Strengths and limitations

Comprising about 800,000 individuals, the AMORIS cohort is unique, with more than 35 million laboratory values from nearly 600 different methods, and linkages to several national high quality registers and research cohorts using the Swedish personal identification number enabling a follow-up of more than 20 years.

The AMORIS cohort constitutes a large and representative part of the population of Stockholm County between1985-1996. Record linkage to Swedish national registers, clinical quality of care registers and research cohorts provides opportunities for detailed prospective observational studies of associations between biomarkers and the risk of chronic disease, morbidity and mortality. All biomarkers were measured on fresh blood/urine samples at the same clinical laboratory, based on well-documented analytical methods and for a large proportion in a routine health screening setting. More than 50% of the cohort had repeated examinations. Furthermore, there is a complete follow-up of mortality, cancer incidence and other major diseases requiring hospitalization for more than 20 years up to 2012. Linkages of the AMORIS cohort to national censuses have shown that it represents the general population of Stockholm County during the inclusion period.

A limitation of the AMORIS cohort is the partial availability of information on lifestyle factors, certain cardiovascular risk factors and specific treatments. However, such information is available in sub-cohorts to confirm results obtained in the entire cohort. For instance, information on major cardiovascular risk factors (including tobacco smoking, hypertension and obesity) is available in a sub-cohort of about 100,000 subjects, constituting about 15% of the total AMORIS cohort, which is sufficiently large in many situations at least as a sensitivity analysis. In addition, from 2005-2012 prescribed drug treatment is available via the National Prescribed Drug Register. If single risk factors are studied, sub-cohorts could amount to several hundred thousand subjects. For the AMORIS cohort there are no frozen blood samples. Also, there is no information on genetic biomarkers. Certain analyses of possible genetic impact may be performed using mono- and dizygotic twin pairs identified in the AMORIS cohort using the Swedish Twin Register. A slightly greater proportion of employed subjects in the AMORIS cohort compared to the general population give a “healthy worker effect” reflected by lower all-cause mortality. While this does not affect the internal validity of studies based on the AMORIS cohort, it could lead to underestimates of prevalence and incidence rates in the general population.

Miranda Beck