As readers of this blog will know I am a big fan of anything that challenges my practice. It’s something that I think in general is a good practice to live by. For many years now when a preterm infant in particular is hypotensive it has been our practice to draw a serum cortisol level and then determine whether the stress response is adequate before starting hydrocortisone for blood pressure support. Having said that, sometimes we start the hydrocortisone and then use the level afterwards to determine if we need to continue. is this approach right though?

Evidence That Shakes Up Our Approach

It turns out the evidence that preterm infants may not be able to produce a robust cortisol response after birth has been around for sometime. In 1994 Hingre et al published Adrenal steroidogenesis in very low birth weight preterm infants. In this paper they documented the diminished ability of infants born < 30 weeks gestational age to produce cortisol finding preterm newborns had low basal cortisol levels “(mean +/- SEM, 207.4 +/- 23.5 nmol/L), and their levels were similar to basal levels reported for healthy full-term newborns (170.7 +/- 26.8 nmol/L; P = 0.31”. It is worth noting here that commonly held beliefs have been that an adequate adrenal response is in the range of 400 – 450 nmol/L or about 15 microgram/dL and these levels are lower than that. Moreover, when the authors measaured precursors of cortisol and found elevations consistent with a deficiency of decreased activity of 11 beta-hydroxylase (11 beta OH). Knowing this then, the use of a baseline cortisol to determine if an appropriate stress response is present before starting hydrocortisone is questionable. Having said that the practice has been that when it is low we assist with hydrocortisone and when it is high we can ease off the support. A new study that has just come out though I think may turn that thinking on its head!

High Cortisol Levels Are Concerning. Not the Lows!

Absence of relationship between serum cortisol and critical illness in premature infants by Prelipcean I et al was just published and looked at 224 infants at the University of Florida who were born under 30 weeks and had baseline cortisol levels drawn for clinical indications prior to 36 weeks PMA. Like many centres the baseline cortisol was done prior to starting hydrocortisone for hypotension. A baseline level under 15 mcg/dL was considered low which equates to about 413 nmol/L for those using those units (like my own hospital). The Simplified Score for Neonatal Acute Physiology II SNAP-II score , neonatal Sequential Organ Failure Assessment (nSOFA) and Vasoactive-Inotrope Score (VIS) were calculated and used as measures of illness severity against the the cortisol levels obtained in a retrospective fashion. Cortisol levels were taken at a median of 3.8 days with an IQR of 1.2 to 14 days). Hydrocortisone was givne to 71% of patients in the study as well.

What emerged from these results might be counterintuitive. From the figure below it was found that those infants with higher baseline cortisol levels were less likely to survive. This result just reached statistical significance. Thinking about this for a moment, we have traditionally worried about the infants with low cortisol and rushed to supplement them. The babies at real risk though here are the ones with a robust pituitary adrenal axis response. Notably another factor that leads to lower cortisol levels in the first few days of life is provision of antenatal steroids so it may be at least in part that the higher baseline levels might be seen in those without the benefit of antenatal steroids and therefore are at higher risk of adverse outcome. Bottom line though, a robust cortisol level would not necessarily appear to be marker of a good thing.

The second thing to be identified is the scatter of results for these infants across birth weight, day of life and gestational age. The authors discovered using a multivariable model that birth weight was the only statistically significant variable to explain cortisol variation. Interestingly for every 100g increase in birth weight cortisol increased an average of 10%.

Additionally, differences in average cortisol level were affected by chorioamnionitis and antenatal steroids. The presence of chorioamnionitis as a variable is not surprising I suppose given the results from the prophylactic steroid trials for BPD that have consistently found chorio predicts a higher rate of BPD.

Where things get really interesting is in the bottom half of the figure below. While weak linear associations with SNAP-II, nSOFA were found ,no correlation between serum cortisol concentration and concurrent critical illness severity objectively measured by SNAP-II and nSOFA scores at time points beyond the first day of life and prior to 36 weeks PMA in these infants were found. Most intriguing was the complete lack of relationship between the VIS and cortisol levels.

This presents a predicament about what to do with these levels. Based on this research the degree of illness and the amount of inotrope one is on (VIS takes into account doses of dopamine, dobutamine, vasopressin, milrinone, epinephrine and norepinephrine) has no relationship to cortisol level. If you are like our centre though you have been considering whether to use hydrocortisone based on the level of cortisol at baseline. Based on this research the message would be that if one wants to know a baseline cortisol it might be useful as a tool to determine how concerned one should be with an infant as risk of mortality is higher if baseline levels are above 413 nmol/L. In terms of determining whether one should support with hydrocortisone though in the face of a sick preterm infant and more specifically a hypotensive one the utility of the baseline measurement I would question. Adding to this the research from 1994 and one has to question if the level is low is that simply because the infant doesn’t have the metabolic machinery yet to produce enough rather than has an abnormal response to stress.

Some qualifiers as with any study like this need to be acknowledged. It is not a study of 1000 patients so the individual numbers of patients at different weight levels will be lower and therefore there could be unusual patients here influencing the results. Having said that, when you combine this information in this study with what is known from before about these preterm infants should we be surprised that there is no relationship between baseline cortisol and illness. If you don’t have the capacity to make it except when exceptionally stressed it would appear that all these baseline cortisols may in fact be good for telling ourselves how stressed we should be about the patient.