In early lactation, dairy cows can develop metabolic imbalances like a significant energy deficit and in consequence subclinical ketosis. First of all β–hydroxybutyrate (BHBA), but also non-esterified fatty acids (NEFA) as well as glucose (Glc) are used to determine the metabolic state of cows in terms of subclinical ketosis (SCK) in blood serum. In this field study in ten commercial dairy farms, ten different non-invasive indicators associated with the risk of developing SCK and obtainable without handling of the cow were investigated in relation to BHBA, NEFA and Glc in early lactating dairy cows. Effects of the indicators were examined using correlation and linear mixed models. Established indicators like fat:protein ratio (FPR) or days in milk showed the largest potential for predicting BHBA and also NEFA and Glc. Parity had no significant effect on the model, except for Glc. In the analysis of covariance, an influence of parity was found between first and 2^nd to 4^th lactation for NEFA and Glc, however the picture was less clear for BHBA. While time for feeding and/or ruminating showed significant influence for all three metabolites, the direction of influence was not always as expected. The study supports the validity of FPR in the prediction of blood BHBA; the other non-invasive indicators investigated can help to identify cows at risk for developing subclinical ketosis, but under the conditions of practical farming like in the present study they were not perfectly suited to replace established systems requiring some handling of the cow.