Table 1 gives a summary of recommended thresholds from several guidelines when using BOHB to help support a diagnosis of ketoacidosis. The evolution of the 3-mmol/L cutoff includes the recommendation stated as a “preliminary guideline,” which suggested this level based on 14 patients admitted with DKA who were diagnosed by other contemporaneous means (11,12). This was reinforced by a highly referenced study of 129 children and 337 adults that recommended a BOHB cutoff of 3.0 and 3.8 mmol/L for children (<16 years) and adults, respectively (13). These thresholds were derived from the blood BOHB concentrations, which were, on average, equivalent to a bicarbonate concentration of 18 mmol/L, so there is something of a circular argument, or at least a collinearity, with the way existing DKA diagnostic criteria were established. Where there was a disparity in that study between HCO3 and BOHB concentrations (the R2 for all ages was between 0.64 and 0.68), it was suggested by the authors that BOHB would be the more reliable guide, although they admitted that this “must be arrived at on theoretical grounds” (13). The subsequent widespread decision to adopt a 3-mmol/L threshold for all age groups based on this work presumably reflected taking a conservative approach. Most guidelines that do not directly refer to these articles refer to documents that either do or do not provide a source for their cutoffs.

A notable guideline exception concerning a diagnostic blood BOHB threshold is the consensus report in 2017 from 8 U.S. professional organizations, including the AACE, the ADA, and the Endocrine Society (4). It recommends that “serum ketones” greater than “the upper limit of the normal range” be part of the DKA definition. If expressed as blood BOHB, then this could equate to concentrations of just 0.6 mmol/L or greater, which has led some to question whether this could lead to overdiagnosis of DKA (14).

Other factors beyond patient age are thought to affect the relationship between bicarbonate and BOHB concentration, an example being end-stage kidney disease. One study showed that in DKA patients with bicarbonate concentrations <18 mmol/L, the mean BOHB concentration in those with end-stage kidney disease was 1.4 mmol/L lower than in patients with preserved renal function (15).

The source of intermediate thresholds, such as 0.6–1.5 mmol/L, suggesting ketonemia, or 1.6–3 mmol/L, suggesting impending DKA (16), seems more difficult to establish with clarity. The preliminary guideline from Wallace et al. suggested that BOHB concentrations <1 mmol/L were not associated with incipient ketosis, whereas concentrations between 1 and 3 mmol/L required further intervention (11). A number of subsequent studies, however, refer to the BOHB concentrations suggested by the ketone meter manufacturers themselves (17–19). In turn, the manufacturer thresholds stated in the Abbott/Medisense instructions seem largely based on a study where an elevated BOHB was defined as being >0.5 mmol/L and where a value >1.5 mmol/L was one of their diagnostic criteria for DKA (20). However, this study did not reference the source of these cutoffs; in fact, >1.5 mmol/L was their diagnostic threshold rather than potentially being an intermediate level. The “upper limit of normal” concentration may date back to a more historic study of fasting patients; this is not clear (21), although this study did identify some normal values, up to 1 mmol/L, in some individuals, as has been noted by others (22). More recently, nonfasting samples have been used to establish a much tighter reference interval of 0.02 mmol/L to 0.28 mmol/L, although the confidence intervals around these limits meant the upper reference value could truly be as high as 0.5 mmol/L; no sex difference was apparent (23).

Despite the variance in cutoffs and the uncertainty in their provenance, there has been some clinical verification of the 1.5-mmol/L value as a useful intermediate threshold: a systematic review (24) identified that in 2 studies a BHOB concentration >1.5 mmol/L, compared with standard measures of DKA, was associated with a positive predictive value of 50% and a negative predictive value of 100% (19,25).

When monitoring the response to treatment of DKA, the JBS reached their conclusion of aiming for BOHB to fall by at least 0.5 mmol/L/h through guideline group consensus and based upon previous work, which suggested “BOHB could be expected to fall by 1 mmol/L per hour” (5,11). ISPAD guidance of approximately 0.5 mmol/L/h cited a study of children and seems to have extrapolated its recommendation from the data found there (7,26).

The recommended levels of blood BOHB used for the identification and further management of euglycemic DKA are identical to that of hyperglycemic DKA (4,6). Indeed, one guideline has dispensed with plasma glucose as a criterion for DKA diagnosis in known diabetes patients, largely because patients can present with euglycemia (4). There is some evidence that resolution from the episode takes longer than anticipated when SGLT2 inhibitors have contributed to its development (27).