Alex Nowbar reviews the latest research from the top medical journals

Covid-19 risk calculator

Risk calculators strike me as being useful only if you can act on the results. Medicine’s treatment options for covid-19 have not really been tested for use in such a way. Prognosis is nice to know, but also not great if there aren’t any interventions. However, the Severe COVID-19 Adaptive Risk Predictor (SCARP) still got my interest antennae up. The calculator has only two inputs: the supplemental oxygen and pulse oximetry recorded at the most hypoxic moment in the past six hours. The prediction of progression to severe illness or death was very accurate. It was developed using machine learning with data from a few thousand patients across five US hospitals. The authors suggest that the tool could help “inform hospital operations to best use resources in meeting the ever-changing demand for intensive care.” This is not unreasonable. But I’m not sure the medical profession is ready to implement something like this yet—perhaps we are ready to have the information, but a little too suspicious to use it to make decisions.

Ann Intern Med doi:10.7326/M20-6754

Risk of severe hypoglycaemia with long acting insulin

Bradley and colleagues compared severe hypoglycaemic event rates (based on the need to come to hospital) between two forms of insulin in a large cohort of people aged at least 65 years in the US. This wasn’t a randomised trial, so the first issue is that any relationship will be confounded by factors leading some people to be given one prescription over the other. Also, the relative doses being delivered in the groups might vary. The vast majority of the over half a million-strong cohort were using long acting insulin analogues insulin glargine or insulin detemir; less than 5% were using NPH insulin (an intermediate-acting insulin). NPH stands for neutral protamine Hagedorn; the first two words being descriptors and the last being the name of the founder of Nordisk (now Novo Nordisk). Despite this being an observational study, the results are convincing. The risk of hypoglycaemia presentations among those taking long acting insulin was much lower than in those taking NPH insulin. However, this analysis excluded the 29% of people who used both long acting insulin and a prandial insulin. In those users, the hypoglycaemia risk was no different from that seen with NPH insulin.

JAMA Intern Med doi:10.1001/jamainternmed.2020.9176

Is medication safety improved by clinical pharmacist intervention?

Gurwitz and colleagues randomised 361 people aged 50 years or over who were discharged from hospital on anticoagulants, diabetes agents, or opioids to a multifaceted clinical pharmacist intervention or to standard care. These common drug classes have potential for adverse effects and prescribing errors that may be particularly harmful. The multifaceted intervention involved an in-home assessment by a clinical pharmacist, evidence based educational resources, communication with the primary care team, and telephone follow-up. It seems obvious that a pharmacist-led, transitional care intervention would reduce adverse drug events, but in this trial the intervention was not associated with lower rates of adverse events or medication errors at 45 days. However, the authors note that the intervention was initiated after discharge and a future intervention might be more effective if started in hospital. It might also have been nice to see some patient-reported outcomes here.

JAMA Intern Med doi:10.1001/jamainternmed.2020.9285

Correction of low thing makes no difference to important thing

In many conditions, a result outside the normal range predicts poor outcomes. Intuitively we want to correct that result (such as by giving replacement therapy) because we believe it will improve outcomes. Unfortunately, trying to correct the result doesn’t necessarily improve outcomes. Albumin in decompensated cirrhosis is a case in point. The ATTIRE study randomised 777 people with decompensated cirrhosis and low serum albumin (<30 g/L) to targeted intravenous 20% human albumin (aiming to maintain serum albumin ≥35 g/L) or placebo in an open-label trial. The primary endpoint was a composite of new infection, kidney dysfunction, or death. Rates of this endpoint did not differ between the two groups. In particular, the expectation was that increasing serum albumin would reduce infections as low albumin is associated with immune dysfunction, but this didn’t pan out.

N Engl J Med doi:10.1056/NEJMoa2022166

Nor does correction of a high thing

Terlipressin is a vasopressin analogue. It is used to treat type 1 hepatorenal syndrome, and, while the CONFIRM trial confirms that it works for improving renal function, this 300 patient, double-blind, randomised controlled trial also found an increase in serious adverse events with terlipressin compared with placebo (particularly respiratory failure). Both groups received albumin. Terlipressin did not reduce the risk of death. This goes to show that blood markers such as creatinine are not necessarily mediators of poor outcomes.

N Engl J Med doi:10.1056/NEJMoa2008290

Alex Nowbar is a clinical research fellow at Imperial College London. Twitter @AlexNowbar