In order to use C-reactive protein (CRP) in risk prediction in individuals, it is necessary to know how to obtain the habitual level of an individual and hence its biological variations: i.e. longitudinal variability within subjects and variability between individuals. This paper provides data on biological variability that is used to propose a strategy for assessing individual low levels of CRP. The longitudinal variability in individuals (intraindividual variability) is essential to know, but only reported in a very limited way. Additional data were calculated from in-house and requested databases for periods of follow-up from 5 days to 1 year. The intraindividual coefficient of variation (CVi) was found to be rather similar for several groups and periods and on average was approximately 30%. Reported analytical coefficients of variation of commercial and in-house methods generally are below 6%, which is well below the desired limit of half the CVi. The distribution of CRP in apparently healthy individuals is wide and skewed with interquartile values ranging between 150-250% of the median and an estimate of the composite coefficient of variation (CVc) of approximately 120%. The distribution is equally broad in several other groups studied such as type II diabetics and pregnant women. It is concluded that the coefficient of variation for the determination of CRP in a single blood sample is as high as approximately 30%, but that this is acceptable for the reliable positioning of individuals within the distribution of CRP in the group with a CVc as high as -120%. CRP can show unexpected outliers (increases) which can sometimes be explained by information from a short questionnaire and definitely identified by the analysis of a second blood sample after an interval of approximately 2 weeks. Similarly, to ascertain high values in a first sample a second blood sample can be analyzed. It should be noted that, in view of the significant intraindividual variability of CRP, the difference between the first and second values may reach 71%. The large intraindividual variability of CRP approximating 30% renders it difficult to position an individual reliably in smaller categories such as tertiles, quartiles or quintiles of the total distribution. It is suggested that it would be most practical to have a goal of a single decision level or threshold only. Positioning an individual into two groups with equally wide distribution is on the borderline of reliability for one blood sample. Multiple blood samplings are required for smaller categories and higher threshold levels. The use of a decision limit should further acknowledge the limited interclass stability of around r = approximately 0.5 for a single blood sample. The above considerations are summarized in a practical working scheme. This scheme can serve as a basis for further refinement, discussions and development of sampling and decision limits to be selected from medical and economical perspectives and tested in practice.