Editorial: Splitting Headache (Off)


  • Michael D. Lockshin

    Corresponding author
    1. Weill Cornell Medical College of Cornell University and Hospital for Special Surgery, New York, New York
    • Weill Cornell Medical College of Cornell University and Hospital for Special Surgery, Barbara Volcker Center for Women and Rheumatic Diseases, 535 East 70th Street, New York, NY 10021. E-mail: lockshinm@hss.edu

    Search for more papers by this author

What is lupus headache? Does it indicate active lupus disease? These questions have for a very long time bedeviled physicians interested in systemic lupus erythematosus (SLE). In this issue of Arthritis & Rheumatism, Hanly and colleagues, using the large inception cohort of the Systemic Lupus International Collaborating Clinics (SLICC), offer some possible answers ([1]).

In their report, Hanly et al profess a humble aim, to examine the frequency and characteristics of headaches. The implications of their study, however, go beyond simply describing and counting episodes. The study suggests uncertainty as to whether lupus headache (and, by analogy, other subjective symptoms) should be defined by expert consensus rather than by direct evaluation of patients, whether a diagnosis of lupus headache is valid, how (if valid) it should be scored, and if biologic and radiologic data aid the understanding of mechanisms of headache. These larger questions also need answers.

How does one define a subjective, poorly characterized lupus symptom?

Central nervous system (CNS) symptoms are common in patients with SLE. They are usually aggregated under the rubric CNS SLE, which includes seizures, cognitive dysfunction, psychosis, demyelination, movement disorders, meningoencephalitis, and spinal cord and peripheral nerve disease. Many components of CNS SLE (psychosis, cognitive dysfunction, peripheral neuropathy) correlate poorly with conventional measures of SLE flare, such as autoantibodies, cytokine profiles, cellular and serologic immunologic and inflammatory markers, radiographic abnormalities, and clinical non-CNS organ system activity. As a result, CNS SLE studies depend heavily on clinical assessment of symptoms that themselves rely on unverifiable subjective reports for diagnosis. Headache is frequent and prominent, but some argue that it should not be counted as a manifestation of SLE ([2]).

Interested professional groups, including the American College of Rheumatology (ACR), have published consensus definitions of CNS SLE, some of which are purely qualitative ([3]) and others of which are both qualitative and quantitative ([4, 5]). To create the definitions, the consensus committees mostly used the methods of paper summaries of patients' clinical characteristics, Delphi processes, and expert opinion. None of the committees tested definitions in real-time clinical populations, nor did they provide followup or treatment response data.

The resulting definitions of lupus headache are vague. The definition describing it as a severe, persistent headache that may be migrainous, but must be nonresponsive to narcotic analgesia is the definition used in the Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2K) and the ACR Subcommittee on Neuropsychiatric Lupus Nomenclature ([4]). The British Isles Lupus Assessment Group 2004 index (BILAG 2004) defines lupus headache as a disabling headache that is unresponsive to narcotic analgesia and that lasts ≥3 days ([5]). The meaning of the italicized adjectives is left to the best guess of the clinician. Despite the vagueness, the lupus headache criterion is widely used in diagnosing and quantitating CNS disease and, with it, lupus disease activity.

Hanly and colleagues are the first to use an inception cohort to examine the headache criterion prospectively ([1]). Furthermore, they investigated clinical and serologic associations with headache, considered outcome, carefully defined subgroups of headaches (migraine, cluster, tension, intractable nonspecific, and intracranial hypertension), queried whether headache reflects lupus flare, and addressed the impact of headache on quality of life. Their followup period was 12 years, and they were able to record headache-related events in 788 of 1,732 total patients, a huge study in this field. Presenting strong and unique data, they concluded that headache does not correlate with any other manifestation of lupus flare.

A weakness of the study is that the authors did not directly examine inter- or intraobserver variability in diagnosing or characterizing headache. They did note that lupus headache, overall a rare event, was recorded at about the same rate by 50% of the study sites in 8 countries, suggesting that the SLICC investigators were consistent in diagnosis. Another weakness is that they did not have a nonlupus control population; instead, the authors cited published general population control data.

Other clinics have studied lupus headache in disease and normal controls, but without the advantage of inception cohorts, very large populations, or long followups. Katsiari et al, in a study in Greece, compared descriptive characteristics, frequency, severity, and treatment responsiveness of headaches in patients with lupus, patients with multiple sclerosis, and normal controls. Over a 1-year prospective evaluation, both patient groups had more tension-type headaches than did normal controls, but otherwise there were no differences among the groups ([6]).

Accepting the fact that the study by Hanly et al lacked simultaneous controls and did not verify observer variability, their use of an inception cohort, prospective analysis, and coordinated diagnostic criteria over more than a decade makes this study as good as can be designed to address this topic. In a broader sense, the study design is also a model for analysis of other vague, subjective, and difficult to quantify symptoms, which are rife in lupus, such as fatigue and malaise.

Is lupus headache distinguishable from nonlupus headache?

Lacking evidence of specific characteristics that would distinguish lupus headache from nonlupus headache, the authors hypothesized that lupus headache will occur at a higher than expected frequency or will be descriptively different from nonlupus headache. Neither hypothesis appears to be true, in part because ordinary headache is common both in healthy subjects and in lupus patients and because, using the restrictive (severe, persistent, nonresponsive) criteria, lupus headache is rare. I would add from my own anecdotal experience that severe, persistent, nonresponsive headaches in lupus patients are followed, usually within a short time, by seizures, magnetic resonance imaging evidence of brain edema or ischemia, or severe hypertension—in fact, the onset of new, severe, persistent headache should trigger immediate evaluation for these other findings, rather than be treated as an isolated symptom by itself. Naming headache as an independent manifestation of CNS SLE, independent of other manifestations, appears to be unnecessary. It may even be misleading. Others have said this in the past ([2]).

That there may be no such thing as a lupus headache is an important conclusion. All clinicians know that it is very tempting to treat an unexplained symptom in a lupus patient with high-dose corticosteroids or other immunosuppressive agents, on the assumption that the symptom indicates lupus activity. Although the authors did not examine treatment of headache directly (corticosteroid use, not specifically for the headache, was associated with shorter duration, but not lower frequency, of headache), their data lend support to the argument that lupus headache does not indicate active disease, and therefore should not be treated as lupus flare.

Should lupus headache be part of the SLEDAI and BILAG scores?

All therapeutic trials and nontrial clinical studies of lupus include a disease activity measure. SLEDAI-2K assigns 0 points (absent) or 1, 2, 4, or 8 points (present) for 26 different lupus-related clinical and laboratory manifestations, the number assignment being preselected for each item, i.e., either 0 or 8 points, without gradation of score (for criteria scored as 8); the maximum total score is 105. Lupus headache receives the highest assignable SLEDAI-2K score, a score of 8, while arthritis merits 4 points, and pleurisy, anti-DNA antibodies, and low complement levels merit only 2 points each. Seven of the 8 items that can earn a score of 8 in the SLEDAI-2K are neurologic signs and symptoms, compared to 4 nephritis-related items of the 6 items rated at 4 points. In the BILAG 2004 index, lupus headache is 1 of 53 items that can each be scored 0 (absent), 1 (improving), 2 (same), 3 (worse), or 4 (new/recurrence). Thus, diagnosis of lupus headache imparts a proportionately higher disease activity score in the SLEDAI-2K than it does in the BILAG 2004.

In clinical studies and trials, the high emphasis on lupus headache in the SLEDAI-2K is often not a problem, because lupus headache, restrictively diagnosed, is rare, and because CNS disease is a common exclusion criterion for trials. The high emphasis will be a major problem, however, if clinicians are not consistent in diagnosing headache. The disproportionate effect of headache in the SLEDAI-2K is a reason for discrepancy between that measure and the BILAG 2004 assessments. Theoretically, this discrepancy may account for differences in treatment trials scored by one or the other outcome measure. The high score given to headache is also a problem if the SLEDAI-2K is used in day-to-day assessment of individual patients, as is now suggested for “accountable care” clinical charts in the United States.

What is the mechanism for lupus headache?

There is no real doubt that lupus patients suffer headache. Despite the existence of numerous immunologic, neuropsychological, and neuroimaging tests, how headache occurs is unknown. Several pathogenic mechanisms for CNS SLE, in general, have been proposed, but none have shown a consistent association with either individual or combined symptoms of CNS SLE ([7-9]).

If lupus headache does indicate active lupus, what might its biologic mechanisms be? No biologic or imaging study of CNS SLE has looked at isolated lupus headache, nor, for that matter, have such studies looked at headache combined with seizure, stroke, or meningoencephalitis, and therefore currently available data offer no clues to pathogenesis. Intuitively, it seems that the search for a cause of isolated headache might focus on a vascular pathogenesis. One possible mechanism is anti–NR2 receptor antibodies (also called anti–NMDA or glutamate receptor), a CNS SLE–associated antibody that activates endothelium ([10]); however, that antibody is associated with cognitive dysfunction, not with headache.

Inability to identify a potential mechanism for headache does not preclude the possibility that headache truly reflects active disease. We are equally uninformed about mechanisms of other nonspecific findings, such as fever or leukopenia, or purely subjective symptoms like fatigue, but we do not question that these findings exist. It may be that our early twenty-first century tools are insufficient to the task of describing why or how headache occurs.

For today's front-line clinicians and their patients, the mechanism of headache and its weight in the SLEDAI-2K and BILAG 2004 are less important than is the instruction about what a physician should do when a lupus patient presents with headache. The study by Hanly et al ([1]) strongly suggests that lupus headache is not a manifestation of lupus activity and may not exist as an entity independent of other neurologic phenomena in lupus patients. In fact, because of its association with other neurologic manifestations, lupus headache is likely a tautology. Headache, when present in a lupus patient, should therefore occasion further investigation and should not be treated for itself alone.

The bias imparted by all of the neurologic scores in the SLEDAI-2K needs reconsideration. Among them, headache is an unreliable measure that contributes too many points. With regard to other neurologic symptoms, experts disagree as to whether scores can be summed or whether only the most prominent single score should be counted. In a clinically common scenario, a patient with headache might, within several days, experience a stroke and seizure, and thus would receive a score of 24 points (36 if psychosis or organic brain syndrome were also present and counted). The need to revisit the entire SLEDAI-2K neurologic symptom scoring system is beyond the scope of this commentary. At the very least, however, the headache criterion no longer seems to serve a useful purpose. It is time for this criterion to be discarded—completely split off, so to speak—from the SLEDAI-2K.


I thank Drs. Doruk Erkan and Alana Levine for their thoughtful comments.