Epidemiology of NMOSD in Sweden from 1987 to 2013

Objective To report the yearly incidence rate and prevalence of neuromyelitis spectrum disorder (NMOSD) in Sweden and to investigate clinical characteristics, treatment, and outcome. Methods We conducted a retrospective study of hospital case records of 294 individuals diagnosed with neuromyelitis optica (NMO) (G36.0 ICD-10, 341.0 ICD-9) in the Swedish National Patient Register from 1987 to end of 2013 or detected by the presence of aquaporin-4 (AQP4) immunoglobulin G (IgG) in serum during the study period. Ninety-two patients (51 NMO and 41 NMOSD) met the 2006 Wingerchuk criteria and were included in the study. Ten patients with an onset of NMO prior to 1987 and alive at the end of 2013 were included when estimating the prevalence. Results The average yearly incidence rate per 1,000,000 individuals increased significantly from 0.30 (confidence interval [CI] 0.19–0.41) between 1987 and 2006 to 0.79 (CI 0.55–1.03) between 2007 and 2013. The prevalence was 10.4 (CI 8.5–12.6) per 1,000,000 individuals at end of 2013. The median time from onset to first relapse was 1.42 years (range 0.58–3.90). The probability of relapse was 60% and 75% after 5 and 10 years after onset. More than 80% were treated with immunosuppressive drugs. Three patients died during the study period. Conclusion The increased incidence rate during the study period was likely due to heightened awareness and increased access to MRI and AQP4-IgG analysis. Incidence and prevalence of NMO in Sweden correspond to other countries with a predominately Caucasian population. We found that most patients were treated with immunosuppressant drugs, presumably resulting in low mortality among the detected cases.

The incidence of NMO per 1,000,000 person-years has varied substantially, from 0.4 in Catalonia to 0.8 in the United Kingdom and central Denmark to 4 in southern Denmark. 1,9,12,22 Although a number of epidemiologic studies on NMOSD have been conducted over recent years, many have methodologic limitations, such as few individuals in each study, limited geographical areas where representativeness is uncertain, short follow-up time, and age limitations. 9,19,22,24,[33][34][35] Furthermore, some studies have only attained cases through positive AQP4-IgG analysis and have not had full access to all patient records. We therefore carried out a nationwide, population-based, cohort study of the incidence and prevalence of NMO and NMOSD in all ages over a long period. In addition, we assessed clinical characteristics, treatment, and outcome.

Data collection
The National Swedish Patient Register (NSPR) lists all patients hospitalized (with total national coverage from 1987) or managed in hospital-based ambulatory care since 2001 in Sweden. 36 In Sweden, the vast majority of neurologic patients are followed up and treated through the hospital system. Few patients with MS are treated only by private neurologists and we know of no such patient with NMO. Each individual's outpatient visit or hospital discharge diagnosis (ICD code) is linked to a unique personal identification number. These data were collected from individual journals of all individuals registered with a diagnosis of neuromyelitis optica (ICD-10 code G36.0 or ICD-9 code 341A) in the database of the Swedish National Board of Health and Welfare from 1987 through 2013. During the study period, analysis of NMO-IgG in Sweden was performed only at Wieslab Laboratories, Malmö, Sweden. The laboratory used the immunoblot assay until 2012; since then, a cell-based analysis has been used. From the laboratory, we received a list of all individuals, their titers, and resident hospital, who tested positive during the study period.
This study used the 2006 Wingerchuk criteria, which are the generally accepted criteria during the latter half of the study period and time of data collection. The criteria require ON and myelitis along with 2 of the following 3 supportive criteria: (1) LETM (≥3 vertebral segments in length), (2) brain MRI with normal findings or with findings not consistent with MS, and (3) AQP4-IgG seropositivity. NMOSD was defined as AQP4-IgG-seropositive status with a clinical event of LETM or ON. 11 After receiving information from both the NSPR and the results from the only laboratory with accredited methods for AQP4 serologic determinations in Sweden, we requested patients' medical records from all clinics and hospitals. Epidemiologic data-including demographic, clinical, CSF (cell count, protein levels, and oligoclonal bands), MRI findings, treatment, and outcome-were obtained from medical records through a structured checklist by a neurology resident. No information could be retrieved for 15 patients, and 17 more records had obvious clerical errors (figure 1). Information from 262 individuals was evaluated against the 2006 Wingerchuk criteria by 2 neurologists. After scrutiny of the records, it was clear that some patients were more likely to have other neurologic disorders such as MS, acute disseminated encephalomyelitis, or infections. These were excluded, leaving 145 patients. Antibody-negative patients who had lone ON, lone transverse myelitis (TM), or ON and short TM were excluded. Following this further exclusion, 92 patients remained and were included into the study cohort for estimation of the yearly incidence rate. Fifty-one patients had Glossary AQP4 = aquaporin-4; CI = confidence interval; ICD = International Classification of Diseases; IgG = immunoglobulin G; LETM = longitudinally extensive transverse myelitis; MS = multiple sclerosis; NMO = neuromyelitis optica; NMOSD = neuromyelitis optica spectrum disorder; NSPR = National Swedish Patient Register; ON = optic neuritis; SIR = standardized incidence ratio; TM = transverse myelitis. classical NMO, while the remaining 41 fulfilled NMOSD criteria (figure 1). Ten patients had an onset of NMO prior to 1987. These were considered when estimating the prevalence. The ratio between NMOSD and MS in Sweden was calculated by using the last prevalence study of MS in Sweden 37 and the number of patients with NMO alive at the end of our study period. This study was approved by the local ethical committee in Stockholm, Sweden (No. 2010/1924-31/3).

Statistical analysis
Population data for estimation of incidence were collected from Statistics Sweden (scb.se) and United Nations (population.un. org/wpp/DataQuery/). The prevalence calculated for the date of December 31, 2013, involved the number of prevalent NMO/ NMOSD cases per 1,000,000 inhabitants. A 95% confidence interval (CI) for the prevalence was estimated using methods described by Vollset. 38 The average yearly incidence rate was defined as the number of new NMO/NMOSD cases (calculated using the date of symptom onset) for the period from January 1, 1987, to December 31, 2013, divided by the total number of people-years at risk (estimated by summing the mid-year census population of each year). This was reported per 1,000,000 people-years. Standardized incidence ratio (SIR) was estimated by adjusting for sex and age distribution and presented with a 95% CI estimated using Byar approximation. Zero estimates were taken into account when estimating SIR for time periods. Data were presented with mean and 95% CI or median with range (25th and 75th percentiles) when appropriate. The distribution of data was presented using histograms with the appropriate number of bins. Trends over time were estimated using linear regression models and reported with CIs when appropriate. The probability to first relapse was evaluated using the Kaplan-Meier method. Right censoring was implemented, and the last day of follow-up was set as December 31, 2013, as long as the patient had not relapsed. The log-rank test was utilized for comparison of relapse probability between strata. p Values <0.05 were considered significant. Data and statistical analysis and data presentation were prepared in R (version 3.4.4) 39 using packages data.table 40 and ggplot2. 41 Data availability Anonymized data of this study will be available from the corresponding author on reasonable request from any qualified investigator, following the EU General Data Protection Regulation.

Prevalence and incidence
The prevalence of NMO/NMOSD in Sweden as of 2013 was 9.3 (CI 7.6-11.5) individuals per 1,000,000 individuals. After standardization to the world population by age, the prevalence was 1.03 (CI 0.66-1.55) for males and 0.99 (CI 0.77-1.26) for females. The ratio between NMO/NMOSD and MS in Sweden based on prevalence was calculated to be 100 vs 17,485 (as of 2008) 37 or a ratio of 1 NMO patient for every 175 MS patients (0.6%). The yearly crude incidence rate showed an increasing trend over time from 1987 to 2013 (figure 2). Age-standardized estimates of yearly incidence rate for each sex are available in  3A). Ten individuals (9.7%) were non-Caucasian. Autoimmune comorbidities were detected in 25.0%. Among the NMOSD group, 36.6% had ON and 63.4% LETM, respectively. Bulbar symptoms were found in 15.2% of all patients. No difference could be detected between patients with NMO and NMOSD in terms of findings of elevated leukocytes in CSF or the presence of oligoclonal bands. The most common drug given was corticosteroid treatment (87.0% received steroids at least once). Azathioprine and rituximab were commonly used as disease-modifying drugs (44.6% and 38.0%, respectively). Four (4.3%) patients underwent autologous hematopoietic stem cell transplantation, mainly due to previous treatment failure. The median follow-up time was 6.2 years (range 3.9-11.7) (figure 3B). Two patients with a disease onset after 1987 and one with an onset prior to 1987 died during the study period (3.3%).

Discussion
In this study, we sought to determine the average yearly incidence rate and the current prevalence of NMO and NMOSD in Sweden. The average yearly incidence rate of NMO/NMOSD was 0.79 (CI 0.55-1.03), and the prevalence was 10.4 patients per million individuals as of the end of 2013. The incidence rate increased steadily from 1987 to 2013, most likely because of increased vigilance but also due to greater usage of diagnostic methods such as MRI (;1990 onwards) and analysis of AQP4-IgG serostatus (2007 onwards).  Although the study is nationwide and population-based, the estimated incidence and prevalence are to be considered minimum estimates. It is known that estimates of incidence and prevalence of rare diseases using registry data are likely to be lower than their true equivalents. 42 Furthermore, we know from clinical experience that patients with NMOSD who do not present with extensive myelitis are at risk of being incorrectly diagnosed with MS. A patient with extensive myelitis is more likely to be tested for AQP4 serostatus and correctly identified as NMOSD than an individual with shorter myelitis or an isolated ON. On the other hand, the high accessibility of publicly funded health care in Sweden in combination with the debilitating symptoms of NMOSD make it unlikely that patients remain unnoticed or receive treatment at private institutions.
The average incidence rate in Sweden did not differ from the incidence rate in central Denmark, Merseyside (UK), or Cuba, was slightly higher than the rate in Catalonia, but was much lower than the rate in southern Denmark ( figure 5A). The prevalence in Sweden was estimated at 10.37 patients with NMO per million individuals, which reflects the prevalence in South East Wales (UK), Merseyside (UK), and Mexico, but is higher than the prevalence in Catalonia and Cuba and lower than the prevalence in southern Denmark ( figure 5B). 1,7,9,12,19,22,24,[33][34][35][43][44][45][46][47][48][49] Thus, the incidence and prevalence of NMO/NMOSD in Sweden is comparable to data reported from countries with a predominately Caucasian population. Approximately 90% of the patients included in this study were of Caucasian origin, which is similar to the Swedish population. Our data, however, do not correspond to Southern Denmark 1 or the United States (Olmsted County) 43 (figure 5). In these 2 studies, all patients who presented with demyelinating disease 42 or an episode of ON or TM and an initial brain MRI not diagnostic for MS 1 were screened for AQP4-IgG and were probably diagnosed earlier and more accurately. Because the demographics are similar in Sweden and Denmark, the absence of such a strategy in Sweden suggests that several cases of NMOSD were not detected. This underpins the importance of screening patients with a demyelinating event for AQP4-IgG serostatus.  In this study, autoimmune comorbidities were detected in 25.0%, which is comparable to another recent populationbased study from Catalonia (27%). 12 We also had very similar female to male ratio (3/1) and average age at onset (41.5 years) as the Catalan study. Other studies have described a higher female to male ratio (up to 9/1 in Martinique) 43 and younger average age at onset, 9,23,43 down to 31 years in Iran. 7 We could not detect any clinical difference between patients with NMO and NMOSD regarding imaging findings, level of leukocytes in CSF, or presence of oligoclonal bands. This supports the increased evidence that NMO and NMOSD follow a similar clinical course and should be regarded as the same disease. To our surprise, the patients with NMO were, on average, 10 years younger than the patients with NMOSD.
We have no explanation for this finding. Another important finding is that the type of symptoms at onset did not affect the probability of a new relapse. To our knowledge, this has not been reported before. As expected, the majority of patients with NMO/NMOSD did relapse, and the probability of relapse measured using the Kaplan-Meier estimator was not different between NMO and NMOSD. In a recent study of pretreatment annualized relapse rate, the relapse rate was 2.26-2.29. 50 In our cohort, the risk for a first relapse was highest during the first 2 years (median 1.42 years). Also, the risk of relapses never disappeared during the study period. Considering the continual risk and severity of NMO relapses, this is in agreement with the prevailing opinion that immunosuppressive drugs should be administered and not discontinued.
To allow for early treatment of NMO/NMOSD and to reduce the risk of disability and death, correct and early diagnosis of NMO/NMOSD and distinction from MS is of utmost importance. In earlier reports, more than 50% of patients with NMO/ NMOSD were blind in one or both eyes or required ambulatory help within 5 years of disease onset. 21 In 1999, before the spectrum of NMO was understood, the calculated 5-year survival rate for NMO was 68%, with all deaths caused by neurogenic respiratory failure. 10 The low mortality (3.3%) observed in our sample might be explained by a high use of immune-modulating drugs. Furthermore, since this was a nationwide study, it is possible that patients with a mild disease course not needing tertiary hospital care were included, thus contributing to the low mortality.
The main limitations of our study are the retrospective design and the dependency on clinicians' diagnosis or suspicion of NMO/NMOSD, which most likely contributes to underestimating the incidence and prevalence of NMO/NMOSD. Another limitation is the fact that we did not have access to records from outpatient care that was not hospital-based. Some milder cases of NMOSD may have been missed. On the other hand, the health system in Sweden is built up in such a way that fewer than 5% of neurologic outpatients are seen by private practicing neurologists. The rule is that all MS/NMO patients should be seen through the hospital system. Furthermore, since the experience of demyelinating diseases varied between clinics and systematic clinical evaluation, we were unable to retrieve and evaluate the Expanded Disability Status Scale score for the majority of patients.
Even though the incidence and prevalence of NMO/ NMOSD in Sweden are in line with most reports from countries with a predominately Caucasian population, we still suspect that NMOSD in Sweden is underdiagnosed, which could be compensated by increased vigilance and routine AQP4-IgG antibody analysis. More than 80% of patients were treated with disease-inhibiting immunosuppressive drugs (mostly azathioprine and rituximab), which possibly affected the low mortality rate among detected cases. We found a high risk of relapse within a fairly short time; considering the severity of relapses, this should be kept in mind when counselling a patient with NMOSD. All patients should receive continuous disease-inhibiting immunosuppressive treatment. An initiative for systematic prospective national follow-up has recently been initiated through the Swedish registry for neurologic diseases. It is our hope that it will reach full national coverage within a limited time frame.

Study funding
The study was supported by The Swedish Brain Foundation, Karolinska Institutet, and Stockholm City Council.
Disclosure D. Jonsson reports no disclosures relevant to the manuscript. O. Sveinsson has received grants from GSK and Biogen as well as personal fees from Biogen and UCB, all outside the submitted work. R. Hakim reports no disclosures relevant to the manuscript. L. Brundin has received travel grants, participated in advisory boards, and received lecturing honoraria from Biogen, Sanofi, Novartis, Amirall, and Genzyme. Go to Neurology.org/N for full disclosures.

Publication history
Received by Neurology September 2, 2018. Accepted in final form February 25, 2019.