Methods

Study population

We used data from 96 patients (46 from FWI and 50 from Cuba). In the FWI, a prospective population based archive of research medical records from patients with NMO has been in place since 1992. In Cuba, 214 neurologists were contacted in order to identify suspected patients with NMO, to report all suspected cases with NMO in each province and to complete a case report form. We included patients with autopsy proven NMO, patients fulfilling the 1999 NMO criteria1 (non NMO-IgG tested) and patients fulfilling the 2006 NMO criteria4 (NMO-IgG tested). NMO-IgG testing5 blinded to all clinical information allowed us to calculate its sensitivity and specificity by using sera from 50 RNMO patients and 137 controls (longitudinal extensive transverse myelitis n = 17; recurrent ON n = 12; severe ON n = 7; multiple sclerosis n = 52; HTLV-1 associated myelopathy (HAM)/tropical spastic paraparesis (TSP) n = 8; asymptomatic HTLV-1 carriers n = 9; and non-demyelinating central nervous disorders n = 20). Between 31 December 1992 and 31 December 2007, no monophasic NMO was seen in FWI. No RNMO patient exhibited a secondary progressive course.

Assessment of patients

Irreversible neurological disability was assessed every 6 months using the Kurtzke Disability Status Scale (DSS). We defined the optic morbidity index as the number of optic bouts leading to fixed visual acuity lower than 20/200. Time between onset of RNMO and the occurrence of unilateral and bilateral severe visual loss was noted. Spinal morbidity index was defined as the number of spinal bouts leading to fixed attack related weakness (Medical Research Council ⩽2) in one or more limbs. We assessed the effect of the following variables on time to death: sex, race, area of residence, age and symptoms at onset, whether the patient had recovered from the first attack (DSS <3, 6 months after onset) and first interval attack. We also included variables concerning treatment by intravenous immunosuppressive drugs which were available for FWI patients. Effect of oral immunosuppressors was not evaluated because only a few patients were treated with azathioprine in Cuba. Cause of death was assessed by death certificates.

Statistics

Data were recorded using StatView 5.0 (SAS Institute, Cary, North Carolina, USA). Values are reported as median (range). Survival was analysed according to the Kaplan–Meier method. The end point was the time to irreversible disability or death, as indicated by scores of 3, 6, 8 and 10 on the DSS. Multivariate regression analysis was performed to identify independent risk factors for mortality. Two sided p values <0.05 were considered to denote statistical significance.

Results

Characteristics of the patients

Median age at onset for the 96 patients (88 women/eight men) was 29.5 years (11–74) (see table in supplementary data available online). NMO-IgG testing5 was found to have a sensitivity of 32% and specificity of 97% in our cohort (table 1). Seven patients had RNMO associated with another autoimmune disorder (systemic lupus erythematosus n = 3; type 1 diabetes mellitus n = 2; sicca syndrome n = 1; and thyroiditis n = 1). Median follow-up was 9.5 years (1–40) (median prospective follow-up 8.5 years). Median time of first inter attack interval was 11.5 months (1–300). The annual relapse rate ranged from 0.1 to 3.6 (median 0.7). There was a significant decrease in relapse rate between the first 5 years of disease and the following fifth ones for optic attacks (p<0.001) and mixed attacks (p = 0.004) (see fig 1 in supplementary data available online). Twenty-seven patients (median age at treatment 38 years) received either alone or in combination mitoxantrone (n = 24), cyclophosphamide (n = 7) and rituximab (n = 9) once diagnosis of RNMO was established.

RNMO course

Sixty-six patients (68.8%) experienced severe visual loss in at least one eye and 46 (47.9%) in both eyes. Median optic morbidity index was 1 (1–4). Median time from onset to unilateral and bilateral severe visual loss was 3 and 15 years, respectively (see fig 2 in supplementary data available online). Median times to reach DSS scores of 3, 6 and 8 from onset of RNMO were 1, 8 and 22 years (see fig 3 in supplementary data available online). Median spinal morbidity index was 3 (1–14). Twenty-four (25%) patients died at a median age of 34.5 years after a disease duration of 11 years. Fifteen (63%) died within 5 years. Respiratory failure during spinal bout occurred in 18 patients but hypothalamic involvement was the cause of death in two patients (severe hypothermia and inappropriate secretion of antidiuretic hormone). Autopsy proven cases had a shorter disease duration (p = 0.005), a higher annual relapse rate (p = 0.005) and a higher attack related severe weakness (p = 0.01) compared with NMO-IgG tested patients (table 1). Half of NMO-IgG positive patients had experienced severe myelitis.

Univariate survival analysis (log rank test)

Time to death was not significantly different between patients from Cuba and FWI (p = 0.50). Sex, race, age at onset and other initial signs of RNMO (motor, sensory, sphincter) did not affect time to death. Time from onset to death was shorter in the group with blindness at the first attack (p = 0.043) and in patients who did not recover from the first attack (p = 0.004). A first inter-attack interval shorter than 1 year was strongly associated with death (p<0.001). In the subgroup of patients who were given immunosuppressive agents, time to death was not significantly shortened (p = 0.19).

Multivariate survival regression analysis

All clinically important prognostic factors were entered into the Cox proportional hazards model. The agreement between this regression model and the data was verified using graphs of cumulative hazard plots. Hazard ratios were generated and expressed with 95% confidence intervals (table 2). This analysis demonstrated that a higher attack frequency during the first year of disease, blindness, sphincter signs at onset and lack of recovery after a first attack were independently associated with a shorter time to death. The regression model did not fit the distribution of patients assigned to immunosuppressive agents. However, the introduction of treatment as a forced variable into this model demonstrated that the clinical predictors of severity were still associated with a shorter time to death whatever the treatment group.

Discussion

Our large series showed, as in other studies,6 7 8 9 10 that RNMO affects almost exclusively women, age at onset is before 40 years of age, relapse rates are high and the second episode usually occurs within 2 years. The 32% value for NMO-IgG seropositivity we found is low compared with other series.4 11 NMO-IgG seropositivity is thought to be correlated with a high relapse rate and severe disease course.12 We can therefore speculate that our NMO-IgG testing in alive RNMO patients favoured results towards a lower sensitivity as our non- tested autopsy proven cases exhibited an aggressive course. Alternatively, one could suggest that intense immunosuppression lowers the level of NMO-IgG. However, the rate of seropositivity was similar between patients from FWI and Cuba, and only 59% of patients from FWI were immunosuppressed at the time of testing. Sixty-eight per cent of our patients became blind. With a shorter follow-up than ours, the rate of blindness was 60% in the Mayo Clinic study.1 Most important is the fact that less than two attacks are sufficient to cause a definitive loss of vision in a given eye in the course of RNMO. The rapidity of RNMO disability is apparent, with half of patients reaching DSS 3, 6 and 8 by 1, 8 and 22 years, respectively. In 46 RNMO patients recruited from15 Italian centres,2 DSS 3 was reached after a median time of 6 months and DSS 6 after 7 years. Our longitudinal data agree with this rate of disability accumulation in RNMO. Late disability in RNMO (DSS 6 and 8) is linked to sequelae of spinal attacks and we found the spinal morbidity index to be as low as 3. Plasmapheresis could be a possible option in preventing loss of motor function during RNMO exacerbations of myelitis.13 A high proportion died during a spinal bout from respiratory failure. This complication of RNMO was found to be the only cause of death in the Mayo Clinic series.1 Our study showed that death may be due to central endocrine disturbances. In the Mayo Clinic series involving 57 patients,1 the authors found that a history of autoimmune disease, a higher attack frequency during the first 2 years of the disease and better motor recovery following the index myelitis event were associated with mortality due to RNMO. In contrast, our study has shown that lack of recovery after the first attack is associated with early death (p<0.01). We also found that blindness (p = 0.04) and sphincter signs (p = 0.02) at onset, and an early attack frequency, but during the first year of disease, was a strong independent predictor of death (p<0.01). Some short cases series have reported that patients with RNMO could be stabilised with azathioprime,14 rituximab,15 mitoxantrone16 or cyclophosphamide,17 but neither these studies nor ours evaluated the effect of immunosuppression after the first attack. It has been reported recently that 62.5% of NMO-IgG positive patients who had experienced only a single attack of longitudinally extensive idiopathic myelitis relapsed with either myelitis or ON within 12 months.18 Our multivariate model of a second attack interval shorter than 1 year is useful in predicting early death and therefore in influencing choice of treatment. In view of these findings, firstline immunosuppressors might be warranted in NMO-IgG positive patients with high risk syndromes in the spectrum of NMO19 in order to delay the second attack of this devastating disorder.