and a posteriori. The a priori red flags being the comorbidities and determinants that may be present leading to the development of SI, and the a posteriori being the signs symptoms and clinical features that are present after the onset of SI. Temporally stratifying red flags in this manner will aid the clinician in building a picture based on both potential risk of SI and presentation consistent with SI.