Alex Nowbar reviews the latest research from the top medical journals

NEJM
A rock and a hard place

Vena cava filters are rarely used in clinical practice—and maybe rightly so after the lack of benefit seen in this randomised controlled trial. When a person is simultaneously at high risk of pulmonary embolism yet also bleeding or at high risk of bleeding, there is a dilemma. Anticoagulation has clear benefit for pulmonary embolism, but it worsens bleeding, so the risk of death is high whether you anticoagulate or not. If anticoagulation can truly not be given then it is hard to stand by and do nothing about the risk of pulmonary embolism. That’s where vena cava filters come in. This well designed trial recruited people with severe injuries who have increased risk of pulmonary embolism as well as bleeding. Participants had to be perceived to have a contraindication to anticoagulation for at least 72 hours to warrant randomisation to filter or not. In these people, the rate of death or symptomatic pulmonary embolism was the same in both groups. In the subgroup of participants who could not receive anticoagulation for seven days, there was a signal that filters prevented pulmonary embolism. So there will still be a role for this device in practice, namely when anticoagulation has to be significantly delayed due to bleeding risk. But placement of a vena cava filter is invasive and not without risk in itself. 

Antibiotics for acute COPD exacerbation

We are supposedly “missing” cases of sepsis and yet we are also overusing antibiotics. This randomised controlled trial found that point of care testing of C reactive protein helped reduce antibiotic prescriptions within the four weeks after randomisation compared with “usual care” in patients who had acute exacerbations of chronic obstructive pulmonary disease. An easy win for advocates of C reactive protein testing and good news for the stewardly-minded. Strictly speaking though, to know if the test result made the difference to the outcome, both groups should have had the test and randomisation should have been to whether the clinician and patient were aware of the result or not. On the absence of a test in the control group, the authors said, “This real-world effect needed to be captured because it may affect health care–seeking behavior, which is critical to assessments of overall benefit.” This argument is often made in trials but surely patients and clinicians only care about a test if knowledge of the test result changes the outcome? If the benefit is in the act of performing the test, is that a genuine advance in medicine?

Cluster headache

Goadsby et al performed a double blinded randomised controlled trial of preventive subcutaneous monthly injections of galcanezumab to reduce episodes of cluster headache. There was a clear signal that the drug improved symptoms compared with placebo in weeks 1-3 (which was the primary endpoint) but not in weeks 4-8. Given the severity of this condition, I imagine even short lived benefit might be meaningful to patients, so I expect we will see more work in this area. There was no adverse safety signal other than pain at the injection site, but these numbers are too small and follow-up too short to have confidence in the safety yet.

Annals of Internal Medicine
A dive into durations

Vaughn et al retrospectively assessed the antibiotics prescribed to more than 6000 people in hospital with pneumonia. They calculated the shortest effective duration of antibiotics based on organism and other factors such as clinical status. More than two thirds of people were given longer courses than necessary. Most of this was beyond discharge. This observational study can’t directly alter clinical practice, but there are a few key results of interest. Not having total treatment duration documented in the discharge summary was associated with higher rates of “excess” antibiotics. Those who had excess antibiotics weren’t less likely to die or be readmitted but were more likely to experience adverse effects.

Pregnancy and lupus

Pregnancy outcomes in women with systemic lupus erythematosus have been a challenge for rheumatologists. Mehta et al studied trends in almost 100 000 pregnant women with the condition in the US from 1998 to 2015, and compared them with pregnant women without systemic lupus erythematosus. Maternal deaths decreased over this time period, which is reassuring, and more so in those with systemic lupus erythematosus compared to those without, which suggests something specific is being done that is protective. An alternative explanation would be that milder cases of systemic lupus erythematosus are being identified in more recent years. Caesarean sections have increased in both those with and without systemic lupus erythematosus, but rates were higher in those with systemic lupus erythematosus. Women with systemic lupus erythematosus had longer lengths of stay than pregnant women without systemic lupus erythematosus over the time period. The study used data from a national database, which adds robustness to the findings.

Lancet
Nerve transfer operation reanimates limbs

Sixteen tetraplegic patients had nerve transfer surgery to restore upper limb function. Ten of them had tendon transfer too. The procedure isn’t new, this study just demonstrates its use in a new patient population with cervical spinal cord injury. There was functional improvement but, without a control group, it is unclear what the benefits attributable to the procedure are. This is promising for such a devastating condition as tetraplegia, but for now it’s a case series.

JAMA Internal Medicine
Prognosis opinions

How long would you guess people your age with similar health conditions usually live? Most people weren’t sure and a third thought they had more than 10 years. The people asked were those undergoing haemodialysis in Seattle and Nashville in the US. In reality, only a fifth of US dialysis patients are estimated to live more than 10 years and most die within five years. Troubling figures. I am not surprised the people who took this survey are more optimistic. The overly optimistic group (those who answered more than 10 years to the question) were more likely to want CPR and mechanical ventilation. This sheds a little light on the challenges in advance care planning in end stage kidney disease, in particular the very high spending at end of life. However, this is not evidence of causation (i.e. we don’t know if unrealistic expectations of prognosis cause people to not make advance care planning decisions). A natural conclusion to this study would be to increase prognostic awareness in end stage kidney disease, but there is no reason to assume knowledge of prognosis would change beliefs and behaviours around end of life decision making. 

People in dialysis studies

Patients in trials aren’t the same as those in clinical practice. That old chestnut. Smyth et al compared 80 104 dialysis-dependent patients from 189 randomised controlled trials to dialysis-dependent patients in the US national registry. Trial patients were younger, a greater proportion were male, and they had less diabetes and lower mortality rates. Trial results aren’t perfect data, but they are all we have. We must remember that they might still be applicable to clinical populations that match the trial’s eligibility criteria, and that trials are only designed to answer a specific question in a specific group of people. Most of the time, they aren’t intended to be applicable to all patients with the condition in question. My concern is that people use the difference between trial populations and clinical populations to justify use of non-randomised study data.

Solutions? Careful design of eligibility criteria could help. More thought could be given to trial design to avoid putting off particular groups of people—for example, make provisions for barriers such as timing, costs, mobility, and literacy. There will always be people who decline to consent, but at least everyone eligible should be considered and approached without making assumptions such as that they are too sick so “It wouldn’t be appropriate” or “They’ll say no.”

When treatments bundles go wrong

You may be familiar with the idea of a treatment “bundle” for a common condition such as community acquired pneumonia, kind of like a “care set.” When bundles go wrong, the issue can be the performance or adherence to the bundle components or that the bundle components aren’t right. In this stepped wedge cluster randomised controlled trial of a pneumonia bundle, the component that strikes me as being the reason it didn’t work and even worsened patient outcomes is the use of steroids. This conflicts with recent randomised controlled data on steroids in this context. Harking back to the Smyth et al study, are the different outcomes explained by different patient characteristics? Quite possibly.

The other components were fairly run of the mill: early mobilization, early switch to oral antibiotics, and screening for malnutrition. The stewardly-minded may be glad that early switch to oral antibiotics was included or might wonder why there wasn’t more of a push to shorten the overall antibiotic duration too. People in the bundle group did no better (measured by mortality rate, readmission rate and length of stay) than those in the control group and had more GI bleeding.

I can’t fault the trial design or conduct. Testing a multi-component intervention has many challenges. The take home message is that interventions we believe to be critical may not be as effective as we might think; proper evidence is as important as ever.

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