Myasthenia gravis epidemiology in a national cohort; combining multiple disease registries
There is a wide variation in reported prevalence and incidence of myasthenia gravis (MG). In this study, we aimed to evaluate the validity of two nationwide databases by comparing prevalence and incidence rates reported from three recent studies using the two databases as case-finding method.
Materials and methods
Two different Norwegian nationwide databases were used: the acetylcholine receptor antibody database (reference cohort) and the Norwegian Prescription Database (NorPD) (study cohort). Presence of acetylcholine receptor antibodies (AChR) is specific for MG. Up to 85% of MG patients are AChR antibody-positive. All samples from the whole country were tested at one laboratory. NorPD contains patient information on all prescriptions of pyridostigmine.
Prevalence was 131 per million in the study cohort and 145 per million estimated from the reference cohort (Jan 1, 2008). No significant difference in prevalence between the study cohort and the reference cohort was found (SIR 1.1, 95% CI 1.0–1.2). The annual incidence rate was 16.0 per million in the study cohort and 8.8 per million estimated from the reference cohort, twofold more new MG patients were found in the study cohort compared to estimated figures from the reference cohort (SIR 1.8; 1.4–2.3).
This study confirms an optimal and unbiased case finding in both databases. Our calculated prevalence and incidence rates are in line with previous population-based studies. There was good agreement in prevalence reported from the two databases. The discrepancy in incidence is expected to diminish as years of study are increasing in NorPD.
Autoimmune myasthenia gravis (MG) is caused by autoantibodies, most often against the acetylcholine receptors (AChR) at the neuromuscular endplate. Impaired neuromuscular transmission leads to fluctuating and fatigable weakness of skeletal muscles. Up to 85% of MG patients are AChR antibody-positive (AChR-MG). Presence of AChR antibodies is specific for the disease. 15% of MG patients do not have detectable AChR antibodies in sera [1-4]. The group of patients without AChR antibodies is heterogeneous: 5–8% have antibodies against the muscle-specific receptor tyrosine kinase (MuSK-MG) [5, 6] and some have antibodies detected only when testing against clustered AChR antibodies . Recently, an MG subgroup with four low-density lipoprotein receptor-related protein 4 antibodies has been identified .
Data on MG prevalence and incidence vary widely: a prevalence rate of 15–179 per million  and an incidence rate of 3–30 per million . Most studies are hospital-based. Other studies combine multiple sources of data. The highest reported prevalence and incidence rates derive from the few population-based and prospective studies . MG epidemiology for the last two decades shows an increasing prevalence . Advances in therapy with a normal life expectancy, an aging population as well as improved case ascertainment are probably the main explanations [9, 10]. Nationwide registries covering whole populations are emerging, providing a sound basis for epidemiological studies [11-13]. The AChR antibody database (AAD) and the Norwegian Prescription Database (NorPD) have been utilized by our research group to elucidate the epidemiology of MG in a complete national cohort [14-16]. One single laboratory performed all AChR antibody tests for the entire country 1983–2007, providing a unique national database of all new positive AChR antibody tests in this period. NorPD contains all prescriptions expedited from every pharmacy in the country from 2004 and onwards, enabling monitoring of individual drug use. Pyridostigmine, an AChR esterase inhibitor, is the drug of choice in symptomatic treatment of most forms of MG [17-19]. Use of pyridostigmine is therefore considered a reliable indicator of MG requiring symptomatic treatment [16, 20, 21]. Three studies with reports on MG prevalence and incidence have until now been published based on these two databases [14-16].
Etiology of MG is unknown. To be able to calculate prevalence and incidence rates that reflect actual number of MG patients in a region is important for discovering etiological clues from an epidemiological point of view. The validity of the databases used in epidemiological studies is therefore crucial. The aims of this study were to evaluate the two nationwide databases, the AAD and the NorPD by comparing incidence and prevalence rates calculated from these two databases, and thereby assessing the true rates.
Materials and methods
NorPD and the study cohort
In this study, our study cohort was the patients identified through the NorPD. Since 2004, all pharmacies in Norway were obliged by law to register all prescription medication dispensed. The database therefore contains information on drug consumption for the entire population of Norway. Identification is made possible through a unique person identifier. This allows us to follow each person through the health system. Until 2008, specific diagnoses were not registered in NorPD. For MG and other chronic conditions with a confirmed diagnosis, medication is reimbursed. We have used the MG reimbursement code as a proxy for the MG diagnosis.
Information on all patients receiving at least one prescription of pyridostigmine January 1, 2004–December 31, 2007 was obtained. Of 723 patients identified, 677 (94%) met the inclusion criteria (one or more) of minimum two prescriptions of pyridostigmine during the study period, a prescription from a specialist in neurology, or prescription for MG as specified in NorPD. For each prescription, we examined the following variables: patient's age, sex, and county of residence. Drug information included the Anatomical Therapeutic Chemical code (N07AA02) and date of expedition.
The prevalence of active MG was defined as the number of patients who had received prescriptions of pyridostigmine according to the inclusion criteria and were alive and living in Norway on prevalence day as recorded in NorPD. Incidence rate of MG requiring symptomatic treatment was defined as new users of pyridostigmine with a first ever dispensed pyridostigmine prescription in the last year of the study period (2007), and who had no prescription of pyridostigmine dispensed during the preceding years of the study period (2004–2006).
AAD and the reference cohort
Two of the three studies have been conducted using the AAD as data source. The first study calculated a nationwide prevalence and incidence rate . The second study examined the geographical variation of seropositive MG in Norway, trying to identify predisposing disease factors . In those studies, the reference cohort was the patients identified in the AAD.
In 1983–1994, human striated muscle from amputated legs was used as antigen for measuring AChR antibody-specific antibodies. Binding of AChR antibodies varied according to antigen preparation [22, 23]. In 1994, the laboratory started using commercial kits  (IBL-Hamburg GmbH, Germany) using AChR from human muscle as antigen in a radioimmunoassay . The concentration of AChR antibodies was measured in nmol/L. The cutoff value of the assay was set to 0.4 nmol/l .
The following variables were registered in the database: name of patient, date of birth, date of sample acquisition, and referring doctor or hospital. For 1983–1995, the registered information was handwritten, and some parameters could be missing. From 1995, all incoming information was computerized and regarded as complete. Prevalence and incidence rates were therefore calculated for the years 1995–2008. Time of first positive AChR antibody test was defined as MG onset date.
The population of Norway increased from 4,348,410 in 1995 to 4,737,171 January 1, 2008 (the selected prevalence day in the three previous studies) (Statistics Norway - SSB, www.ssb.no). Both AAD and NorPD are based on this population. Symptom debut age of 50 years was used as a cut off to distinguish between early- and late-onset MG cases. Prevalence and incidence rates were calculated per million inhabitants.
Myasthenia gravis patients with and without AChR antibodies would receive symptomatic treatment with pyridostigmine, that is, both MG patients with and without AChR antibodies would be included in our study population. We assumed that all patients with AChR-MG are identified in the AAD. As AChR-MG constitutes 85% of all MG patients [1-4], a stipulated 15% share of MG without AChR antibodies was calculated to give estimates on MG with and without AChR antibodies.
Comparisons of age- and sex-specific prevalence and incidence rate among MG patients in our study and reference cohorts were done by calculating the standardized incidence ratios (SIRs) of observed number of patients identified in the NorPD (the study cohort) compared to the number of MG patients calculated from the numbers of patients identified in the AAD (the reference cohort). 95% confidence intervals (CIs) were calculated assuming a Poisson's distribution. All statistical analyses were performed in Microsoft Excel (2010).
At prevalence day, 619 MG patients were identified through the study cohort, while 547 MG patients were identified in the reference cohort. Estimated prevalence rates were 131 per million for the study cohort and 145 per million for the reference cohort. For the latter figure, the estimated 15% share of MG patients without detectable AChR antibodies was added. In the study cohort, mean age of onset was higher for both sexes compared to the reference cohort: 55 vs 53 years for females and 64 vs 59 years for males. The female to male ratio (F:M) in both early- and late-onset MG was higher in the study cohort compared to the reference cohort: 3.0:1 vs 2.6:1 in patients <50 years and 1.5:1 vs 1.1:1 in patients >50 years.
There was no significant difference in the estimated total number of MG patients in the study cohort compared to the reference cohort (SIR = 1.1; 95% CI 1.0–1.2) (Table 1). Also, there was no significant difference in the number of females found in the two cohorts (1.0; 0.9–1.1). A slightly higher number of males were found in the study cohort compared to the reference cohort (1.2; 1.1–1.4). In total, the number of early-onset patients did not differ significantly between the two cohorts (0.9; 0.7–1.0), but fewer female MG patients <50 years in the study cohort was found compared to the reference cohort (0.7; 0.6–0.8), while threefold more male MG patients <50 years were found in the study cohort compared to the reference cohort (3.0; 2.2–3.9). For patients >50 years, significantly more patients were found in the study cohort compared to the reference cohort (1.3; 1.2–1.4). There was no difference in the number of males identified in the two cohorts (1.0; 0.9–1.2), but more females >50 years were identified in the study cohort compared to the reference cohort (1.6; 1.4–1.8).
Table 1. Age- and sex-specific prevalence of Myasthenia Gravis (MG) patients per January 1, 2008
|Men||216||150||1.2 (1.1–1.4)||53||15||3.0 (2.2–3.9)||163||135||1.0 (0.9–1.2)|
|Women||403||329||1.0 (0.9–1.1)||160||197||0.7 (0.6–0.8)||243||132||1.6 (1.4–1.8)|
|Total||619||479||1.1 (1.0–1.2)||213||212||0.9 (0.7–1.0)||406||267||1.3 (1.2–1.4)|
For the year 2007, we found 74 new MG cases in the study cohort, yielding an incidence rate of 16.0 per million inhabitants. Estimated incidence rates in the reference cohort were calculated as mean annual rate, yielding a rate of 8.8 per million per year. The actual annual number of new patients in the reference population varied between 27 and 42 for the period 1995–2007. Age- and sex-specific incidence rates in the study cohort were 6.7 per million for patients <50 years and 34.0 per million for patients >50 years – 18.0 per million for females and 14.0 per million for males. In the reference cohort, age- and sex-specific incidence was 4.6 per million for early-onset female MG patients and 2.2 per million for early-onset male patients, while a rate of 10.0 per million was found for late-onset females and 31.0 per million for late-onset males.
In the study cohort, we found a near twofold higher number of incident MG cases in total compared to the reference cohort (1.8; 1.4–2.3) (Table 2). Nearly threefold more new female MG patients were found in the study cohort compared to the reference cohort (2.7; 2.0–3.7). No significant difference was found in the number of new male MG patients identified in the two cohorts (1.2; 0.8–1.7). The total number of new patients <50 years found was twice as high in the study cohort compared to the reference cohort (2.0; 1.2–3.0). Twice as many females were found in the study cohort compared to the reference cohort (2.3; 1.3–3.7). No significant difference in the number of males <50 years was found between the two cohorts (1.4; 0.5–3.3). For patients >50 years, nearly twofold more new patients were found in the study cohort compared to the reference cohort (1.7; 1.3–2.3). Threefold more new female MG patients >50 years were found in the study cohort compared to the reference cohort (3.2; 2.1–4.6), but the number of new male MG patients was not significantly different in the two cohorts (1.2; 0.8–1.8).
Table 2. Age- and sex-specific incidence of Myasthenia Gravis (MG) patients, 2007
|Men||32||22||1.2 (0.8–1.7)||5||3||1.4 (0.5–3.3)||27||19||1.2 (0.8–1.8)|
|Women||42||13||2.7 (2.0–3.7)||16||6||2.3 (1.3–3.7)||26||7||3.2 (2.1–4.6)|
|Total||74||35||1.8 (1.4–2.3)||21||9||2.0 (1.2–3.0)||53||26||1.7 (1.3–2.3)|
There was no difference in the total prevalence of MG patients comparing the results calculated based on AAD and NorPD. This indicates that most MG patients have been identified by these two databases. The NorPD and AAD do not contain specific MG diagnosis, but include very specific indirect markers. Our calculated MG prevalences of 131 and 145 per million in the NorPD and the AAD, respectively, are in good agreement with other recent population-based studies on MG prevalence of 117–179 per million [9, 20, 27]. The heterogeneity in previously reported MG prevalence ranging from 15 to 179 per million is probably due to differences in case ascertainment and inclusion criteria . Use of commercial AChR antibody assays and improved epidemiological methodology are main explanations for the trend of increasing prevalence with year of study . This trend is also seen in our data. As expected, higher prevalence was found in the AAD which covered a period of 25 years compared to 4 years in the NorPD. A longer time span minimizes the possibility of missing patients in remission. Nonetheless, this difference was not significant.
Our data revealed a threefold higher number of prevalent male MG <50 years and a doubling in the number of prevalent female MG >50 years in the NorPD compared to the AAD. The reason for this is unknown. AChR antibody tests are likely to be taken of all patients with symptoms of MG regardless of age or gender. Age or gender was not known for 68 of the patients in the AAD. There is a theoretic possibility that the observed discrepancy may be caused by a larger proportion of the unknown cohort belonging to either of these two subgroups of MG patients. Epidemiology of other subgroups of MG than AChR-MG and MuSK-MG remains undescribed . Differences in age and gender ratios could be due to a different distribution of age and gender among MG patients without AChR antibodies being missed in the AAD, but registered in the NorPD. Only 50% of ocular cases have AChR antibodies . Progression to generalized disease is seen in 80% after 2 years, but 20% remain purely ocular MG . Some patients with milder symptoms might not have been tested for AChR antibodies and/or received pyridostigmine medication without a positive antibody test.
The incidence rates of 16.0 and 8.8 per million calculated in the NorPD and the AAD, respectively, are in agreement with other prospectively conducted studies, ranging from 11 to 30 per million . As for prevalence, reports of MG incidence are heterogeneous, ranging from 1.7 to 21.3 per million for population-based studies [9, 10]. This wide range reflects the methodological challenges due to small study populations and data collection from multiple data sources, probably also real geographical variation. The trend has been toward a decreasing heterogeneity with enhanced study quality . As incidence could be calculated for 1 year only in the NorPD, we were not able to assess time trends.
We found that the total incidence rate in the NorPD was twice as high as in the AAD. Discrepancies in incidence rates were observed for both early- and late-onset MG. Inclusion of some relapsing MG cases in the NorPD is the most likely explanation. The issue of relapsing cases was the reason why we for the NorPD data calculated the incidence for the last year of the 4-year period only. We expect this issue to resolve with increasing time span in the NorPD. MG remission is more often associated with early-onset MG [30, 31], whereas there is no association between remission and gender . Age of MG debut differs between males and females. Male disease onset peaks around 60–70 years, while female disease onset is bimodal with one peak around 20–30 years and one peak around 60–70 years [14, 16, 33, 34]. Females who terminate drug treatment during pregnancy may have caused bias in estimates of MG incidence for early-onset females in the NorPD. However, symptomatic MG treatment with pyridostigmine is rarely discontinued during pregnancy. One study refers that in 66 of 135 pregnancies, pyridostigmine was used and continued throughout the pregnancies in 61 of the cases . Females are twice as likely as males to have an AChR antibody test performed . It is therefore unlikely that more females than males with AChR-MG would have been missed in the AAD.
In this study, two nationwide databases used to assess MG epidemiology within a well-defined national cohort were evaluated. The study confirms an optimal and unbiased case finding in both databases. Our calculated prevalence and incidence rates are in line with previous population-based studies. There was good agreement in prevalence reported from the two databases. The discrepancy in incidence can be explained by the different case-finding methods, and is expected to diminish as years of study are increasing in the NorPD. Pyridostigmine is not prescribed on a regular basis for other diseases than MG, and the drug is only reimbursed for MG. A positive AChR antibody test is sensitive and 100% specific for AChR-MG. The specificity of pyridostigmine prescriptions and a positive AChR antibody test for the MG diagnosis makes these two databases suitable for epidemiological studies, providing valid calculations of MG prevalence and incidence.
The authors have no acknowledgments to declare.
Conflict of interests
The authors declare no potential conflicts.