Making a diagnosis of ankylosing spondylitis in patients with chronic back pain can be difficult at an early stage—that is, before radiographic sacroiliitis is definitely present (also referred to as axial spondyloarthritis (SpA) at the preradiographic state). We recently proposed to diagnose patients at this early stage by probability estimations¹ based on a pretest probability (p_pre) of 5% in patients with chronic back pain.² To facilitate the probability calculation in each patient, we subsequently³ proposed the use of likelihood ratios (LR).⁴ We suggested that the diagnosis could be considered definite if the post-test probability (p_post) is ⩾90% (LR product ⩾171), probable if the post-test probability is 80–90% (LR product 76–171) and unlikely if the post-test probability is ⩽10–20% (LR product <2–4).^1,3

Mainly because of the complicated mathematics, we previously³ concentrated on the use of positive likelihood ratios—that is, in case the parameter is present. However, when making a diagnosis in daily practice, a negative test result (absence of a certain parameter) sometimes helps to rule out a diagnosis. In axial SpA, a few parameters, if absent, clearly render the diagnosis less likely. These include negativity for human leucocyte antigen-B27, a negative magnetic resonance image (showing no signs of inflammation), the absence of the inflammatory type of back pain, a normal C reactive protein level or erythrocyte sedimentation rate, no good response to non-steroidal anti-inflammatory drugs and, probably, a negative family history (discussed already by Rudwaleit et al¹). On the other hand, other mostly clinical parameters should not be considered to be definitely absent if not present at disease onset, as these may occur later in the disease course and therefore are rather a function of disease duration. These include peripheral arthritis, enthesitis, dactylitis, acute anterior uveitis, psoriasis and inflammatory bowel disease. These parameters are helpful in increasing the disease probability if present, but should be ignored if absent at an early disease stage.

Table 1 shows the list of LR+ values for positive test results supplemented by LR− values for negative test results. The likelihood ratio product is calculated by multiplying the relevant LR+ and LR− values as derived from table 1, according to the presence or absence of particular features as appropriate. The final post-test probability can be read from fig 1, which presents a probability curve showing the dependency of the post-test probability on the LR product, again based on a pretest probability of 5%. The curve in fig 1 has been calculated using the formula

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

 Dependency of the post-test probability of axial spondyloarthritis (SpA) on the resulting likelihood ratio (LR) product for an assumed pretest probability of 5% (according to Underwood and Dawes²). This probability curve is meant to be applied in patients with chronic back pain suspected to have axial SpA.

where p_post is the post-test probability, Π_LR the product of likelihood ratios and p_pre the pretest probability.

Thus, taking into account all positive and negative diagnostic test results as appropriate, the disease probability of axial SpA at the preradiographic stage in a patient with chronic back pain can now be easily assessed at the bedside with the help of table 1 and fig 1.