Back pain associated with ankylosing spondylitis (AS) is referred to as inflammatory back pain (IBP). The value of the clinical history in differentiating IBP from mechanical low back pain (MLBP) has been investigated in only a few studies. In this exploratory study, we sought to evaluate the individual features of IBP and to compose and compare various combinations of features for use as classification and diagnostic criteria.
We assessed the clinical history of 213 patients (101 with AS and 112 with MLBP) younger than 50 years who had chronic back pain. Single clinical parameters and combinations of parameters were compared between the AS and MLBP patient groups.
Morning stiffness of >30 minutes' duration, age at onset of back pain, no improvement in back pain with rest, awakening because of back pain during the second half of the night only, alternating buttock pain, and time period of the onset of back pain were identified as independent contributors to IBP. Importantly, none of the single parameters sufficiently differentiated AS from MLBP. In contrast, several sets of combined parameters proved to be well balanced between sensitivity and specificity. Among these, a new candidate set of criteria for IBP, which consisted of morning stiffness of >30 minutes' duration, improvement in back pain with exercise but not with rest, awakening because of back pain during the second half of the night only, and alternating buttock pain, yielded a sensitivity of 70.3% and a specificity of 81.2% if at least 2 of these 4 parameters were fulfilled (positive likelihood ratio 3.7). If at least 3 of the 4 parameters were fulfilled, the positive likelihood ratio increased to 12.4.
A new set of criteria for IBP performed better than previous criteria in AS patients with established disease. A prospective study is needed to validate the diagnostic properties of the new candidate criteria set in patients with early disease.
Back pain in ankylosing spondylitis (AS) is usually referred to as inflammatory back pain (IBP). IBP is the key clinical symptom of AS, but it can also be present in other spondylarthritides in which there is axial involvement (1–5). Since IBP has certain characteristic features, obtaining the clinical history has been proposed as a screening test to identify patients with AS among those who have chronic back pain (6, 7). However, the value of such a screening test has been questioned (8). Indeed, the clinical history has been considered to be of low to moderate value at best (9, 10). Given the low (∼5%) prevalence of AS and its early, preradiographic stages among patients with chronic back pain (11), a powerful test for IBP as the leading symptom in AS would be of great help not only for screening (12), but also for diagnostic evaluation of patients with back pain (13, 14).
The first precise clinical description of IBP dates back to the report by Hart and colleagues in 1949 (15). In the patients described in that report, the pain was located in one or both buttocks, also occasionally in the mid-lumbar region, and was accompanied by stiffness. Hart et al wrote, “A frequent feature of the pain and stiffness was the aggravation caused by immobility. Waking in the morning stiff and in pain, the patient gradually became more supple during the day, feeling at his best from the afternoon until bedtime. One patient noted that by frequent exercise, his condition was kept in check, but confinement to bed for any cause made him worse. Another woke himself up [every 2 hours] throughout the night to exercise his spine as otherwise, he suffered unduly in the morning.” Those investigators also noted that a few patients were exceptional, in that rest appeared to ease the pain. In some patients, heavy exertion or strain acted as an aggravator (15). The description by Hart nicely illustrates the typical features of IBP, but even at the outset, it indicates that these features do not apply to all patients.
The first set of criteria for IBP was proposed in 1977 by Calin et al (6). The criteria, which were derived from a study of 42 AS patients and 24 patients with back pain of other origin, consist of 5 features: 1) insidious onset, 2) age at onset <40 years, 3) duration of back pain ≥3 months, 4) associated with morning stiffness, and 5) improve with exercise. IBP was considered to be present if at least 4 of the 5 features were fulfilled (sensitivity 95%; specificity 76% [in patients with mechanical low back pain]) (6). The modified New York criteria for AS, which were published in 1984 (16), incorporated aspects of the findings by Calin et al and defined the back pain associated with AS as “low back pain and stiffness for more than 3 months, which improves with exercise, but is not relieved by rest.” Subsequent studies found the specificity of Calin's screening test to be about 75% (17, 18) and the sensitivity to be only about 23% (17) and 38% (16), respectively. The disappointingly low sensitivity led to new proposals for IBP criteria, such as “getting out of bed at night” as a single item, which had a sensitivity of 65% and a specificity of 79% in one study (17).
The recent broadening of treatment options for AS, from physical therapy plus nonsteroidal antiinflammatory drug (NSAID) treatment only to tumor necrosis factor (TNF) blockers as additional highly efficacious agents (19), has raised interest in the correct and early diagnosis of AS. The aim of this study was to evaluate again in an exploratory study the individual features of IBP and to compose various combinations of features and compare them with each other. Since AS is a differential diagnosis in patients with chronic back pain, whereas acute back pain often is nonspecific in nature and subsides without intervention within 3 months in 90% of patients (20, 21), and since ∼95% of all AS patients develop back pain before the age of 40–45 years (4, 22, 23), we decided to focus on patients younger than 50 years who had chronic back pain.
PATIENTS AND METHODS
Patients (n = 213) recruited for this study were seen between August 2000 and February 2001 at various hospital-based rheumatology, orthopedic surgery, and neurosurgery departments, as well as private practices, in Berlin, Germany. Patients had to be ≤50 years old at the time of study inclusion and had to have chronic low back pain for at least 3 months. The diagnosis in each patient had been made prior to this study by the treating rheumatologist or other specialist. All patients previously diagnosed as having AS satisfied the modified New York criteria (16). The diagnoses in patients with back pain due to causes other than AS (referred to herein as control patients with mechanical low back pain [MLBP]) were obtained from the treating physician. In all MLBP patients, a radiograph of the pelvis or lumbar spine that ruled out radiographic sacroiliitis was available and was evaluated by one of us (MR). Patients were recruited for this study nonconsecutively (“convenience sample”), provided that they met the inclusion criteria, had been diagnosed as having AS or MLBP, and were willing to participate.
Data on HLA–B27 status and C-reactive protein (CRP) levels were taken from the patient's clinical notes, if available. If such data were not available, blood was drawn at the time of the interview and then later analyzed at the Department of Clinical Chemistry (for CRP) and the rheumatology research laboratory (for HLA–B27), both of which are located at the Charité–Campus Benjamin Franklin. HLA–B27 status was assessed by polymerase chain reaction analysis, as described previously (24).
Since the study was rather exploratory, without an a piori hypothesis to be tested, there was no formal calculation of the sample size needed. We assumed that ∼100 patients per group should yield a representative spectrum of findings. Approval for this study was obtained by the local ethics committee, and all patients gave informed consent.
Administration of questionnaire and measurement of spinal flexion.
A face-to-face interview using a standardized questionnaire was conducted by the same trained examiner (AM) when patients came to the clinic for a routine visit. The examiner was not provided with the diagnosis of each of the patients who were to be interviewed. However, blinding of the examiner was incomplete, since patients were interviewed and examined on-site at the clinic at which they were treated (i.e., the rheumatology, orthopedic surgery, neurosurgery clinic, etc.), a circumstance that allowed for certain inferences to be made.
The questionnaire covered various aspects of disease history, such as age at onset of back pain, duration of disease, time period of onset of back pain (i.e., within 1 hour, 1 day, 1 week, between 1 and 4 weeks, or within 1 year), presence of morning stiffness (current and ever), duration of morning stiffness, improvement in back pain with exercise, improvement in back pain with rest, buttock pain, radiation of pain to the leg, awakening because of back pain during the night, time of awakening during the night, preceding events (such as trauma, mental or emotional stress, or infection), time point at which a doctor was first consulted, overall number of doctors consulted, number of days of sick leave taken over the previous 4 weeks, current intensity of back pain, and impact of back pain on well-being. In addition, anterior spinal flexion (by Schober's test) and lateral spinal flexion were measured (25, 26).
Chi-square tests and Student's t-tests were used to compare AS and MLBP patient groups and to identify parameters that discriminated between the 2 groups. Multivariate logistic regression was used to investigate which parameters were independent predictors of IBP. Stepwise forward logistic regression was applied to identify the parameter combination that best predicted IBP. In addition, best-subset logistic regression was applied to rank the possible combinations of 2, 3, 4, 5, and 6 parameters according to their ability to discriminate between AS and MLBP. The results of these analyses were used to build candidate criteria from combinations of the parameters. Sensitivity, specificity, positive and negative likelihood ratios (LRs), and 95% confidence intervals were calculated by the Wilson method (for sensitivity and specificity) or the normal approximation method (for positive and negative LRs). SPSS version 11.5 (SPSS, Chicago, IL) and SAS version 9.1 (SAS Institute, Cary, NC) computer software were used.
Basic demographic data.
A total of 213 subjects (101 with AS and 112 with MLBP) were evaluated. The mean ± SD duration of back pain was 12.9 ± 8.6 years in AS patients and 12.2 ± 9.6 years in MLBP control patients. The AS patients were slightly younger than the MLBP patients (mean ± SD age 35.9 ± 7.9 years versus 39.1 ± 7.5 years; P = 0.003), and there were slightly more men in the AS group (64.4%) than in the MLBP group (58.9%; P not significant). The diagnoses in the MLBP patients are shown in Table 1.
Table 1. Diagnoses in patients with mechanical low back pain
No. of patients
Osteoarthritis of the spine
Frequencies of individual parameters in AS and MLBP groups.
Age at onset of back pain. The number of patients with an age at onset of back pain <40 years was comparable in the AS and MLBP groups (96.0% of AS patients versus 86.0% of MLBP controls). An age at back pain onset of <35 years was significantly more frequent among the AS patients than among the MLBP control patients (93.1% versus 77.7%; P = 0.002), as was an age at back pain onset of <30 years (84.2% versus 59.8%; P < 0.001) and age at back pain onset of <25 years (64.4% versus 42.0%; P = 0.002) (Table 2).
Table 2. Frequencies of clinical features in patients with AS and MLBP and associated positive and negative LRs*
% of AS patients (n = 101)
% of MLBP patients (n = 112)
Positive LR (95% CI)
Negative LR (95% CI)
AS = ankylosing spondylitis; MLBP = mechanical low back pain; LR = likelihood ratio; 95% CI = 95% confidence interval.
Age at onset of back pain
Time period of back pain onset prior to evaluation
Within 1 hour
Within 1 day
Within 1 week
Between 1 and 4 weeks
>1 month (within 1 year)
Trauma (accident, lifting, carrying, etc.)
Duration of morning stiffness
No morning stiffness
Improvement in back pain
With exercise, but not with rest
Any buttock pain
Awakening at night because of back pain
At any time
First half of the night only
Second half of the night only
Both halves of the night
Preceding events. Accidents or traumas (lifting and carrying heavy weights, sports injuries, falls) preceding the onset of back pain were slightly more frequent among MLBP controls (20.5%) than among the AS patients (9.9%; P = 0.037) (Table 2). Enteritis as a preceding infection was more often reported by AS patients (5%) than by MLBP controls (0.9%; P not significant), but there were no differences in the reported occurrences of urogenital or upper respiratory tract infections.
Time period of onset of back pain. In the AS patients, back pain onset was reported for all of the time periods evaluated: very acute onset (within 1 hour) 23.7%, onset within 1 day 14.4%, within 1 week 19.6%, between 1 and 4 weeks 17.5%, and within 1 year 24.7% (Table 2). A very slow onset of back pain (i.e., within 1 year) was more frequent among the AS patients than among the MLBP controls (24.7% versus 9.8%; P = 0.008), whereas a very acute onset (within 1 hour) was reported significantly less often by AS patients than by MLBP controls (23.7% versus 41.2%; P = 0.01). There were no significant differences between AS and control patients for the other time periods of onset.
Morning stiffness. Significantly more AS patients than MLBP patients had a prolonged duration of morning stiffness. A duration of >1 hour of morning stiffness was reported by 48.5% of those with AS compared with 17.4% of those with MLBP (P < 0.001) (Table 2). Conversely, no morning stiffness or very brief morning stiffness (<10 minutes) was reported by only 16.8% of the AS patients but 55.0% of the MLBP controls (P < 0.016). The cut-off duration that best separated the 2 groups was a morning stiffness duration of >30 minutes (64.3% versus 24.7%; P < 0.001).
Improvement in back pain with exercise or rest. Of the AS patients, 78.2% reported improvement in back pain with exercise, whereas 50.0% of the MLBP controls reported improvement with exercise (P < 0.001). In contrast, improvement in back pain with rest was more frequent in the MLBP group (58.0%) compared with the AS group (31.7%; P < 0.001). The best separation between AS and MLBP patients was achieved by combining the 2 features into a single conditional parameter: improvement with exercise but not with rest was seen in 55.4% of the AS patients compared with 21.4% of the MLBP patients (P < 0.001) (Table 2).
Buttock pain. Both AS and MLBP patients frequently reported buttock pain (63.4% and 57.1%, respectively). If further specified, the unilateral type of buttock pain was reported more often by the MLBP patients than by AS patients (33.9% versus 14.9%; P = 0.001). In contrast, the alternating type of buttock pain was significantly more frequent in AS patients than in MLBP controls (36.6% versus 11.6%; P < 0.001). There was no difference in the frequency of bilateral buttock pain between the 2 patient groups (Table 2).
Awakening at night. Both AS and MLBP patients woke up frequently at night because of back pain (75.0% and 62.7%, respectively) (Table 2). Again, if specified further, the AS patients reported that they woke up significantly more often in the very early morning hours (second half of the night) as compared with the MLBP patients (44.0% versus 22.7%; P = 0.001). In contrast, there were no significant differences with regard to waking up only during the first half of the night or waking up during both halves of the night.
Laboratory results. As expected, HLA–B27 was found in the majority of AS patients (89.1%). The frequency of HLA–B27 in the MLBP group was 5.5% and, thus, in the range reported for the general population in Germany (2) (Table 3). CRP levels were elevated in significantly more patients in the AS group than in the control group (Table 3).
Table 3. Pertinent laboratory and spinal mobility findings in patients with AS and MLBP and associated positive and negative LRs*
% of AS patients (n = 101)
% of MLBP patients (n = 112)
Positive LR (95% CI)
Negative LR (95% CI)
Tests for spinal mobility (anterior and lateral spinal flexion) were performed in 97 ankylosing spondylitis (AS) patients and 103 mechanical low back pain (MLBP) patients. LR = likelihood ratio; 95% CI = 95% confidence interval; CRP = C-reactive protein.
Anterior spinal flexion <5 cm
Lateral spinal flexion <10 cm
Spinal mobility. Anterior spinal flexion (Schober's test) and lateral spinal flexion were assessed in 97 AS patients and 103 MLBP patients (Table 3). A restriction in anterior spinal flexion of <5 cm was found in 90.7% of the AS patients and in 35% of the MLBP controls (P < 0.001). Similarly, lateral spinal flexion was restricted (<10 cm) in significantly more AS patients than control patients (34.4% versus 4.9%; P < 0.001).
Independent association of clinical history parameters with IBP.
Stepwise logistic regression analysis revealed that morning stiffness of >30 minutes' duration (P < 0.001), age at onset of back pain (P = 0.008), no improvement in back pain with rest (P = 0.011), alternating buttock pain (P = 0.018), awakening because of back pain during the second half of the night only (P = 0.021), and time period (sudden/insidious) of back pain onset (P = 0.039) were independently associated with IBP.
Positive and negative LRs of individual parameters.
Positive LRs and negative LRs were calculated to assess the clinical usefulness of individual features of IBP and of combinations of features as diagnostic tests (27). As shown in Table 2, the positive LRs of features of the clinical history that were significantly different between the AS and MLBP groups ranged between 1.4 and 3.2: positive LR of 1.4 for age at onset of back pain <30 years; positive LR of 2.5 for onset of back pain within 1 year; positive LR of 2.7 for morning stiffness of >30 minutes' duration; positive LR of 2.8 for morning stiffness of >60 minutes' duration; positive LR of 2.6 for improvement in back pain with exercise but not with rest; positive LR of 3.2 for alternating buttock pain; and positive LR of 2.0 for awakening because of back pain during the second half of the night only.
None of the features had sufficient power to rule out the disease (AS). The best parameters in this regard were age at onset of back pain <40 years (negative LR 0.3) and improvement in back pain with exercise (negative LR 0.4).
Combination of parameters for the classification of IBP.
Since none of the single parameters from the questionnaire had both a high enough sensitivity and a high enough specificity (and positive LR), we selected relevant parameters based on results obtained by univariate comparisons, logistic regression analysis (see above), and clinical reasoning, and we generated numerous sets of possible combinations. The resulting sets consisted of combinations of 2, 3, 4, 5, or 6 parameters. These various sets of criteria were then compared with regard to the most appropriate tradeoff between sensitivity and specificity.
Three combinations of 4 parameters each performed almost equally well in distinguishing between AS and MLBP patients. These 3 sets of combinations were as follows (Table 4): set 8c, which contained the parameters morning stiffness of >30 minutes' duration, improvement in back pain with exercise but not with rest, alternating buttock pain, and age at onset of back pain <30 years (score test value 58.5; P < 0.001); set 7a, which contained the parameters morning stiffness of >30 minutes' duration, improvement in back pain with exercise but not with rest, awakening because of back pain during the second half of the night only, and age at onset of back pain <30 years (score test value 58.2; P < 0.001); and set 8a, which consisted of the parameters morning stiffness of >30 minutes' duration, improvement in back pain with exercise but not with rest, alternating buttock pain, and awakening because of back pain during the second half of the night only (score test value 57.1; P < 0.001). These 3 combinations of 4 parameters were therefore used to build candidate criteria.
Table 4. Various sets of combinations of 3, 4, 5, or 6 different parameters associated with inflammatory back pain and the resulting sensitivities, specificities, and positive and negative LRs in 101 ankylosing spondylitis patients and 112 mechanical low back pain patients*
No. of parameters in criteria set, criteria set number
LRs = likelihood ratios; 95% CI = 95% confidence interval.
a = morning stiffness of >30 minutes' duration; b = morning stiffness of >60 minutes' duration; c = improvement in back pain with exercise but not with rest; d = improvement in back pain with exercise; e = alternating buttock pain; f = awakening because of back pain during the second half of the night only; g = age at onset of back pain <30 years; h = onset of back pain within 1 year; i = onset of back pain within ≥1 month; j = age at onset of back pain <40 years.
The minimum number of parameters required for each set was selected on the basis of receiver operating characteristics (i.e., the best tradeoff between sensitivity and specificity for each set).
This set contains inflammatory back pain parameters as defined in the modified New York criteria of 1984 (see ref.16). All patients in this study had a duration of back pain of >3 months, and this parameter was therefore not listed separately.
This set defines inflammatory back pain according to Calin et al (see ref.6 and the Results section of the present study). All patients in this study had a duration of back pain of >3 months, and this parameter was therefore not listed separately.
As can be seen in Table 4, there were also several other criteria sets with good sensitivities and specificities (70–80%), such as sets 9b, 7b, 10a, and 8i. As expected, however, a sensitivity higher than 80% was usually accompanied by a drop in specificity and vice versa.
Performance of published criteria for IBP.
In the modified New York criteria for AS (16), IBP was defined as “low back pain and stiffness for more than 3 months, which improves with exercise but not with rest.” This definition of IBP—without further specification of morning stiffness—gave a sensitivity of 54.8% and a specificity of 66.2% (positive LR 1.6) in our cohort. When defining “stiffness” as “morning stiffness of 30 minutes or more,” the modified New York criteria definition of IBP gave a sensitivity of only 35.6%, but a high specificity of 94.6%, resulting in a positive LR of 6.6 (Table 4, criteria set 5).
In the screening test by Calin et al (6), no minimum duration of morning stiffness and no definition of “insidious onset” were given. When we defined insidious onset as the onset of back pain within a period of ≥1 month and when we defined morning stiffness as a duration of at least 30 minutes, Calin's screening test (at least 4 of 5 criteria) revealed that 64.4% of our AS patients and 23.2% of our MLBP controls met the criteria, resulting in a positive LR of 2.7 (Table 4, set 6a). If insidious onset was defined as the onset of back pain within a period of 1 year, the Calin criteria had a sensitivity of 59.4% and a specificity of 83.9% (positive LR 3.7).
Proposal of new classification criteria for IBP.
For use in daily practice, the simplicity and feasibility of criteria sets are as important as good sensitivities and specificities. In this regard, a set of criteria containing ≤4 parameters would be preferable to a set containing 6 parameters, provided that the diagnostic value is about the same. Based on a good balance between sensitivity and specificity and on feasibility, we propose that criteria set 8a, which consists of 4 parameters (morning stiffness of >30 minutes' duration, improvement in back pain with exercise but not with rest, alternating buttock pain, and awakening because of back pain during the second half of the night only), is a useful candidate set of criteria for IBP that could be applied in daily practice (Table 5). The best tradeoff between sensitivity and specificity was found if at least 2 of the 4 parameters in this set were fulfilled (Figure 1), which resulted in a sensitivity of 70.3% and specificity of 81.2% (positive LR 3.7). The superiority of candidate criteria set 8a over the Calin criteria in this cohort of patients is shown in Figure 1.
Table 5. Proposed new criteria for inflammatory back pain in young to middle-aged adults (<50 years old) with chronic back pain, and application as classification and diagnostic criteria*
The posttest probability of having ankylosing spondylitis/predominantly axial spondylarthritis is based on an assumed prevalence (pretest probability) of 5% among patients with chronic back pain (see ref.11). 95% CI = 95% confidence interval; LR = likelihood ratio.
Best tradeoff between sensitivity and specificity.
Individual parameters of the inflammatory back pain criteria
1. Morning stiffness of >30 minutes' duration
2. Improvement in back pain with exercise but not with rest
3. Awakening because of back pain during the second half of the night only
4. Alternating buttock pain
Application as classification criteria
The criteria are fulfilled if at least 2 of the 4 parameters are present†
Positive LR 3.7
Application as diagnostic criteria
If none of the 4 parameters are present
Sensitivity 10.9% (95% CI 6.2–18.5)
Specificity 57.1% (95% CI 47.9–65.9)
Positive LR 0.25 (95% CI 0.14–0.46)
Posttest probability 1.3%
If 1 of the 4 parameters is present
Sensitivity 18.8% (95% CI 12.4–27.5)
Specificity 61.6% (95% CI 52.4–70.1)
Positive LR 0.5 (95% CI 0.3–0.8)
Posttest probability 2.6%
If 2 of the 4 parameters are present
Sensitivity 36.6% (95% CI 27.9–46.4)
Specificity 83.9% (95% CI 76.0–89.6)
Positive LR 2.3 (95% CI 1.4–3.7)
Posttest probability 10.8%
If ≥3 parameters are present
Sensitivity 33.6% (95% CI 25.1–43.3)
Specificity 97.3% (95% CI 92.4–99.1)
Positive LR 12.4 (95% CI 4.0–39.7)
Posttest probability 39.4%
Set 7b, with the 3 parameters of morning stiffness of >30 minutes' duration, improvement in back pain with exercise but not with rest, and age at onset of back pain <30 years (2 of 3 positive), would also be a valuable option in daily practice because of its ease of assessment. The slightly higher sensitivity of set 7b yields a lower specificity than set 8a (Table 4), resulting in a positive LR of 2.8.
The use of criteria set 8a for diagnostic purposes is shown in Table 5. The presence of none of the 4 parameters (positive LR 0.25) or only 1 of the 4 parameters (positive LR 0.49) of criteria set 8a makes AS highly unlikely, whereas the presence of 3 or 4 of the 4 parameters (positive LR 12.4) represents a high diagnostic gain (27).
Intensity of back pain, impact on daily activities, number of days of sick leave taken, and number of doctors visited.
On the day of the interview, there was no difference in the intensity of back pain, as measured by a 0–10 rating scale, between AS patients and MLBP patients (mean ± SD score 3.62 ± 2.9 versus 3.71 ± 2.8; P not significant). Continuous (daily) pain over the previous 12 months was reported by 90.1% of AS patients, as compared with 69.9% of MLBP patients (P < 0.001). The disease had a substantial impact on activities of daily living over the previous 6 months, as reported by 43.6% of the AS patients and 37.5% of the MLBP patients. The mean ± SD number of days of sick leave taken over the previous 4 weeks was 8.5 ± 12.7 days for AS patients and 10.6 ± 12.6 days for MLBP patients. Interestingly, during their overall disease course, the AS patients visited more doctors than did the MLBP patients, with a mean ± SD of 5.3 ± 2.7 doctors visited by AS patients compared with 3.5 ± 2.4 visited by MLBP patients (P = 0.04). Visits to ≥5 different doctors were reported by 56.4% of the AS patients but only 21.8% of the MLBP controls.
The identification of patients with suspected AS among the many patients with chronic low back pain is of relevance in clinical practice (12, 20, 21). The typical features of IBP associated with AS, such as morning stiffness or improvement in back pain with exercise, have been known for some time (15), but there are only a few systematic studies that have sought to establish criteria for IBP. Furthermore, the results of those earlier studies were discordant (6, 16, 17), which may be partly related to the small numbers of either study patients (17) or controls (6). Moreover, parameters such as awakening because of back pain during the night were assessed in some studies (17, 28) but not in others (6, 29). These shortcomings motivated us to explore again the clinical history features of IBP in a large cohort of patients with AS and MLBP.
The frequencies of a variety of IBP features differed significantly between AS and MLBP patients. Of note, logistic regression analysis demonstrated the independent contribution of several of these features to the association with IBP. However, none of the parameters we tested had both a sufficiently high sensitivity and a sufficiently high specificity to be used as a single discriminating test. On the other hand, none of the single parameters or combinations of parameters of the clinical history were useful in ruling out AS. The highest sensitivity was found for improvement with exercise, since this was reported by 78.2% of AS patients. However, improvement with exercise was also reported by ∼50% of patients with MLBP, indicating a poor specificity of this item. Our creation of the conditional feature of improvement with exercise but not with rest clearly resulted in a better discrimination between patients and controls (55.4% versus 21.4%), but at the cost of sensitivity.
In the study by Calin et al (6) as well as in other studies (16, 28, 29), the duration of morning stiffness in AS was not defined. To our knowledge, Gran (17) was the only investigator who assessed the duration of morning stiffness and found that a duration of >30 minutes was relevant in AS (sensitivity 64% and specificity 58%). In our study, we also found morning stiffness of >30 minutes' duration to differentiate well between AS and MLBP patients, thereby confirming the finding by Gran (17).
Likewise, there has been no consistent definition of insidious onset of back pain. While Calin et al (6) distinguished only sudden onset from slow onset, Gran (17) differentiated between acute (within 1 week), subacute (within 1 month), and insidious (within >1 month) onset. Møller et al (28) differentiated sudden onset (minutes to hours) from onset of a few days to weeks and from insidious onset, but did not give a time period for the latter. While in the study by Møller et al, an insidious onset of back pain was reported by ∼73% of AS patients, in the study by Gran (17), insidious onset did not differentiate between AS patients and controls. Interestingly, in our study, about the same percentage of AS patients reported a very acute onset of back pain (within 1 hour; 23.7%) as reported a very slow onset (within 1 year; 24.7%). Therefore, if one considers “insidious” as representing the onset of back pain within 1 year, this indeed discriminated between AS and MLBP patients, but would apply to only one-fourth of the AS patients. If one considers “insidious” as representing the onset of back pain within a period of more than 1 week, this did not differentiate between AS patients (42.3%) and MLBP patients (33.3%) in our study.
Since none of the single parameters of IBP sufficiently distinguished AS patients from MLBP patients, we generated various sets of IBP parameters. We mainly considered the parameters that were significantly different between the 2 groups. A well-balanced tradeoff between sensitivity and specificity was found for several combinations, such as sets 9b, 7b, 8a, 10a, and 8i, all of which showed a range of sensitivity and specificity between 70% and 80%. Somewhat surprisingly, none of these candidate sets performed outstandingly better than the others. Whatever the combination of parameters tested, the sensitivity and specificity generally did not exceed 70–80% each, indicating a substantial overlap of the clinical history of AS patients and patients with MLBP. Among the candidate sets, criteria set 8a, which consisted of the 4 parameters of morning stiffness of >30 minutes' duration, improvement in back pain with exercise but not with rest, awakening because of back pain during the second half of the night only, and alternating buttock pain (at least 2 of 4 positive), evolved to us as an interesting candidate set of criteria for the classification of IBP. Set 8a also performed somewhat better than the Calin criteria, which had both a lower sensitivity and a lower specificity. As an alternative to set 8a, set 7b, which consisted of the 3 parameters of morning stiffness of >30 minutes' duration, improvement in back pain with exercise but not with rest, and age at onset of back pain <30 years (2 of 3 positive), would also be a good option because of its ease of application and its acceptable test properties, making it another interesting candidate set of criteria.
Whereas for classification purposes, a dichotomous approach (criterion fulfilled: yes/no) is generally applied, making a diagnosis in an individual patient requires more flexibility. In this regard, a multilevel approach to the application of IBP criteria may be conceivable. For example, if 3–4 parameters of set 8a were present in a patient, the positive LR would be 12.4, which represents a substantial diagnostic gain (27): assuming a 5% prevalence of AS and other spondylarthritides with predominantly axial involvement (11), the resulting posttest probability of ∼40% would make AS likely, and the primary care physician may choose to immediately refer the patient to a rheumatologist. If, in contrast, none of the 4 parameters of set 8a are present in a patient, the positive LR would be 0.25, and the resulting posttest probability of ∼1.3% would render AS very unlikely. The presence of just 2 parameters (positive LR 2.3) renders the diagnosis of AS a little bit more likely (posttest probability 11%), whereas the presence of only 1 parameter renders the disease less likely (posttest probability 2.5%).
Although the multilevel application of the criteria may seem complicated at first glance, this very much reflects clinical decision-making by experts in everyday practice (the more parameters present, the more likely the diagnosis). In fact, a multilevel application of the clinical history criteria for IBP helps to resolve the discrepancy between the consideration by many experts that the clinical history is indeed a useful diagnostic test in the individual patient and the rather poor performance of established criteria for IBP if applied at certain cutoff levels (criterion fulfilled: yes/no) that are aimed at being balanced between sensitivity and specificity.
The new candidate criteria for IBP proposed herein arose from studying a cohort of patients with definite diagnoses and established and longstanding disease who received medical care at secondary and tertiary centers. This implies several limitations, however. First, patients were selected for this study nonsystematically (“convenience sample”). Further, the examiner who took the clinical history was not effectively blinded to the diagnosis. To reduce this potential bias, a structured questionnaire was applied in the same way to every patient. Recall bias in patients with longstanding disease could also have influenced the results. Thus, the findings of this exploratory study need to be prospectively validated in another study. In such a validation study, primary care physicians might be invited to refer to rheumatologists all young adults who have chronic back pain and who are suspected of having AS, but in whom there is no clear diagnosis. AS might be suspected because of clinical manifestations that include features of IBP, a family history of AS, or the presence of HLA–B27. In all consecutively referred patients, features of IBP should be assessed and documented prior to making the diagnosis, ideally by an independent and blinded examiner (27, 30, 31).
In this study, the frequencies of HLA–B27 and of elevated levels of acute-phase reactants were also assessed, and tests for spinal mobility were performed. As expected, there were significant differences between AS and MLBP patients. Of interest, the frequency of HLA–B27 among MLBP patients was within the range of the background frequency in the general population (2), thus demonstrating the lack of association between HLA–B27 and MLBP. When the parameters of clinical history of IBP (set 8a; at least 2 of 4 parameters positive), HLA–B27 positivity, and elevated CRP level were combined, the presence of at least 2 of these 3 parameters was found in 80.1% of the AS patients but in only 6.3% of the MLBP patients. This confirms the use of these parameters, in particular, HLA–B27, as additional helpful diagnostic tests for AS, as was recently proposed (14). In established and longstanding AS, spinal mobility is frequently restricted; this was also the case in our study. Of note, in this cohort of AS patients with a mean duration of back pain of 12.9 years, the test for lateral spinal flexion was far more specific for AS than was the test for anterior spinal flexion (Schober's test), which is not consistent with the findings of a previous study (17). The diagnostic value of the tests for spinal mobility in early disease, however, is unknown.
Interestingly, during their overall course of disease, the AS patients changed doctors more often than did the patients with MLBP: the mean number of doctors visited by AS patients was 5.3, compared with 3.5 doctors visited by MLBP patients. Visits to 5 or more different doctors were reported by 56.4% of AS patients, but only by 21.8% of MLBP patients. Unfortunately, we cannot deduce from our data whether this was done mainly in search of a diagnosis or in search of relief from back pain.
In summary, in this exploratory study, we defined new candidate criteria for the clinical history of IBP in a large cohort of patients. Detailed emphasis was placed on several features of the clinical history, and numerous combinations of parameters were compared. For diagnostic purposes, a multilevel approach to the application of IBP criteria is conceivable. We have proposed a new set of IBP criteria that appears to be superior to existing sets, but this needs to be validated in a prospective study.
We thank all of the patients for their participation in the study and the following physicians, rheumatologists, orthopedic surgeons, and neurosurgeons for contributing their patients to this study: Drs. S. Allers, C. Begemann, M. Bohl-Bühler, M. Brock, P. Emberger, V. Franke, H. Haibel, R. Haux, K. Karberg, S. Kary, P. Matussek, J. Ramsbacher, A. Reeg, P. Reeg, M. Rischke, H. Sörensen, and I. Spiller.