Abstract
Background In a small proportion of patients experiencing unspecified back pain, a specified underlying pathology is present.
Objective The purposes of this study were: (1) to identify the prevalence of physician-specified causes of back pain and (2) to assess associations between “red flags” and vertebral fractures, as diagnosed by the patients' general practitioner (GP), in older adults with back pain.
Methods The Back Complaints in the Elders (BACE) study is a prospective cohort study. Patients (aged >55 years) with back pain were included when consulting their GP. A questionnaire was administered and a physical examination and heel bone densitometry were performed, and the results determined back pain and patient characteristics, including red flags. Participants received a radiograph, and reports were sent to their GP. The final diagnoses established at 1 year were collected from the GP's patient registry.
Results Of the 669 participants included, 6% were diagnosed with a serious underlying pathology during the 1-year follow-up. Most of these participants (n=33, 5%) were diagnosed with a vertebral fracture. Multivariable regression analysis showed that age of ≥75 years, trauma, osteoporosis, a back pain intensity score of ≥7, and thoracic pain were associated with a higher chance of getting the diagnosis of a vertebral fracture. Of these variables, trauma showed the highest positive predictive value for vertebral fracture of 0.25 (95% confidence interval=0.09, 0.41) and a positive likelihood ratio of 6.2 (95% confidence interval=2.8, 13.5). A diagnostic prediction model including the 5 red flags did not increase these values.
Limitations Low prevalence of vertebral fractures could have led to findings by chance.
Conclusions In these older adults with back pain presenting in general practice, 6% were diagnosed with serious pathology, mainly a vertebral fracture (5%). Four red flags were associated with the presence of vertebral fracture.
Most patients with back pain experience unspecified back pain, although in a minority of the cases, a specified underlying pathology is present.1–3 Of these specified causes, vertebral fractures, malignancies, infection, cauda equina syndrome, and ankylosing spondylitis are considered serious pathologies and are estimated to account for 1% to 5% of low back pain in primary care.4,5 Vertebral fractures are the most common underlying serious pathology in patients with back pain.2 These fractures particularly are more common in older patients.1 In patients with vertebral fractures, disability is more common than in patients with unspecified back pain.6 Identifying patients with vertebral fractures is necessary because vertebral fractures can also be an indicator for osteoporosis.7 In turn, treatment of osteoporosis can prevent future vertebral fractures.8,9
To identify specified causes of back pain, most clinical guidelines recommend the use of “red flags.”10,11 These alarming symptoms, derived from history taking or physical examination, or both, are suggested to have an association with serious pathology as a cause of back pain. The prevalence of serious pathology as a cause of back pain rises with age, and red flags, consequently, may be more important in patients aged >55 years.12 Guidelines differ in their recommendations regarding which red flags should be used and what the consequence should be when red flags are present. For example, the Dutch guideline for general practitioners (GPs) makes no statement about the direct consequences if red flags are present, and further diagnostic actions are at the discretion of the patient's GP.13 The guideline of the American Pain Society advises diagnostic imaging and testing if, based on the presence of red flags, a serious underlying pathology is suspected.4 A recent systematic review showed that only a limited number of red flags are of diagnostic value.14 The variables older age, corticosteroid use, and significant trauma are red flags for a vertebral fracture.14,15 Most individual red flags show poor diagnostic accuracy. There are indications that predictive performance can be improved when combinations of red flags are used. A first step in combining red flags was a diagnostic model for detecting vertebral fractures in primary care, described by Henschke et al.2 However, this diagnostic model has not yet been evaluated in other populations.
The aims of this study were: (1) to identify the prevalence of physician-specified causes (eg, vertebral fractures) of back pain, as identified by the patients' GP, in older adults with back pain and (2) to assess the associations between red flags and vertebral fracture in a subgroup of patients diagnosed with vertebral fracture by their GP.
Method
Data from the Back Complaints in the Elders (BACE) study, a prospective observational cohort in the Netherlands, was used for this study. Patient inclusion (n=675) took place between March 2009 and September 2011 in a representative sample of 49 general practices in and around Rotterdam. Patients aged >55 years were included when they consulted a GP with a new episode of back pain. Back pain was defined as pain in at least a part or the whole region from the top of the shoulder blades to the first sacral vertebra, with or without pain radiation to the leg. If a patient had not visited a GP with the same back pain in the preceding 6 months, this was considered a new episode. Patients were invited to join the study by their GP during the consultation or in writing within 2 weeks after the consultation. Patients were excluded if they were unable to fill out the questionnaires due to cognitive impairment, were not able to read and write in Dutch, or were unable to undergo physical examination (eg, patients using wheelchairs). Details of the BACE study design are described elsewhere.16
Data Collection
After inclusion in the BACE study and providing written informed consent, baseline measures included a questionnaire, a structured physical examination of the back, a radiograph of the back, and heel bone densitometry. The questionnaire asked about participant characteristics, features of the back pain, and the presence of red flags. Red flags for vertebral fractures were assessed in a questionnaire during the physical examination and were retrieved from the GP's patient registry. The red flags for vertebral fractures were age, sex, trauma, sudden decrease in height, acute onset of pain, osteoporosis, and prolonged corticosteroid use. Red flags were chosen based on those reported in clinical guidelines.4,11,13,17,18 Corticosteroid use in the year before consulting the GP with back complaints was retrieved from the GP's patient registry because this red flag was not included in the questionnaire. Other determinants that were considered important for diagnosing vertebral fractures were percussion tenderness of the spine,19 disability, back pain intensity score, osteoarthritis in the hip or knee, and thoracic pain. These variables also were assessed in the questionnaire or during the physical examination.
A radiograph of the lumbar spine was performed. If participants had back pain in the region of the thoracic spine, a thoracic radiograph also was performed. The radiographic findings reported by the radiologist were sent only to the GPs, who could use this information in their final diagnoses. Diagnoses were established by the GPs as they do in regular daily practice using the clinical guidelines for GPs regarding back pain.13,20
The final diagnosis regarding the back pain was determined at 1-year follow-up from the GP's patient registry. For each patient included in the cohort, the corresponding diagnosis was retrieved via the associated International Classification of Primary Care (ICPC)21 code (L03–unspecified low back pain without radiating pain, L86–unspecified low back pain with radiating pain, L76–fracture of the musculoskeletal system, L71–malignancy of the musculoskeletal system, L84–osteoarthritis and spondylosis) and by searching in the free-text field. All GPs checked the collected diagnoses of their participating patients. The final diagnoses were mostly manifest shortly after consulting the GPs; however, the 1-year follow-up was decided on for the present study to ensure that all diagnoses would be evident at that time.
Diagnoses collected were categorized as unspecified back pain and specified back pain. Of the specified back pain diagnoses, vertebral fracture, spinal malignancy, ankylosing spondylitis, vertebral infection, and cauda equina syndrome were considered serious pathologies.
Participants' perceived severity of back pain averaged over the previous week was measured on an 11-point numeric rating scale (NRS),22 with 0 representing “no pain” and 10 representing “worst pain ever.” Severe pain was defined as an NRS pain score of ≥7, as it is shown to be a discriminative cutoff value of severe pain in patients with back pain.23 Disability was measured with the Roland-Morris Disability Questionnaire (RDQ).24 The RDQ scores range from 0 to 24, and a score of ≥17 was defined as severe disability. Quality of life was measured with the 36-Item Short-Form Health Survey (SF-36), Dutch version.25 The SF-36 measures 8 dimensions: physical function, role–physical function, bodily pain, general health, vitality, social function, role–emotional function, and mental health. These 8 dimensions can be recoded into 2 summary scores: a physical component summary score and a mental component summary score. Each dimension and summary score is scored from 1 to 100, with a higher score representing better health.26,27 Summary scores were calculated with adapted z-score values, in view of the higher mean age of our study population.25
Depression was measured with the Center for Epidemiologic Studies Depression Scale (CES-D) (range: 0–60 points). Patients with a higher score are more prone to depression.28 Pain catastrophizing was measured with the Pain Catastrophizing Scale (range: 0–52 points), with a higher score representing a higher risk for catastrophizing.29 Back beliefs were investigated with the Back Beliefs Questionnaire (range: 9–49 points), with a higher score representing more positive thoughts on recovery.30 Lifestyle factors included smoking (yes/no) and drinking alcohol. Alcohol consumption was measured with the Alcohol Use Disorders Identification Test (AUDIT-C).31,32 Women were defined as possible hazardous drinkers if they scored ≥3 on the scale, and men were considered possible hazardous drinkers if they scored ≥4. During the physical examination, body weight and height were measured and converted to body mass index (BMI). Low education level was present if the participant had no education or if the highest level of education was primary school or lower vocational education. Trauma, sudden decrease in height, acute onset of pain, and osteoarthritis in the hip or knee were self-reported. Corticosteroid use was defined as oral or inhalation corticosteroid use for more than 90 days. Percussion tenderness of the spine was assessed in the physical examination. Heel bone densitometry was performed using Achilles quantitative ultrasound assessment as a proxy for identification of osteoporosis. In every participant available for the physical examination, both heel bones were measured, a T-score of ≤2.5 on at least one side was considered as osteoporotic.33
Data Analysis
Descriptive statistics were used to present prevalence of back pain, participant characteristics, back complaint characteristics, psychological factors, and red flags, using frequencies for categorical data and mean and standard deviation for continuous variables. Analyses were performed using complete cases. To assess the associations between red flags and vertebral fractures, red flags and other determinants that were considered important for diagnosing vertebral fractures were separately included in a univariable logistic regression analysis, with diagnosis at 1-year follow-up as outcome. Variables scoring a P value of <.05 were examined for correlation using the Pearson test; if variables were correlated (r>.6), only one variable (after consensus) was entered in the multivariable logistic regression analysis (backward Wald method, entry .05, removal .10). The multivariable regression analysis was used to find the best fitted model for vertebral fractures in our study population.
For all red flags, the sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios were calculated to establish diagnostic accuracy. Reported P values were from 2-sided tests, and a P value of <.05 was defined as statistically significant. All analyses were performed using SPSS software (version 20 for Windows, SPSS Inc, Chicago, Illinois).
Role of the Funding Source
The study was funded by the Department of General Practice, Erasmus University Medical Center, Rotterdam, the Netherlands, and the Coolsingel Foundation, Rotterdam, the Netherlands. This study also was partly funded by a program grant from the Dutch Arthritis Foundation.
Results
Participant Characteristics
Of the 1,402 invited patients, 675 were included in the BACE study; 118 patients did not meet the inclusion criteria, 291 patients were not willing to participate, and 318 patients did not respond to the invitation. Six patients (0.9%) 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, the diagnosis could not be retrieved.
Table 1 presents the baseline characteristics of the 669 included participants. The age of the included participants ranged from 55 to 91 years (X̅=66, SD=7.7); 40% of the participants (n=269) were male; mean severity of back pain was 5 (SD=2.7); and 87 participants (13%) reported that they experienced a first episode of back pain. Of all 669 participants, 95% underwent a radiograph of the back.
Baseline Characteristics of Included Participants Aged >55 Yearsa
Prevalence
Of all 669 participants, unspecified back pain with or without radiation below the knee was diagnosed in 384 of them (57%) (Tab. 2). Serious pathology was present in 6% of our study population; vertebral fracture was the most common diagnosis and was seen in 33 participants (5%), of whom 30 were diagnosed with an osteoporotic vertebral fracture. Four participants (1%) were diagnosed with a spinal malignancy, and no spinal infections or cauda equina syndrome were diagnosed. Of the other specified diagnoses that were not defined as serious pathology, vertebral osteoarthritis was the most common (173 patients, 26%); disk herniation was present in 5% of the participants (Tab. 2).
General Practitioners' Diagnosis Within 1 Year of Follow-up (N=669)a
Association Between Red Flags and Vertebral Fracture
In the univariable analyses, age of ≥75 years, prolonged corticosteroid use, trauma, osteoporosis, severe disability, a back pain intensity score of ≥7, and thoracic pain were individually associated with the diagnosis of vertebral fractures (Tab. 3). These variables were included in the multivariable regression analysis. In this model, age of ≥75 years, trauma, osteoporosis, a back pain intensity score of ≥7, and thoracic pain were associated with the diagnosis of vertebral fractures.
Univariable and Multivariable Association of “Red Flags” With Vertebral Fractures in Participants Aged >55 Years With Back Paina
Table 4 shows the diagnostic value of the red flags. Age had a positive predictive value of 0.14 (95% confidence interval [CI]=0.07, 0.20) and a positive likelihood ratio of 3.1 (95% CI=2.0, 4.7). Osteoporosis had a similar positive predictive value of 0.14 (95% CI=0.07, 0.21) and likelihood ratio of 3.2 (95% CI=1.9, 5.2). The positive predictive value and the positive likelihood ratio of trauma were 0.25 (95% CI=0.09, 0.41) and 6.2 (95% CI=2.8, 13.5), respectively. These values for trauma were the highest among all variables and raise the probability from .05 to .25 when the test is positive. The negative likelihood ratio for trauma was 0.8 (95% CI=0.5, 1.3), which lowers the probability of a vertebral fracture if there was no trauma from .05 to .04. A diagnostic prediction model with 4 red flags combined did not increase these diagnostic values.
Diagnostic Value (95% Confidence Interval) of “Red Flags” and Other Determinants for Vertebral Fractures (N=33) in Participants Aged >55 Years With Back Pain (N=669)a
Discussion
Summary of Results
The present study assessed the prevalence of physician-specified underlying pathologies of back pain, as identified by the participants' GP, in older adults with back pain seen in primary care and the associations of red flags with vertebral fracture. Based on the final diagnoses of back pain retrieved at 1-year follow-up, 57% of the participants were diagnosed with unspecified back pain with or without radiation below the knee, and 6.1% of the participants were diagnosed with serious pathology (vertebral fracture, spinal malignancy, and ankylosing spondylitis). Vertebral fractures were the most common serious pathology (5% of the participants). Red flags associated with vertebral fractures were age of ≥75 years, trauma, osteoporosis, a back pain intensity score of ≥7, and thoracic pain. However, the positive predictive value and positive likelihood ratio of the combined red flags did not increase more than the values for trauma alone.
Interpretation of Findings
In our study population, vertebral fractures, malignancies, and ankylosing spondylitis accounted for 6% of the causes of back pain presented to the GP. Vertebral infection and cauda equina syndrome were not identified in our study population. The 6% prevalence in our study is similar to prevalences found in other studies performed in primary care, namely 1% to 5%.1,2,4 Most studies on specified causes of back pain describe only serious causes of back pain. Only one study performed in primary care reported all specified diagnosis of back pain.1 Comparing our distribution of specified causes of back pain with the previous findings, it appears that the prevalences of herniated disk, spinal stenosis, vertebral fractures, and spondylolisthesis are about the same in both populations. Only the prevalence of unspecified back pain was somewhat higher in the study by Deyo and Weinstein,1 and the prevalence of degenerative pathology, such as osteoarthritis, was higher in our study population. These findings were due mainly to our older study population because spine degeneration increases with age.34 The finding of osteoarthritis might not alter the management of back pain in general practice because there is no specific treatment for this group of patients.
In our multivariable model, age of ≥75 years, trauma, osteoporosis, and a back pain intensity score of ≥7 were associated with vertebral fractures in our population of older patients with back pain. Various cutoff points have been used to determine older age in studies investigating red flags in low back pain. We used the cutoff value of 75 years of age because a recent Cochrane review showed this cutoff value to be the most informative for detecting vertebral fractures in patients with back pain.15 The finding that patients with older age are more likely to have a vertebral fracture is consistent with other studies.2,34 Also, the result that trauma, as a red flag, is associated with a vertebral fracture is in line with other studies in primary care.14,15 However, only 21% of our participants with a fracture reported a trauma.
A back pain intensity score of ≥7 has not been shown in research on red flags to be associated with vertebral fractures. In tertiary care, the presence or absence of pain has been studied,35–38 although the use of a cutoff value in patients with back pain has not previously been tested. This finding should be further evaluated in other populations with back pain and might be of diagnostic value in patients with back pain seen in primary care.
Prolonged use of corticosteroids appeared to be related with a higher risk of vertebral fractures in older patients with back pain. Oral and inhalation corticosteroids were analyzed because both are associated with an increased risk of vertebral fracture.39–41 Other topical applications of corticosteroids (injection, dermal, nasal, ocular, and auricular) have not been reported to increase the risk of vertebral fracture.41,42 In the present study, prolonged use was defined as 3 months of corticosteroid use in the year previous to consulting a GP for back complaints. Two previous studies2,37 on corticosteroid use as a red flag in primary care showed likelihood ratios of 4.037 and 48.5,2 whereas the likelihood ratio was only 2.5 in our study. In our study population, corticosteroid use showed an association with vertebral fractures in the binary analysis, but this variable was not included in the final model. Osteoporosis is thought to be a mediator in the relationship between prolonged use of corticosteroids and fractures. It is possible that corticosteroids cause osteoporosis; however, in our population of older adults, it is plausible that osteoporosis also was present without the prolonged use of corticosteroids.
The diagnostic prediction model reported by Henschke et al2 also was assessed in our study population. We used the same red flags (age of >70 years, female sex, trauma, and prolonged use of corticosteroids) but could not reproduce the results as reported. When 2 red flags were present, the positive likelihood ratio in the study by Henschke et al was 15.5 (95% CI=7.2, 24.6), although we found a positive likelihood ratio of only 2.6 (95% CI=1.9, 3.7). Therefore, we used the red flags emerging from the multivariable regression analysis in our study population. The diagnostic values were somewhat better, but the positive likelihood ratios in this new diagnostic model were not as high as the likelihood ratios reported by Henschke et al.2 The differences between these diagnostic models might be due to the low prevalence of vertebral fractures in both studies or to the somewhat different population of older patients in our study.
Strengths and Limitations
To our knowledge, this is the first study assessing the diagnostic value of red flags in older patients with back pain seen in primary care. Older patients may be more prone to a specified cause of back pain, so it is important to evaluate whether red flags are present in this population and predict the most common type of serious pathology (ie, vertebral fractures).
If serious pathology is not identified at the first consultation, disease manifestation may become clearer over time. Also, almost all participants underwent a radiograph of the spine, and these reports were sent to the GPs. We assumed that specified diagnoses, even those not detectable on the radiograph of the spine, became evident within 1 year and were incorporated in the GP registry. Therefore, we identified the final diagnoses regarding specified causes of back pain in the registry of the participants' GP after the 1-year follow-up and assumed that, in this way, we retrieved the most reliable specified diagnoses.
One limitation of the study is the low prevalence of vertebral fractures. Because only 33 participants were diagnosed with a fracture and multiple variables were tested, this limitation could have led to findings by chance. The diagnoses of 6 participants were not available for follow-up, but we expect that they were random missing patients and did not have an important impact on the results of this study. Furthermore, incorporation bias could have occurred. We wanted to know if the red flags are useful to predict vertebral fractures, but they also could have been used by the GP to determine if a patient had a vertebral fracture. In that case, the strength of the associations could be overestimated. Although almost all of our study participants underwent a radiograph and the findings were sent to their GP, it is more likely that the diagnosis of vertebral fractures was obtained from the radiographic findings rather than from the red flags at baseline.
It is possible that patients with a vertebral fracture were missed, because not all included patients received a radiograph of the back and no standardized imaging protocol was used. The radiographs of the back were made at patients' nearest hospital, as is usual in normal daily practice. Any additional investigation for the back pain during follow-up was up to the GP; therefore, it is possible that we underestimated the prevalence of vertebral fractures. A final limitation might be that the diagnosis of vertebral fractures was obtained from radiographic findings, and it could have occurred that the fracture found on the radiograph was not linked to this episode of back pain.
Clinical Impact
In older adults with back pain seen in primary care, red flags alone and combined in a diagnostic prediction model were not very accurate in predicting vertebral fracture. Also, in this population of older adults, prevalence of vertebral fractures was low. Patients with a traumatic vertebral fracture need to be identified as soon as possible. These patients had a trauma; therefore, if this red flag is present, further diagnostic analysis should be performed. The fractures missed using this red flag alone were all osteoporotic fractures, which do not require immediate treatment, besides pain medication. A “wait-and-see” policy in the patients without a trauma can be justified, and further diagnostic testing can still be considered if pain lasts longer.
In these older adults with back pain seen in general practice, 43% were diagnosed with specified back pain and 6% with a serious underlying pathology. Most of these patients were diagnosed with a vertebral fracture (5%), and red flags associated with vertebral fracture were age of ≥75 years, trauma, osteoporosis, and a back pain intensity score of ≥7. The red flag of trauma had the highest diagnostic value, and a diagnostic prediction model did not increase this value.
Footnotes
Ms Enthoven, Dr Scheele, Professor Bierma-Zeinstra, Dr Bohnen, Dr van Tulder, Professor Peul, Professor Berger, Professor Koes, and Dr Luijsterburg provided concept/idea/research design. Ms Enthoven and Ms Geuze provided writing. Ms Enthoven, Ms Geuze, Dr Scheele, and Dr Luijsterburg provided data collection. Ms Enthoven, Ms Geuze, and Dr Luijsterburg provided data analysis. Ms Enthoven, Professor Bierma-Zeinstra, Dr Bohnen, Dr Bueving, Professor Koes, and Dr Luijsterburg provided project management. Professor Bierma-Zeinstra, Professor Koes, and Dr Luijsterburg provided fund procurement. Professor Bierma-Zeinstra and Dr Bueving provided facilities/equipment. Ms Geuze, Dr Scheele, Professor Bierma-Zeinstra, Dr Bohnen, Dr Bueving, Dr van Tulder, Professor Peul, Professor Berger, Professor Koes, and Dr Luijsterburg provided consultation (including review of manuscript before submission).
The Medical Ethics Committee of Erasmus University Medical Center, Rotterdam, the Netherlands, approved the study protocol.
The study was funded by the Department of General Practice, Erasmus University Medical Center, Rotterdam, the Netherlands, and the Coolsingel Foundation, Rotterdam, the Netherlands. This study also was partly funded by a program grant from the Dutch Arthritis Foundation.
- Received November 29, 2014.
- Accepted July 5, 2015.
- © 2016 American Physical Therapy Association
References
Issue highlights
