Red flags presented in current low back pain guidelines: a review

Overview of recommendations on red flag screening in low back pain guidelines.

We searched for clinical guidelines in primary health care concerning adults with low back pain (date of last search January 30, 2016). Our starting point was a previously published review article including 15 national and international guidelines for diagnosis and treatment of low back pain [12]. First, we checked for updates of these 15 guidelines. Additionally, we searched for other clinical practice guidelines using electronic databases: Medline, PEDro (key words: low back pain, practice guidelines, clinical guidelines), National Guideline Clearinghouse (http://​www.​guideline.​gov; key word: low back pain), and National Institute for Health and Clinical Excellence (http://​www.​nice.​org.​uk; key word: low back pain). Furthermore, we performed searches via Google, performed snowballing and citation tracking on publications found and consulted experts in the field. The search was aimed at finding all the clinical guidelines that exist. No language or date restriction was applied. We defined ‘guideline’ as: “… contains systematically developed statements including recommendations intended to optimize patient care and assist physicians and/or other health care practitioners and patients to make decisions about appropriate health care for low back pain under the auspices of a medical specialty association; relevant professional society; public or private organization” (according to the National Guideline Clearinghouse). When one country had more than one guideline, we selected the most recent multidisciplinary guideline.

We extracted data on the number and type of red flags for serious pathology for each guideline using a standardized form. For each red flag, we scored if the red flag was supported by the literature presenting its diagnostic accuracy (e.g., data on sensitivity/specificity, predictive values, etc.), if it was supported by consensus of the guideline committee only, or if no information was given to support the endorsement of red flags. One author (NP) extracted the data, which were checked by a second (APV). The data were summarized using tables.

First, of the original 15 guidelines of previously published review article [12], we excluded the European guideline for chronic low back pain [13], given that red flags were presented in the European guideline for acute low back pain only [4]. Eight countries updated their guideline (Austria, Canada, Finland, Germany, Netherlands, Norway, Spain, and United States) [17‐24]; of three countries, we found more than one updated guideline (Austria, Netherlands, and United States). We found two updated guidelines from Austria including an update of a multidisciplinary guideline from 2007 and one specifically for radiologists [25] of which we selected the multidisciplinary one [17]. The updated guidelines from The Netherlands included a multidisciplinary guideline and one specifically for physiotherapists [26] of which we selected the multidisciplinary one [21]. The United States had two multidisciplinary guidelines [24, 27] and one specifically for physiotherapists [28] of which we selected for this overview the latest multidisciplinary guideline [24] linked to a website [29]. The guidelines of Finland and Norway were not available in English, so colleagues were contacted to extract the relevant data.

Next, we performed a broad search aiming to identify additional guidelines. In total, we identified 21 guidelines, of which four were excluded (see above) as we selected one guideline per country. We found three new guidelines (Philippines, Malaysia, and Mexico) of which one guideline (Mexico) [14] could not be retrieved [15, 16]. Finally, 16 discrete guidelines were included in this review (see Table 1).

The guidelines were published between 2000 (France) and 2015 (Finland), with the publication date of one guideline unknown (Malaysia). The target population was mostly adults (>15 or 18 years) with low back pain. Nine guidelines used the term nonspecific low back pain, three guidelines also included people with radiculopathy, four guidelines specifically focused on patients with acute low back pain (defined as a duration less than 3 months), and one guideline included patients with acute and/or recurrent low back pain (New Zealand) (see Table 1).

All guidelines recommended screening patients for suspected serious pathologies by using red flags. Eight guidelines presented red flags for various forms of serious underlying disease specifically (Australia, Finland, France, Germany, Italy, Netherlands, United Kingdom, USA) [19, 20, 24, 30‐33]; one guideline combined red flags for malignancy and infection (Canada) [18]; two guidelines presented general red flags, but separately for cauda equina syndrome (Europe, New Zealand) [4, 34]; and five guidelines presented red flags without targeting a specific underlying pathology (Austria, Malaysia, Norway, Philippine, Spain) [15‐17, 22, 23].

The pathologies most commonly referred to in the guidelines were: malignancy (9 guidelines); fracture (9 guidelines) of which one guideline focused on compression fractures only (Finland), and three guidelines distinguished between traumatic and osteoporotic fractures (Canada, Netherlands, United States); infection (8 guidelines) of which one focused only on ankylosing spondylitis (Netherlands), two guidelines separately focused on infection and spondyloarthropathies (Italy, United States) and two on infection and ankylosis spondylitis (Canada, France); cauda equina syndrome (7 guidelines); aneurism (3 guidelines); myelopathy (United States) and severe spondylolisthesis (Netherlands). We found 46 different guideline endorsed red flags for malignancy, fractures, infection and cauda equina syndrome (see Table 2).

None of the guidelines provided a detailed definition of each red flag nor a precise description of when a red flag could be considered positive, e.g., when does a patient have ‘osteoporosis’ or ‘loading pain’. For the presentation, we clustered red flags when the wording suggested a comparable definition or description, e.g., some guidelines state as a red flag for a fracture the ‘use of steroids’ or ‘corticosteroid use’, while others add the prefix ‘systemic’, ‘chronic’ or ‘prolonged’. Others categorize corticosteroid use with ‘immunosuppressive use’.

There are a wide variety of recommended red flags for malignancy. In total, 14 red flags were specifically related to malignancy. Two red flags were mentioned in almost all guidelines: a ‘history of cancer’ was included in all guidelines, and ‘unexplained or unintentional weight loss’ was included in all but three guidelines (Spain, United Kingdom and United States). Almost all guidelines mentioned pain as a red flag, but the description of the kind of pain differed. Most often ‘pain at rest’ or ‘pain at night’ was considered as a red flag. Nine red flags for malignancy were mentioned in a single guideline only: ‘multiple cancer risk factors (unspecified)’ and ‘strong clinical suspicion’ (United States), ‘reduced appetite’ and ‘rapid fatigue’ (Germany), ‘elevated ESR’ and ‘general malaise’ (The Netherlands), ‘fever’, ‘paraparesis’ and ‘progressive symptoms’ (Finland). One guideline presents a combination of red flags for malignancy: ‘Patient over 50 (particularly over 65), with first episode of severe back pain and other risk factors for malignancy, such as history of cancer/carcinoma in the last 15 years, unexplained weight loss, failure of conservative care (4 weeks)’ (Canada), see Table 1.

In total, 11 red flags were considered to be specifically related to fractures. All but one guideline (United Kingdom) mentioned ‘major or significant trauma’ as a red flag, and ‘use of steroids or immunosuppressors’ was mentioned as a red flag in nine guidelines. Seven guidelines mentioned ‘older age’ as a red flag, but the cut-off varied between 50 and over 70 years. Five red flags for fracture were mentioned in a single guideline only: ‘previous fractures’ (Italy), ‘low body weight’ and ‘increased thoracic kyphosis’ (The Netherlands), ‘structural deformity’ (Canada) and ‘minor trauma’ (Germany). Three guidelines mentioned female gender as a red flag specifically for osteoporotic/compression fractures (Italy, Netherlands, and United Kingdom). Two guidelines presented a combination of red flags to be related to (compression) fractures: ‘minor trauma if age is over 50 and there is a history of osteoporosis and corticosteroid use’ (Australia) or ‘severe onset of pain with minor trauma, age >50, prolonged steroid intake or structural deformity (for compression fracture)’ (Canada).

Overall, 13 red flags were recommended in relation to infection. The most frequently mentioned red flags were: fever (12 guidelines), use of corticosteroids or immunosuppressant therapy (10 guidelines) and intravenous drug abuse (11 guidelines). Five guidelines mentioned pain as red flag: ‘pain worse at night’ (France); ‘intense nocturnal pain’ (Germany); ‘night and rest pain’ (Italy); ‘fever/chills in addition to pain with rest or at night’ (United States) or ‘bone tenderness over the lumbar spinous process’ (Australia).

Nine red flags were recommended in relation to cauda equina syndrome (CES), of which two were frequently mentioned: ‘saddle anesthesia (perineal numbness)’ and ‘(sudden onset of) bladder dysfunction’, both in nine guidelines. Only one red flag (‘sciatica’) is endorsed by one guideline (France).

Seven guidelines presented 23 red flags unrelated to a specific disease (Austria, Europe, New Zealand, Norway, Philippine, Spain, Malaysia). Of these red flags, some were endorsed for a specific disease by other guidelines; 9 were endorsed for malignancy, 4 for fracture, 3 for infection and 6 for CES. In total, three unique red flags were presented and 6 unique pain items of which ‘pain under 20 or over 50 years’ and ‘thoracic pain’ were the most presented in 6 and 5 guidelines, respectively, see Table 3.

Nine guidelines (Austria, Canada, Europe, Finland, Germany, Norway, Philippine, Spain, United States) based their recommendations for red flags on previous guidelines, of which two also included additional references (Europe, United States) and one explicitly stated that there was a consensus procedure (Germany), see Table 1. Four guidelines did not present any reference supporting their choice of red flags (Italy, Netherlands, New Zealand, United Kingdom); two guidelines presented references to support the choice of red flags (Australia, Malaysia), see Table 1. One guideline (France) presented diagnostic accuracy data (sensitivities and specificities) for the individual red flags. In the short version of the French guideline they only presented these data for two red flags for malignancy (‘history of malignancy’, ‘unexplained weight loss’), while in their full paper all published accuracy data for red flags for malignancy and ankylosing spondylitis were presented.

We included 16 discrete guidelines for the management of patients with low back pain in the primary care setting presenting 46 different red flags for the four main categories of serious underlying pathologies (malignancy, fracture, infection and CES). Five guidelines endorsed red flags without targeting a specific pathology. Overall almost all guidelines endorsed two red flags for malignancy (‘history of cancer’ and ‘unintentional weight loss’) and two for fracture (‘major or significant trauma’ and ‘use of steroids or immunosuppressors’). Red flags such as ‘pain at night’ or ‘at rest’ were recommended for various underlying pathologies. Existing accuracy data supporting the choice and endorsement of red flags was rarely used in the selected guidelines.

Our findings that most guidelines vary in terms of the red flags endorsed, and contain little information on the diagnostic accuracy of the red flags, are in line with previous studies [12, 35, 36]. Although all guidelines present red flags and recommend their use to screen for serious pathology, only a few provide evidence of their accuracy. The American Pain Society presented an ‘Evidence review’ on the clinical evaluation and management of low back pain with a date of last search in July 2008 [37]. This report presents a clear overview of the known diagnostic accuracy of red flags for the detection of pathology including malignancy, fracture, infection and CES. Several guidelines have been developed or updated since [27, 38], but without presenting the level of evidence to endorse red flags as cited in the evidence report (or refer to it). For example, the United States guideline (2014) endorses a greater number of red flags, but seldom underpins their recommendations with evidence.

Change in evidence is one of the reasons for updating guidelines [39]. New evidence can prompt the update of a guideline, but our review suggests that evidence related to screening for serious pathology has not prompted update of the guidelines studied. One exception is the United States physiotherapy guideline (excluded as it was not multidisciplinary), which presents a comprehensive table with red flags and their accompanying diagnostic accuracy data were available [28].

A recent paper summarizing two Cochrane diagnostic systematic reviews found nine studies evaluating the diagnostic accuracy of in total 29 red flags for fracture and 24 for malignancy [8]. There were differences in the red flags that demonstrated diagnostic utility and those endorsed by guidelines. It makes sense that red flags that do not show acceptable diagnostic accuracy are not endorsed in guidelines. Nevertheless, most red flags endorsed by the guidelines have never been evaluated for their diagnostic accuracy; 8 out of 14 red flags for malignancy and 6 of the 11 red flags for fracture.

For malignancy, the systematic review concluded that only ‘history of cancer’ is based on acceptable validity; it increases the probability of having cancer from 0.7 % (pre-test) to 33 % (95 % CI 22–46 %) [8]. Nevertheless, this conclusion is based on one study set in primary care and another in an emergency department where 36 % of patients were referred to because of a significant trauma [40, 41]. It is argued that ‘history of cancer’ is not very useful as a red flag, as it does not consider the type of primary cancer or the time since diagnosis [42]. For example, a history of recent (less than 5 years) breast cancer might be a more useful red flag than a history of leukemia greater than 20 years ago.

According to the systematic review, the red flags ‘severe trauma’, ‘use of corticosteroids’, ‘older age’ and ‘presence of a contusion or abrasion’ each increased the probability of a fracture from 4 % (pre-test) to between 9 and 62 % [8]. Three of these red flags were most often mentioned in the guidelines, but one (‘presence of a contusion or abrasion’) was absent from all guidelines.

An Australian population-based prospective cohort study of 1172 consecutive patients presenting to primary care for low back pain calculated the increased probability of fracture when a combination of red flags were positive [43]. When any three of the red flags ‘female’, ‘age >70’, ‘severe trauma’, and ‘prolonged use of corticosteroids’ were present, the probability of fracture increased from 4 % (pre-test) to 90 % (95 % CI 34–99 %). Combining red flags to inform clinical decision-making remains largely unexplored in the literature. In addition, external validation of red flags used in combination to raise suspicion of disease is even more rare.

The European guideline reports explicitly “If any of these are present, further investigation (according to the suspected underlying pathology) may be required to exclude a serious underlying condition, e.g., infection, inflammatory rheumatic disease or cancer” [4]. Later in their guideline, the advice is diluted: “Individual ‘red flags’ do not necessarily link to specific pathology but indicate a higher probability of a serious underlying condition that may require further investigation. Multiple ‘red flags’ need further investigation.” Nevertheless, the majority of guidelines inferred that the presence of a red flag was absolute by recommending further diagnostic workup (e.g., advanced imaging). Given that up to 80 % of patients presenting to primary care may have at least one positive red flag [43], when combined with weak evidence in support of many red flags, this advice may cause harm to many patients through unnecessary imaging (increased radiation and health care costs), unnecessary alarming the patients (resulting in reduction of quality of life) and unnecessary treatment (including unnecessary surgery) [42, 44].

For this overview, we searched for clinical guidelines. This required a broad and sensitive search of electronic databases, the World Wide Web and personal communication with experts in the field as most often clinical guidelines are made by (a combination of) professional bodies and published on national websites in their native languages. Not all guidelines have been translated into English, so it is possible that some non-English guidelines have been missed. Notwithstanding, we believe this would not have significantly influenced our conclusions. Furthermore, we selected a multidisciplinary guideline when more than one guideline per country was available. This resulted in an a priori selection of guidelines that might have influenced our conclusions. For instance, the United States physiotherapy guideline endorsed another set of red flags with accompanying diagnostic accuracy data where available, compared to the included multidisciplinary guideline [24, 28]. Hence, we have clustered red flags based on their assumed definition or description. Lack of standardization was evident when defining or describing red flags. For example, red flags related to nocturnal pain comprised ‘increasing pain at night’, ‘intense night pain’, ‘unbearable night and rest pain’, ‘pain at night not eased by prone laying’ or ‘pain with recrudescence at night’. Similarly, there was a range of age cut-off for suspicion of fracture (>50, >60, >70, and ‘older age’). This lack of standardization may introduce confusion for the clinician, reduce the ability to describe red flags, and decrease the accuracy of any pooled results. Nevertheless, we do not think this clustering has influenced our conclusions.

We found a wide variety of red flags, a lack of standardized description, and an overall lack of (presentation of their) diagnostic accuracy supporting their use. This highlights the need for a (limited) core set of red flags, ideally underpinned with acceptable diagnostic accuracy and endorsed by all guidelines. Next, the conduct of high quality diagnostic accuracy studies with clear operational definitions for each red flag should be commenced to assess the validity of these red flags individually or in combination (diagnostic model). Furthermore, guidance for primary care clinicians on how to ask for red flags needs attention, as there appeared little consensus between physiotherapists in a small qualitative study [45]. Given that the risk of serious disease for patients who present to primary care with low back pain is already low (e.g., infection <0.1 %, cancer about 0.7 %), red flags are of limited use when ruling out pathology. This is in contrast to other diagnostic models such as the Ottawa ankle rule where a negative test result may decrease the probability of ankle fracture from about 15 % to less than 2 % [46‐48]. Therefore, diagnostic models that demonstrate an increased ability to detect serious disease should be explored. Some diagnostic models of red flags for fracture have been developed to identify patients with a greater risk of a fracture (up to 90 %), but they are yet to be validated [43, 49].