Critique 310 – Alcohol consumption and incident osteoarthritis: a nationwide population-based longitudinal study

Cho, S., Yong, S., Ha, E., Kim, J.-H., Kim, Y.-J. Osteoarthritis and Cartilage 2026, in press

https://doi.org/10.1016/j.joca.2026.09.005

Abstract

Objective: To examine the association between alcohol consumption and incident osteoarthritis (OA) and whether it persists after restriction for liver-related and weight-related factors.

Design: We analysed South Korean National Health Insurance Service data for adults aged ≥40 years who underwent health screening in 2009-2011 and had no recorded OA at baseline. Alcohol consumption was assessed by self-reported drinking frequency and estimated daily ethanol intake. Participants were followed until first claims-based OA diagnosis (ICD-10 M15–M19, excluding traumatic OA), death, or December 31, 2023. Multivariable Cox models estimated adjusted hazard ratios (HRs). Prespecified restricted analyses were conducted in participants with normal liver function and no liver disease history and in participants with normal weight and no obesity history.

Results: Among 1,233,179 participants with drinking-amount data (mean age, 50.8 years), the mean follow-up period was 9.9 years, during which 571,077 cases of OA occurred. Compared with non-drinkers, the light-drinking group had an adjusted HR of 1.037 (95% CI, 1.030–1.045), the moderate-drinking group had an adjusted HR of 1.081 (95% CI, 1.072–1.091), and the heavy-drinking group had an adjusted HR of 1.186 (95% CI, 1.174–1.197). A similar dose-response pattern was observed for drinking frequency and across sensitivity analyses. The pattern also persisted in the normal liver function and normal weight cohorts, and women showed elevated OA risk at lower levels of alcohol consumption than men.

Conclusion: Alcohol consumption was associated with incident OA in a dose-dependent manner that persisted after liver-related and weight-related restrictions, suggesting that this association is unlikely to be explained solely by hepatic dysfunction or obesity. These findings identify alcohol consumption as a factor associated with incident OA, warranting further investigation as a potentially modifiable contributor to OA risk.

ISFAR Summary

Osteoarthritis (OA) is a major and growing cause of pain, disability and reduced mobility worldwide, particularly as populations age. A large South Korean study by Cho et al. (2026), involving more than 1.2 million adults followed for almost 10 years, found that alcohol consumption was associated with a higher likelihood of developing OA, with the association strengthening as consumption increased. Compared with non-drinkers, the relative risk was approximately 4% higher among light drinkers, 8% higher among moderate drinkers and 19% higher among heavy drinkers. Although the study’s large size, long follow-up and extensive additional analyses strengthen its findings, the increases at light and moderate levels of consumption were relatively small, and their clinical significance remains uncertain.

While the findings are potentially important, caution is warranted in interpreting them as proof that alcohol causes OA. Alcohol consumption was self-reported, previous drinking histories were not fully established, and important influences such as genetic susceptibility, previous joint injury and physically demanding occupations may not have been adequately accounted for. The study also raises questions about differences between men and women and whether the findings from a South Korean population apply equally to other countries. Overall, the research adds to the investigation of potentially modifiable risk factors for OA, but further studies in other populations, with better measurement of alcohol consumption and other contributing factors, are needed before firm conclusions can be drawn.

Background

Osteoarthritis (OA) is a chronic, degenerative joint disease characterised by the gradual breakdown and thinning of articular cartilage. OA also changes the underlying bone and is accompanied by inflammation of the synovial fluid (Kraus et al., 2015). OA differs from rheumatoid arthritis (RA); although both affect joints, their aetiologies differ. RA is an autoimmune disease that primarily causes painful inflammation, swelling, and stiffness in the joint lining, whereas OA involves cartilage degeneration. Also, RA usually occurs in young to middle adulthood, whereas OA usually occurs after middle age.

In 2020, 595 million people were estimated to have OA worldwide with an estimated global age-standardised prevalence of almost 7000 per 100 000 people. It is a leading cause of pain and disability worldwide. The study found that 61% of all global OA cases occur in women compared with 39% in men (Steinmetz et al., 2023). However, the likelihood of developing OA shifts across different life stages. At a young age, OA occurs more frequently in men, usually driven by joint deformities or sports- and work-related traumatic injuries, whereas in middle age the incidence spikes in women. During this window, women develop OA at nearly twice the rate of men.

Current management strategies for OA include non-pharmacological treatments such as exercise and other forms of physical therapy, weight loss and surgical treatments including joint replacement (Kolasinski et al., 2020). No approved pharmacological treatment currently cures or reverses OA. Pharmacological treatment may include pain medication, non-steroidal anti-inflammatory drugs and corticosteroids.

Recent insights into OA pathophysiology reveal a complex interplay between genetic and lifestyle risk factors. Twin and other family-based inheritance studies have estimated OA heritability at 40–65%, depending on the joint site (Kraus et al., 2007). Overweight and age are important risk factors that act through joint degeneration and overloading. These processes are thought to be triggered by biomechanical injury or microtrauma, which may interact with genetic susceptibility and other environmental factors (Kloppenburg et al., 2025). Other risk factors include bone mineral density (BMD). Large prospective cohort studies show that individuals with high BMD have a 2.3–2.9-fold greater risk of developing knee OA. Recent Mendelian randomisation studies confirm a direct causal relationship (R. Zhu et al., 2025). A recent systematic review and meta-analysis evaluated a large number of potential risk factors for incident knee OA, indicating that overweight/obesity and previous knee injury account for 14% of incident knee OA (Duong et al., 2025).

This review also noted two studies examining alcohol consumption as a risk factor for knee OA. The authors concluded that pooled estimates showed no association between alcohol intake and incident knee OA. In this context, Cho et al. (2026) is notable because this nationwide, population-based longitudinal study found that alcohol consumption is positively associated with incident OA in a dose-dependent manner.

Critique

Cho et al. (2026) concluded that alcohol consumption is positively associated with incident OA. They observed a dose-dependent association even after adjusting for liver- and weight-related factors. Although the extensive corrections, sensitivity analyses and repeated assessment of alcohol exposure seem robust, relative risks increase only modestly with drinking amount; a 4% increase in light drinkers (defined as < 10 g/day for men and women) and an 8% increase in moderate drinkers (defined as 10-30 g/day for men and 10-20 g/day for women). The relative risk increases at light and moderate consumption levels are small, and their clinical significance remains uncertain. The adjusted 10-year cumulative incidence was 34.54% among non-drinkers, 35.52% among light drinkers, 36.64% among moderate drinkers and 39.21% among heavy drinkers. These estimates correspond to differences of approximately 1.0, 2.1 and 4.7 percentage points, respectively, compared with non-drinkers. Presenting these alongside relative risks would help in assessing the magnitude of the association. These are adjusted observational estimates, not proof of alcohol-caused excess cases.

Alcohol exposure was assessed using self-administered questionnaires completed during health examinations at two time points, which makes the results susceptible to under-reporting and other errors associated with self-reported alcohol consumption. The participants classified as non-drinkers may have included former drinkers, as drinking histories before baseline were not fully established. Although drinking frequency was assessed, more detailed drinking patterns, such as episodic heavy drinking, were not reported. The longitudinal design establishes that alcohol exposure was assessed before the recorded diagnosis of OA, strengthening the temporal evidence compared with cross-sectional studies; however, it does not establish causation.

Participants were categorised into four groups, with drinking amounts differing between men and women, whereas relative risks were expressed for light, moderate and heavy drinkers. Although the rationale for using different drinking amounts by gender seems logical, it is based on the assumption that women will be more susceptible in this specific exposure-outcome relation than men. In their discussion, the authors report a stronger association in women and a null or slightly inverse association in men who consume lesser amounts of alcohol. Although the authors present the observations in women as stronger, they attribute the observations in men to residual confounding, exposure misclassification, or drinking patterns.

A strength of this study is its large cohort, which provides substantial statistical power to evaluate the association of interest, as the authors note in the discussion section. While this may help detect small effects, substantial statistical power may also reveal small differences that confounding factors could explain. One such unaccounted-for confounding factor may be genetic predispositions present in the Korean population. Although modern-day ethnic Koreans are considered one of the most genetically homogeneous populations in the world (An et al., 2025), individual differences in gene variants will exist. Variants in genes such as ALDH may result in substantial differences in alcohol metabolism. Interestingly, ALDH2 variants are associated with bone mineral density, an established risk factor for OA. The ALDH1A2 variant has also been identified as strongly associated with severe and erosive hand OA (L. Zhu et al., 2022). Although ALDH is briefly mentioned by the authors in relation to alcohol metabolism, no confounding by genetic background is discussed or assumed for a disease with an estimated heritability of 40–65%.

Other important confounding factors, such as the role of trauma and repetitive or physically demanding work, have only been partially accounted for. The influence of pre-existing but undiagnosed OA at baseline was minimised by excluding participants diagnosed within one year of the health examination. However, the study captured the role of repetitive or physically demanding work only through regular exercise and low or high income, which may also confound the analysis.

Overall, this well-conducted and impressive study provides further evidence of an association between alcohol consumption and incident osteoarthritis (OA). However, the relative risks at light and moderate consumption levels were modest, and their clinical significance remains uncertain. The findings warrant confirmation in studies that better account for other potential contributing factors, including genetic susceptibility and occupational biomechanical stress. The longitudinal design, dose-response pattern and extensive sensitivity analyses strengthen the evidence for an association. Nevertheless, this remains an observational study, and residual confounding, exposure misclassification and potential selection bias cannot be ruled out. The findings do not, by themselves, establish that alcohol consumption causes OA. In addition, the study was conducted in South Korea, where drinking patterns, social practices, genetic backgrounds and other population characteristics may differ from those in other countries. Replication in other populations would help establish how broadly the findings apply.

Specific member comments

Forum member Skovenborg comments that “the authors of this interesting study propose low-grade inflammation as a result of alcohol consumption as a possible mediator because low-grade systemic inflammation has been implicated in OA pathogenesis. The soluble urokinase plasminogen activator receptor (suPAR) is a biomarker of low-grade inflammation (and mortality risk) in various patient populations. In a Danish study lifestyle, demographic, and cardiovascular disease (CVD) risk factor data were collected from 5,538 participants in the Danish population-based Inter99 study (Haupt et al. 2014). Their suPAR levels were measured using a sandwich enzyme-linked immunosorbent assay. In the final adjusted model, smoking and morbid obesity were strongly associated with higher suPAR levels (P , 0.001). An unhealthy diet and alcohol abstinence in men were also associated with higher suPAR levels. Physical activity in leisure time had a modest impact on suPAR levels in univariate analysis, but not in the final adjusted model. In conclusion, smoking and morbid obesity were strongly associated with higher serum suPAR levels in this general population. Diet and alcohol consumption (in men) also seemed to impact suPAR levels. In adjusted analyses alcohol showed the same sex-specific associations as in univariate analysis: higher suPAR levels in abstinent men and lower suPAR levels in women who overused alcohol (P , 0.01). That might explain the stronger association between alcohol intake and OA in women and the null or slightly inverse association found in men.”

Forum member Zhang suggests that “The evidence regarding the association between alcohol consumption and osteoarthritis (OA) remains inconsistent, and an independent association has not been firmly established. Indeed, several systematic reviews and meta-analyses have not identified alcohol consumption as an established risk factor for OA. For example, To et al. (2021) found that the apparent inverse association between alcohol consumption and OA was no longer statistically significant when analyses were restricted to studies that adjusted for potential confounders. The findings reported by Cho et al. should therefore be interpreted in the context of this limited and conflicting evidence, rather than viewed as definitive evidence that alcohol consumption is an independent risk factor for OA. Several aspects of the study warrant careful scrutiny as follows:

1. Potential misclassification of OA diagnosis

OA ascertainment was based on ICD diagnostic codes, but the validity of this approach in the South Korean setting is unclear. In a validation study of knee OA identification in UK primary-care records, Yau et al. (2022) reported a positive predictive value (PPV) of 64% for a single diagnostic code against a reference standard combining knee pain and imaging findings consistent with OA. Requiring at least two OA codes over six months increased the PPV to 92%, although sensitivity decreased substantially to 19%. These findings suggest that misclassification of OA based on physician-recorded diagnostic codes may not be trivial. However, how does this misclassification affect the current study results remains unclear.

2. Lack of information on occupational physical workload

The authors state that the analyses were adjusted for lifestyle factors but do not clearly specify all the variables included in the adjusted models. In particular, occupational physical workload does not appear to have been specifically assessed. Certain manual occupations are associated with higher prevalence of heavy alcohol consumption, while occupational activities involving heavy lifting, kneeling, squatting, and repetitive joint loading are established risk factors for OA at particular joint sites. Therefore, the lack of detailed information on occupational physical demands may have resulted in residual confounding and potentially influenced the observed association between alcohol consumption and incident OA.

3. Combining OA across different joint sites

The study appears to have assessed incident OA as a composite outcome rather than reporting separate associations for knee, hip, and hand OA. However, the risk factors for OA may differ by joint site, particularly with respect to occupational biomechanical loading. Combining OA at different sites may therefore obscure potentially important differences in the association between alcohol consumption and OA. If data on joint replacement from OA at knee or hip joints are available, that may be a better outcome to investigate.

4. Potential competing risk of death

The study used Cox proportional hazards models to examine the association between alcohol consumption and incident OA but did not explicitly account for death as a competing event. Because individuals who die can no longer subsequently receive an OA diagnosis, competing mortality may be relevant, particularly if mortality differs across alcohol consumption categories. Heavy alcohol consumption may be associated with higher mortality, potentially affecting the observed incidence of OA during follow-up. However, the direction and magnitude of any resulting bias cannot be determined a priori. The authors should report mortality across alcohol consumption categories and consider a competing-risk analysis, such as the Fine–Gray subdistribution hazards model, alongside the cause-specific Cox model. This would help assess the robustness of the findings and clarify the association between alcohol consumption and the cumulative incidence of OA in the presence of competing mortality.

5. Covariates adjustment in the Cox proportional model

The authors stated that the multivariable adjustment model included sex, age, income, health examination measurements, lifestyle variables, and a history of chronic diseases. However, the specific lifestyle variables included in the model are unclear. Was body mass index (BMI) included among these variables? In addition, the rationale for adjusting for a history of chronic diseases as potential confounders requires further clarification. Alcohol consumption is a habitual lifestyle behavior, and some chronic diseases may develop after drinking habits have been established. If these diseases are consequences of alcohol consumption rather than true confounders, adjusting for them could introduce selection bias. Furthermore, Table 2 shows that the crude 10-year risk of OA decreased as alcohol consumption increased, from 44.07% to 37.06% and 33.26%, before increasing to 36.53% across the four exposure categories. However, the association appeared to reverse after multivariable adjustment. It would be helpful if the authors could clarify which covariates contributed most to this reversal and provide a plausible explanation for the substantial change in the estimated association. This marked change raises important questions about the potential influence of confounding on the observed association and warrants further investigation.

Forum member Romano concludes her extensive comments, which are available in full on the ISFAR website,with “Cho et al. (2026) provide epidemiological evidence of a positive association between self-reported alcohol intake and subsequent recorded OA diagnosis in a large South Korean cohort. Longitudinal follow-up and multiple statistical analyses are notable strengths. Nevertheless, they do not fully offset limitations in participant selection, missing exposure data, drinking classification, administrative outcome ascertainment and incomplete adjustment for the multifactorial aetiology of OA. Sex-specific, non-equivalent intake categories; incomplete information on drinking patterns and duration; inadequate measurement of biomechanical and hormonal determinants; and violation of the proportional hazards assumption all constrain interpretation. High statistical power can detect small associations without demonstrating clinical importance or causality. The light- and moderate-drinking findings are especially susceptible to residual confounding. The larger, comparatively consistent association in heavy drinkers likewise cannot be attributed unequivocally to a direct ethanol effect.

Ultimately, the study demonstrates an adjusted statistical association but does not establish alcohol as an independent cause of OA or show that reducing intake prevents the disease. Its conclusion should acknowledge the need for better exposure characterisation, validated joint outcomes and more comprehensive control of the disease’s multifactorial determinants.”

Forum member Ellison considers this to be “a very well-done and intensive analysis by the authors, with testing of their results with a large number of sensitivity sub-analyses that supported their main finding. 

In their summary, the authors agreed with many limitations of their paper.  A key limitation is that they had to base their estimate of alcohol exposure only on the self-reports by subjects of the number of drinks they consumed, with inadequate data on the pattern of such consumption.  Drinking patterns include frequency, whether consumption is regular or in binges, and especially the type of beverage.  It is also important to know whether or not it is consumed with or without food, data not provided in this paper.

Further, these analyses are from Korean subjects who consume, among other beverages, certain ones common only in Korea.  Thus, the reported effects of alcoholic beverages in this paper might apply or not apply to other populations where beer, wine, and/or spirits are the usual beverages.  It was also noted that the rates of abnormal liver function or a history of liver disease were rather high for this population, making one consider that causes other than alcohol might be playing a role.  These factors help determine how useful this study’s results might be for setting appropriate drinking guidelines for the general public. 

It may have been useful to evaluate the risk of OA for individuals reporting more frequent but lower amounts of alcohol per occasion, rather than just the total amount consumed over a period of time.  Regular (even daily) consumption of small amounts is usually found to be associated with the lowest risk of cardiovascular disease, diabetes, and total mortality.  Further, in almost all epidemiologic studies, wine intake is associated with healthier outcomes than the intake of other beverages.  Data on subjects consuming predominantly wine were not presented in their analyses.  “Daily” intake was estimated by using the total amount and frequency of drinking (while the number of beverages may differ markedly between weekdays and weekends).  It is believed that health outcomes may be based especially on the blood alcohol level attained from a beverage, which would be higher for greater amounts consumed on only one or two days per week than when spread out over the week.”

Forum member Harding “supports the comments of other Forum members, particularly the conclusion that the study does not establish alcohol as an independent cause of osteoarthritis, and the observation on the lack of data on drinking patterns.  Daily alcohol consumption was estimated based on weekly drinking frequency and the amount consumed per drinking occasion.

ISFAR Critique #295 also concerned a Korean cohort and noted that most alcohol consumed in Korea was beer (67%) and soju (32%).  The alcoholic strength of soju varies widely, so the assumption that it is 21% ABV (Reference 18, Supplement) might not be robust.”

References

An, K., Jeon, S., Kwon, Y., Choi, Y., Yoon, C., Jeon, Y., Bhak, J., Shin, D.-H., Choi, H.-J., Lee, H., Kim, Y. J., Shin, E.-S., Ryu, H., Blazyte, A., Park, S., Cho, J., Bolser, D., Joe, S., Yang, J. O., … Bhak, J. (2025). 10,239 whole genomes with multi-omic and clinical health information as the Korean population multi-omic reference dataset. BioRxiv, 17, 2025.11.17.688763. https://doi.org/10.1101/2025.11.17.688763

Cho, S., Yong, S., Ha, E., Kim, J. H., & Kim, Y. J. (2026). Alcohol consumption and incident osteoarthritis: A nationwide population-based longitudinal study. Osteoarthritis and Cartilage. https://doi.org/10.1016/j.joca.2026.09.005

Duong, V., Abdel Shaheed, C., Ferreira, M. L., Narayan, S. W., Venkatesha, V., Hunter, D. J., Zhu, J., Atukorala, I., Kobayashi, S., Goh, S. L., Briggs, A. M., Cross, M., Espinosa-Morales, R., Fu, K., Guillemin, F., Keefe, F., Stefan Lohmander, L., March, L., Milne, G. J., … Cooper, C. (2025). Risk factors for the development of knee osteoarthritis across the lifespan: A systematic review and meta-analysis. Osteoarthritis and Cartilage, 33(10), 1162–1179. https://doi.org/10.1016/J.JOCA.2025.03.003

Haupt et al. Risk Factors Associated with Serum Levels of the Inflammatory Biomarker Soluble Urokinase Plasminogen Activator Receptor in a General Population. Biomarker Insights 2014:9 91–100 https://doi.org/10.4137/BMI.S19876.

Kloppenburg, M., Namane, M., & Cicuttini, F. (2025). Osteoarthritis. The Lancet, 405(10472), 71–85. https://doi.org/10.1016/S0140-6736(24)02322-5

Kolasinski, S. L., Neogi, T., Hochberg, M. C., Oatis, C., Guyatt, G., Block, J., Callahan, L., Copenhaver, C., Dodge, C., Felson, D., Gellar, K., Harvey, W. F., Hawker, G., Herzig, E., Kwoh, C. K., Nelson, A. E., Samuels, J., Scanzello, C., White, D., … Reston, J. (2020). 2019 American College of Rheumatology/Arthritis Foundation Guideline for the Management of Osteoarthritis of the Hand, Hip, and Knee. Arthritis & Rheumatology (Hoboken, N.J.), 72(2), 220–233. https://doi.org/10.1002/ART.41142

Kraus, V. B., Blanco, F. J., Englund, M., Karsdal, M. A., & Lohmander, L. S. (2015). Call for standardized definitions of osteoarthritis and risk stratification for clinical trials and clinical use. Osteoarthritis and Cartilage, 23(8), 1233–1241. https://doi.org/10.1016/j.joca.2015.03.036

Kraus, V. B., Jordan, J. M., Doherty, M., Wilson, A. G., Moskowitz, R., Hochberg, M., Loeser, R., Hooper, M., Renner, J. B., Crane, M. M., Hastie, P., Sundseth, S., & Atif, U. (2007). The Genetics of Generalized Osteoarthritis (GOGO) study: study design and evaluation of osteoarthritis phenotypes. Osteoarthritis and Cartilage, 15(2), 120–127. https://doi.org/10.1016/j.joca.2006.10.002

Steinmetz, J. D., Culbreth, G. T., Haile, L. M., Rafferty, Q., Lo, J., Fukutaki, K. G., Cruz, J. A., Smith, A. E., Vollset, S. E., Brooks, P. M., Cross, M., Woolf, A. D., Hagins, H., Abbasi-Kangevari, M., Abedi, A., Ackerman, I. N., Amu, H., Antony, B., Arabloo, J., … Kopec, J. A. (2023). Global, regional, and national burden of osteoarthritis, 1990-2020 and projections to 2050: a systematic analysis for the Global Burden of Disease Study 2021. The Lancet Rheumatology, 5(9), e508–e522. https://doi.org/10.1016/S2665-9913(23)00163-7

To K, Mak C, Zhang C, Zhou Y, Filbay S, Khan W. (2021) The association between alcohol consumption and osteoarthritis: a meta-analysis and meta-regression of observational studies. Rheumatology International, 41, 1577–1591. https://doi.org/10.1007/s00296-021-04844-0

Yau MS, Dubreuil M, Li S, et al. Validation of knee osteoarthritis case identification algorithms in a large electronic health record database. Osteoarthritis and Cartilage Open. 2022;4(1):100229.

Zhu, L., Kamalathevan, P., Koneva, L. A., Zarebska, J. M., Chanalaris, A., Ismail, H., Wiberg, A., Ng, M., Muhammad, H., Walsby-Tickle, J., McCullagh, J. S. O., Watt, F. E., Sansom, S. N., Furniss, D., Gardiner, M. D., & Vincent, T. L. (2022). Variants in ALDH1A2 reveal an anti-inflammatory role for retinoic acid and a new class of disease-modifying drugs in osteoarthritis. Science Translational Medicine, 14(676). https://doi.org/10.1126/SCITRANSLMED.ABM4054

Zhu, R., Xing, X., Bian, J., Zhang, X., Ge, L., & Cai, G. (2025). Causal association between bone mineral density and the risk of joint replacement in patients with osteoarthritis: a Mendelian randomization study. Clinical Rheumatology, 44(2), 823–830. https://doi.org/10.1007/S10067-024-07289-5

Comments on this critique by the International Scientific Forum on Alcohol Research were provided and supported by the following members:

Henk Hendriks, PhD, Independent consultant and partner of the Nutrition Consultants Cooperative, Netherlands

Creina Stockley, PhD, MBA, Independent consultant and Adjunct Senior Lecturer in the School of Agriculture, Food and Wine at the University of Adelaide, Australia

Erik Skovenborg, MD, specialised in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark

Yuqing Zhang, MD, DSc, Clinical Epidemiology, Boston University School of Medicine; Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA

R. Curtis Ellison, MD, Section of Preventive Medicine/Epidemiology, Boston University School of Medicine, Boston, MA, USA

Raquel Romano, PhD, Independent consultant and Professor of Applied Technology at the University of Aconcagua, Argentina

Matilda Parente, MD, Consultant in Molecular Pathology/Genetics and Emerging Technologies, San Diego, CA, USA

Richard Harding, PhD, Formerly Head of Consumer Choice, Food Standards and Special Projects Division, Food Standards Agency, UK

Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy