Trim-Sulfa and Sudden Death

Co-trimoxazole and sudden death in patients receiving inhibitors of renin-angiotensin system: population based study.

Fralick M, Macdonald EM, Gomes T, Antoniou T, Hollands S, Mamdani MM, Juurlink DN; Canadian Drug Safety and Effectiveness Research Network (CDSERN).

PMID: 25359996

And

HARMONIZE

Effect of Sodium Zirconium Cyclosilicate on Potassium Lowering for 28 Days Among Outpatients With Hyperkalemia

Kosiborod M, Rasmussen HS, Lavin P, Qunibi WY, Spinowitz B, Packham D, Roger SD, Yang A, Lerma E, Singh B

PMID: 25402495

TweetChat Dec 2nd

Summary

NephJC returned from the Kidney Week hangover in full force on December second where we dissected two articles on hyperkalemia.

Prior to the discussion Eoin O'Sullivan, winner of our coffee challenge, put together a great introduction on both articles.

If you were fortunate enough to have attended Kidney Week this year, you may have stumbled across any one of a number of posters describing trials of new anti-hyperkalemic agents. There were 5 posters on ZS-9 alone!

The last fortnight has seen a flurry of publications in the field with 3 separate trials of oral potassium binding agents within a week of each other (Sodium Zirconium in JAMA, and NEJM and Patiromer in NEJM) and a potentially related observational trial on the risks of co-trimoxazole in patients on RAAS blockade in the BMJ.

With all that reading to get through, this week’s NEPH JC will be a double whammy. We will look at the HARMONISE trial of ZS-9, and a large study of co-trimoxazole and potential associations.

Trim-Sulfa and Sudden Death

Co-trimoxazole and sudden death in patients receiving inhibitors of renin-angiotensin system: population based study.

Fralick M, Macdonald EM, Gomes T, Antoniou T, Hollands S, Mamdani MM, Juurlink DN; Canadian Drug Safety and Effectiveness Research Network (CDSERN).

PMID: 25359996

The first paper for discussion is a large, Canadian, case control series, by the Canadian Drug Safety and Research Effectiveness Network, published in the BMJ.

The hypothesis is the risk of sudden death in patients on RAAS blockade is higher following administration of specific antibiotics rather than amoxicillin.

The scale of this study is impressive. To answer their question, they searched 17 years of records representing over 1.6 million patients. They identified 39,879 with a label of sudden death and a subsequent group of 1,027 that had a prescription for the target antibiotics in the 7 days prior to dying.

Once the dizzying the scale of the database wears off, I found it interesting to take out a calculator and note that over 17 years, sudden death, in those over 66 years and on RAAS blockade, occurred in 2.5% of patients. Without the data to calculate incident rates per 1000 patient years it’s hard to compare this to other groups, but remember that this is a fairly rare occurrence here, but still represented 1 in 40 deaths.

Before getting to the meat of the results, it’s worth reviewing the case-control groups. 

The first striking feature to me was just what an elderly group of patients this was. A median age of 82, and the majority were female. Cases were notably sicker:

• Twice as likely to have heart failure

• Twice as likely to have renal failure

• More co-morbidities

• More likely to be on a loop

So perhaps, not an ideal control group.

The initial results are striking.

The authors write:

“In the primary analysis, co-trimoxazole was associated with a significantly increased risk of sudden death within seven days relative to amoxicillin (O.R 1.8 C.I 1.5-2.24)”

Ciprofloxacin was associated with a somewhat lower risk of sudden death (1.29 1.02 – 1.62)

I found it strange that norfloxacin, which has similar QT prolonging properties to ciprofloxacin, had had no such risk. Nitrofurantoin actually conferred a lower risk of sudden death than amoxicillin. The smaller number of cases may have contributed to these results.

The absolute risk is not clearly detailed which I always find frustrating as it makes it hard to gain some perspective on relative risks. 

What is expressed, is a risk of sudden death of 1/1,000 per prescription of amoxicillin, compared to 3/1,000 with trimethoprim. Well and good, but what is the risk of sudden death without any antibiotics?

The authors speculate this observed association may be due to trimethoprim’s activity as an ENAC antagonist. This well known phenomena can cause hyperkalemia in patients given prolonged courses or high doses of trim-sulfa.

Well this is easy. So we simply hold our RAAS blockade during certain antibiotics usage, and maybe black label trimethoprim, right?

Not so fast. There are a number of important limitations to consider. With no indication for antibiotics given, it is hard to assume that each antibiotic group had a similar underlying infections (nitrofurantoin and amoxicillin are interchangeable usually only in UTI for example).

The cases and controls had some important differences. 

The relative risk of trimethoprim was reduced when the authors factored in the presence of heart failure. We have no way of knowing what the other co-morbidities in higher disease burden of the cases were ( coronary disease? Previous arrhythmia? ) and these were not accounted for in the statistics. And the higher use of loop diuretics could have a variable affect on trimethoprim. Induce a hypokalaemia and counteract the cotrimoxazole, or increase dehydration and cause AKI?

Making things a little more difficult again to extrapolate to our own patients, only 8.2% of the cases had renal disease, the stage of which was unclassified. 

The authors can only speculate about a possible mechanism involving hyperkalaemia as a no K levels were obtained for any of these patients, nor any ECG to help explain the affect of ciprofloxacin. 

Reported rates of cardiac dysrhythmia in ciprofloxacin are 1 in a million which is disparate to the rates of death suggested here, implying a different mechanism unrelated to the drug usage.

An accompanying editorial wisely cautions against assuming a causal link and being seduced by the plausibility of a pathophysiological mechanism.

Despite these shortcomings, this paper generates some interesting questions.

• Is this common enough to justify prospective controlled trial?

• The authors have previously shown that trimethoprim is a risk factor for hospitalisation due to hyperkalemia, is the plausibility of unrecognised hyperkalemia as a cause of death supported by this trial?

• How can we explain the discordant risks of the different antibiotics results and are the differences between cases and controls enough to bias the results?

• How significant is the lack of detail, such as biochemistry and antibiotic indications? Is it a deal breaker for the working nephrologist?

• Most importantly, how will you incorporate this paper into your practice?

Harmonize

Effect of Sodium Zirconium Cyclosilicate on Potassium Lowering for 28 Days Among Outpatients With Hyperkalemia

Kosiborod M, Rasmussen HS, Lavin P, Qunibi WY, Spinowitz B, Packham D, Roger SD, Yang A, Lerma E, Singh B

PMID: 25402495

ZS-9 is a zirconium silicate, a non-absorbable potassium binding agent. It is an inorganic cation exchanger crystalline with the capacity to bind both potassium and ammonium in the GI tract. Its creators tout its non-absorbable nature as the key to minimising systemic side effects. 

ZS-9’s investors certainly seem encouraged of late.

HARMONISE is a phase 3, multicenter, randomized, double-blind, placebo-controlled trial spanning 44 centres.

Inclusion criteria was simply a serum K of ≥ 5.1 on 2 occasions 5 minutes apart. The important exclusion criteria to be aware of were pseudohyperkalemia, dialysis requirement, life expectancy less than 3 months, pregnancy, cardiac arrhythmias requiring immediate treatment, and diabetic ketoacidosis.

So these were relatively well outpatients; not 4 AM emergency department paic attacks with acute hyperkalemia and ECG changes.

Initially, 258 patients who met eligibility criteria were given ZS-9 10g three times daily. If they achieve normokalaemia within 48 hours, they were then randomised to a placebo, or increasing doses of ZS-9 once daily. The 10g tid dose only failed to induce normokalaemia in 2 patients.

An essential question all nephrologists need to ask when reviewing trials is whether or not their patient cohort is represented in the trial, thus allowing us to extrapolate the findings to them.

Despite the preponderance of white patients, the trialists have otherwise done a good job representing a typical nephrology patient, with an average eGFR around 46 ml/min/1.73m2 range, and accurately representative comorbidity status with high rates of heart failure, diabetes and RAAS blockade.

As mentioned, no dialysis or ESKD patients are represented.

So an encouraging start; these findings may apply to our patients. Did it actually work?

The short answer is yes.  Quickly and predictably. Zs-9 had a reasonable rapid rate of onset, shown nicely in this graph.

Within 2 hours, serum Potassium has dropped by  −0.4 mEq/L (95% CI, −0.5 to−0.4) and was - 1.1 mEq/L by 48hours.

I’m impressed with how tight the confidence intervals are; it's exhibiting a very predictable effect on the serum levels.

Once randomised to the variable dose, long term maintained arms of the study, the data can best be summarised in the  graph below, which demonstrates a brief rise at day 2 as they study participants decreased their dose from TID to once daily, and then a fairly predictable dose dependant response thereafter.

This is all well and good, but if patients can’t tolerate the drug, it’s as good as useless.

The good news is that it seems generally well tolerated, particularly at the lower 5g dose. However the most significant signal in the side effects data was increasing rates of oedema and hypokalemia as the doses increased. At the maximum dose of 15g od, these occurred in 14% and 10% of patients respectively. This is not an insignificant adverse event profile and would at the very least prompt warning patients about a 1/10 chance of oedema repeating serum biochemistry on patients started on 10g or more.

In conclusion then, this is a tidy and well executed phase 3 trial and  Zs-9 has potential to be a relatively well tolerated and predictable treatment option for hyperkalaemia in selected patients. 

The authors quite rightly point out we still have no data beyond 4 weeks, nor have we any meaningful endpoint such as mortality or hospital admissions. It is an encouraging study none the less, and should lead to FDA approval and another tool in our kit.

I wonder what new therapeutic strategies this agent could allow nephrologist to pursue in future? Will ZS-9 it allow some leeway in those patients who we so desperately want to put on an ACEi but are limited by hyperkalemia?

• Would you start ZS-9 in patients with modest hyperkalaemia?

• Do you think they previously would have come to no harm?

• Do you see a role for it in your practice, and if so in whom and at what potassium level?