Study design and data source

This pilot, retrospective, population‑based cohort study was conducted using administrative health care claims from the NHF. Additionally, cross‑sectional assessment of MS burden in the Polish population was performed. Currently, there is no dedicated national registry for patients with RA or MS in Poland. The NHF is the only public health care payer, with coverage of various medical services across numerous hospitals, outpatient clinics, and medical centers holding a reimbursement contract. The NHF payment depends on patient employment and citizenship status, with minor exceptions. The role of the private sector in Poland is limited, in that it is highly unlikely for individuals to hold no record of public health care contact, particularly in the context of chronic diseases.^18-20 To maintain consistency in the timeframe of exposure, unique patient identifiers were cross‑referenced with the death register maintained by the Ministry of Digitital Affairs of Poland. Due to the administrative nature of the data (personal identification number [PESEL] linkage for all citizens), incomplete or inaccurate (but not missing) data represent the major source of bias.

This report covers all persons over 16 years old that could be identified using PESEL codes within the NHF database from January 1, 2009, to December 31, 2021. As the observation began in 2009, left truncation limited assessment of the disease onset before electronic database inception. Therefore, our analyses were reported with a look‑back (so‑called “wash‑in”) period between 2009 and 2010, with analyses of aggregated data from January 1, 2011. Patients were progressively identified throughout the study timeframe, with cohort size changing over time. The maximum follow‑up window was 13 years.

We report 3 main analyses based on annual cross‑sectional datasets: 1) changes in annual period prevalence of MS among RA patients and general population–based controls (2011–2021); 2) changes in the number of newly detected MS cases (annual incidence) in both groups (2011–2021); and 3) cumulative prevalence comparison of MS between the RA and matched control groups, stratified by sex, age, and RA serotype.

For the two temporal analyses (prevalence and incidence), annual estimates were calculated using the number of individuals registered as alive and insured in the NHF at mid‑year as the denominator for each study group. For prevalence, the numerator was the number of individuals fulfilling the MS case definition in a given calendar year. For incidence, the numerator was the number of individuals meeting the MS case definition for the first time in that calendar year.

Of note, our study describes living individuals who met the case definitions at a given follow‑up time. Differences in survival may lead to competing risk and surveillance bias. Therefore, we analyzed mortality rates in both RA and control groups. To some extent (likely limited), postdiagnosis survival rates do not capture undetected cases (eg, due to real‑life diagnostic errors, or limitations of the chosen case definition) and premature deaths. Pragmatically, the prevalence estimates in older patient groups are most likely to be affected by surveillance bias, though given the cumulative approach to case identification, the distribution of bias should be (relatively) uniform. Nevertheless, our estimates should be treated conservatively.

Case identification

The patients with MS were identified based on NHF claims if, within any 3 consecutive years, they had at least 3 health care encounters coded with the International Classification of Diseases, Tenth Revision (ICD‑10) G35 code (ie, outpatient specialist visits, inpatient admissions, or use of rehabilitation services). However, at least 1 of these encounters was required to occur in a neurology setting. Participation in a dedicated MS drug program was treated as equivalent to fulfilling the latter criterion. It also represents 1 of the 3 required health care contacts, rather than an additional requirement. Disease onset was ascertained based on the date of the earliest fulfillment of the case definition criteria.

Due to the secondary nature of claim data, a combination of different definition elements is an approach that is commonly used to increase the reliability of proxy disease diagnoses.²¹ Our definition follows the methodology of validated administrative data algorithms that aim to balance specificity and sensitivity.^21-23 This approach is also consistent with the previously developed definition of MS that was determined by a panel of NHF data experts and specialists in neurology.

RA cases were identified by at least 2 health care encounters with ICD‑10 codes M05 or M06, occurring at least 90 days apart, alongside at least 1 reimbursed prescription for a disease‑modifying antirheumatic drug (DMARD) based on Anatomical Therapeutic Chemical codes. Eligible DMARDs included methotrexate (L01BA01, L04AX03), sulfasalazine (A07EC01), leflunomide (L04AA13), or a biologic / targeted synthetic DMARD (as part of an NHF‑financed program). In the case of RA, the disease onset was defined at the second recorded health care encounter. The patients were further classified as seropositive RA (SPRA; ICD‑10 M05 claim) or seronegative RA (SNRA; ICD‑10 M06 claim) based on available diagnostic records. The patients recorded with both M05 and M06 codes during the observation period were excluded from all analyses; this approach was chosen to reduce double‑counting and incorrect classification, though it introduces an unknown degree of bias in terms of unspecified or transitional RA patterns. General characteristics of the patients with MS and RA are provided in Table 1.

Control group

We used the Polish public payer database, covering over 38 million people with unique PESEL identifiers, as the recruitment pool. The controls were identified from the general population using random sampling (1:1) based on age, sex, and calendar year of the specific RA case index date (ie, the year when the RA patient to be matched fulfilled the case definition). No geographic strata were used. Each control was required to have at least 1 recorded health care encounter of any type (inpatient, outpatient, or rehabilitation service) within the same calendar year. This criterion was applied to ensure inclusion of individuals actively using the public health care system during the same period, which is particularly relevant for younger persons. Calendar‑year matching was performed to maintain a sufficient sampling pool (Figure 1).

Ethics approval and consent to participate

This analysis relied solely on data obtained from the NHF electronic databases. Obtaining institutional ethical approval was not required.

Availability of data and materials

The datasets used and analyzed in this study were derived from the administrative database of health care claims of the Polish NHF and the data regarding mortality under the Ministry of Digital Affairs. Due to data protection regulations and restrictions imposed by the NHF, the full dataset is not publicly available. However, aggregated data on MS prevalence and incidence in Poland can be accessed through the publicly available repository maintained by the Polish Ministry of Health at https://analizy.mz.gov.pl/html/zpa_sm/.

Access to the complete anonymized datasets analyzed in this study may be granted upon reasonable request and subject to approval by the relevant regulatory authorities.

Statistical analysis

All analyses were performed using R 4.4.1 package (R Core Team, 2024; R Foundation for Statistical Computing, Vienna, Austria). Annual MS incidence rates were calculated as the number of newly identified MS cases per calendar year divided by the mid‑year population at risk in each cohort, with 95% CIs estimated using the exact Poisson methods. Prevalence rates were calculated per 10 000 persons, with 95% CI estimated using the Wilson method. Differences in MS prevalence between RA patients and the comparator group were evaluated separately by sex and age strata using 2‑sample tests for proportions without continuity correction. The Fisher test was utilized if any observed or expected cell counts were below 5. The Spearman correlation was used to assess temporal incidence trends. Effect sizes were estimated using prevalence ratios (PRs) with 95% CIs calculated by the Katz log method for each strata. Pooled PRs for age and sex strata were calculated using the Mantel–Haenszel method. To account for multiple exploration comparisons, the Holm adjustment was applied and reported alongside raw P values. As a form of additional sensitivity analysis, the Poisson regression with robust SEs was used to compare RA serostatus subgroups (M05 and M06) with the control group, after adjusting for sex and age. All statistical tests were 2‑sided, and a P value below 0.05 was considered significant.

Prevalence and incidence of multiple sclerosis in the Polish population

Using the adopted case definition, in 2021 there were 55 053 individuals in Poland living with MS, mostly women (70.9%), at a median age at diagnosis of 50 (32–52) years. In the same year, a total of 2588 new MS cases were identified, mostly in women (70.1%) at a median age of 37 (33–54) years. Overall, the prevalence of MS in Poland was estimated at 14.46 per 10 000 inhabitants, with higher rates in women (19.84) than in men (8.7). The incidence rate for MS in Poland in 2021 was 0.68 per 10 000 individuals, with a higher incidence in women (0.92 vs 0.42 in men).

Age‑dependent burden of multiple sclerosis in rheumatoid arthritis

Among RA patients followed between 2009 and 2021, we identified 464 individuals (0.15%) with concurrent MS diagnosis, as compared with 813 MS cases (0.26%) among the matched controls. In general, RA patients had by 43% lower MS prevalence than the comparator group (Mantel–Haenszel PR, 0.57; 95% CI, 0.51–0.64; P <⁠0.001). The majority of prevalent MS‑RA cases were women (87.7%), at a median (interquartile range [IQR]) age at diagnosis of 52 (41–58) years. Temporal trends in the burden and demographic structure of the MS‑RA population are shown in Figure 2. Over time, generally stable female predominance was observed, with median age at MS diagnosis increasing from the fifth to sixth decade of life (Table 2). In combination with declining incidence over time (z, –0.92; P <⁠0.001; for incidence rates see Table 2), this suggests a relationship with changing demographic structure of the Polish population. Age and sex strata specific results are reported in Table 3 and Figure 3.

In women, the association between RA and MS prevalence differed across age groups. Young women (16–39 y) with RA showed higher MS prevalence (PR, 1.39; 95% CI, 1–1.92; P = 0.048; adjusted P = 0.19). Clear differences were observed among older women, in whom reduced MS prevalence was found (PR range, 0.56–0.4; all adjusted P <⁠0.001). Overall mortality was similar in the RA patients (34/464; 7.3%) and the control group (68/813; 8.4%). The distribution of time intervals between RA and MS diagnoses is shown in Figure 4.

Effect of rheumatoid arthritis serostatus on multiple sclerosis prevalence

Within the RA group, the SPRA patients had lower MS prevalence than the SNRA ones (pooled PR, 0.68; 95% CI, 0.57–0.82; P <⁠0.001; Table 4). In women, we observed significant differences in the older subgroup (40–55 y and ≥66 y; Table 4). In men, the small number of cases resulted in wide uncertainty. In the sensitivity analysis using the robust Poisson regression with shared controls, both subtypes had lower prevalence than the controls (PR for SNRA, 0.73; 95% CI, 0.58–0.92; PR for SPRA, 0.49; 95% CI, 0.39–0.61). Direct comparison between SPRA and SNRA suggests a reduced PR in seropositive disease (0.67; 95% CI, 0.5–0.9; P = 0.008).

Conclusions

This nationwide population‑based study describes the epidemiologic link between RA and MS. Rheumatologists and general practitioners caring for RA patients should remain alert to the possibility of MS, including atypical neurological symptoms. A high index of diagnostic suspicion should be retained in younger women aged from 16 to 40 years. While further prospective, observational studies on the age, sex, and serotype specific RA‑MS risk association need to be conducted, closer monitoring is warranted. The coexistence of autoimmune conditions may reflect an interface between shared and divergent mechanisms of immune dysregulation, which could explain the variable risk association across different age groups. Our findings point to the clinical need of tailored screening strategies among patients with autoimmune disorders, particularly those that may be present in clusters.