Multiple Sclerosis Journal 2016, Vol. 22(8) 983–985 DOI: 10.1177/ 1352458516655218 © The Author(s), 2016. Reprints and permissions: http://www.sagepub.co.uk/ journalsPermissions.nav MULTIPLE SCLEROSIS MSJ JOURNAL http://msj.sagepub.com 983 The identification of temporal trends in the incidence of disease or in the phenotypic characteristics of com- plex diseases such as multiple sclerosis (MS) over short time intervals may lead to the recognition of contributing environmental factors. In recent years, investigators have reported a rising incidence of MS in some world regions, an increasing female:male sex ratio, and a progressively earlier age of symptom onset. 1 However, analyses of temporal trends can suf- fer from multiple biases, leading to erroneous conclu- sions as previously reported for age of onset. 2,3 In this issue of Multiple Sclerosis Journal, Westerlind et al. examined temporal trends in the diagnosis of primary progressive multiple sclerosis (PPMS). They used the Swedish MS registry to identify 16,915 persons with MS, representing six birth cohorts (1946–1975) and seven date of diagnosis cohorts (1980–2014). 4 They found that the proportion of per- sons with PPMS at onset decreased from 19.2% in the 1946–1950 birth cohort to 2.2% in the 1971–1975 birth cohort, a remarkable decline. The authors have thoughtfully attempted to address many of the poten- tial biases that might account for this result, including survival bias, inadequate duration of follow-up in younger birth cohorts, and the effects of changing diagnostic criteria. Nevertheless, it remains difficult to accept this rapid decline in the occurrence of PPMS, and this may reflect incomplete control of these biases, which work together in complex ways. First, survival bias is probably operating as evi- denced by the older average age of onset of 46.3 in the persons with PPMS in the oldest birth cohort consistent with frequently reported age at onset as compared to 34.6 years in the youngest birth cohort. The average age at onset also declined in the relaps- ing-onset cohorts, from 38.4 in the oldest cohort to 29.8 in the youngest. This also illustrates a potential survivorship bias that may or may not be the same as that in the PPMS cohort. Second, the exclusion of the large proportion of unclassified patients (8.6%), a proportion nearly as large as those identified with PPMS (8.6%), may influence the results. This may have contributed dif- ferentially to reducing the denominator over time, impacting their findings. We note that one of the sen- sitivity analyses combined the two groups which greatly reduced the incidence rate ratio to 0.89. Finally, and perhaps most importantly, the inadequate and differential duration of follow-up by birth cohort may lead to a substantial overestimate of the decline in the proportion of PPMS. Consider the fact that the mean age at diagnosis for the 1945 birth cohort was 46.3. The current maximum age for the youngest cohort is 41–45 in 2016 showing that the younger birth cohorts are missing the older persons who get diag- nosed with MS because they have not been diagnosed yet, thus the differential follow-up time truncates the period over which to make the diagnosis within a birth cohort, affecting the prevalence in the cohorts. Consider a simple mathematical example to illustrate the effect of these issues. Suppose we have 1000 MS patients in each birth cohort and we estimate the pro- portion with PPMS to be 10% for each of those birth cohorts. Therefore, each birth cohort would include 100 PPMS and 900 relapsing-onset multiple sclerosis (ROMS) patients. We will adjust these numbers for deaths, which decline by birth cohort; for the follow- up time, which reduces the prevalence of PPMS in the cohort; and, finally, for the elimination of the unclas- sified group from the ROMS population. The adjust- ments have been informed by the literature, but are only meant to illustrate how such assumptions and estimates combine to produce remarkable declines in the prevalence of PPMS as shown in Table 1 of the Westerlind article. We assumed that 30% of the 1946–1950 birth cohort will have died by 2016, based on findings from the Declines in the diagnosis of primary progressive MS: A critical change in phenotype or critical measurement error? Gary R Cutter, Amber Salter and Ruth Ann Marrie Correspondence to: GR Cutter Department of Biostatistics, School of Public Health, The University of Alabama at Birmingham, 1665 University Boulevard, Ryals Public Health Building, Birmingham, AL 35294-0022, USA. cutterg@uab.edu Gary R Cutter Department of Biostatistics, School of Public Health, The University of Alabama at Birmingham, Birmingham, AL, USA Amber Salter Department of Biostatistics, Washington University in St. Louis, Saint Louis, MO, USA Ruth Ann Marrie Departments of Internal Medicine and Community Health Sciences, University of Manitoba, Winnipeg, MB, Canada; Multiple Sclerosis Clinic, Winnipeg, MB, Canada 655218MSJ 0 0 10.1177/1352458516655218Multiple Sclerosis JournalGR Cutter, A Salter research-article 2016 Editorial