Assigning causality in AKI – Untying the Gordian Knot?

#NephJC Chat

July 12 9 pm Eastern

July 13 8 pm BST, 12 noon Pacific

J Pharm Pract. 2016 Feb 23. pii: 0897190016628960.

Prospective Comparison of Acute Kidney Injury During Treatment With the Combination of Piperacillin-Tazobactam and Vancomycin Versus the Combination of Cefepime or Meropenem and Vancomycin.

Peyko V, Smalley S, Cohen H.

PMID: 26912532


AKI is a common and important clinical problem affecting about 15% of patients admitted to hospital. It is associated with increased morbidity and mortality, increased lengths of stay and higher health care costs. (Fischer, 2005; Lafrance, 2010). There is strong evidence that AKI is not recognized early (NCEPOD, 2009, PDF link) and NHS Kidney Care estimates that a large proportion of cases of AKI are preventable.

AKI is a term that describes a variety of diseases with heterogeneous aetiologies that lead to renal dysfunction, all of which have different prognostic outcomes. Whilst there are many causes of AKI, one of the major causes in critical illness is sepsis. When a patient presents with sepsis one of the first treatments is broad-spectrum antibiotics. Numerous antibiotics have been implicated in the development of AKI including gentamicin, vancomycin and polymyxins. Piperacillin-tazobactam (pip/tazo) is an anti-pseudomonal agent, whose use has become more over the last decade as initial broad spectrum antibiotic.

Though product literature suggests that like other beta-lactam antibiotics, rates of AKI with pip/tazo are low, several retrospective studies have suggested that rates of AKI may be higher in patients who receive pip/tazo plus vancomycin as opposed to cefepime plus vancomycin. This study aimed to prospectively assess whether there is increased AKI in patients treated with pip/tazo plus vancomycin compared to patients who received another anti-pseudomonal (meropenem or cefepime) and vancomycin.


This single-center, open-label study was run at a busy teaching hospital in Brooklyn, NY. Patients were enrolled after admission to the acute care unit. Patient were treated according to the local sepsis protocol with either pip/tazo + vancomycin, meropenem + vancomycin or cefepime + vancomycin – there was no randomization and clinicians assigned patients to different agents depending on disease state. Patients were identified through the pharmacy medication ordering system and once enrolled, data were extracted daily, including serum creatinine, demographic data, medications that may affect serum creatinine, diagnosis and use of contrast media.

Patients were followed until they developed AKI (staged using the KDIGO criteria), antibiotics were discontinued, or when the patient was discharged from hospital. The authors suggest that 120 patients per group would be needed to have 80% power to detect a difference between the groups. This was based on ‘studies’ showing a 20% difference between rates of AKI in patients who received tazocin/vancomycin or alternative anti-pseudomonals and vancomycin. Surprisingly these studies aren’t referenced and details of this ‘power calculation’ are not really discussed.


The study reviewed 210 patients and final numbers included 59 who received pip/tazo + vancomycin and 26 who received meropenem + vancomycin or cefepime + vancomycin.


Baseline demographic details were well matched, but there were differences in admitting diagnosis with patients treated with cefepime or meropenem more likely to have either a UTI or an intra-abdominal infection. Importantly, there were no differences in trough levels of vancomycin between the groups, with mean trough levels being within target range.

So, the result! The study reported development of AKI was observed in 37.3%  of patients receiving the combination of piperacillin-tazobactam and vancomycin vs only 7.7%  of patients receiving the combination of either cefepime or meropenem and vancomycin (χ2=7.80, P=0.005).


Whilst this is an interesting study and an important question, care must be taken in interpreting the results. On the face of it the study suggests that AKI is more likely to develop in patients treated with pip/tazo + vancomycin than in patients who receive meropenem + vancomycin or cefepime + vancomycin. There are however a number of limitations that mean this conclusion cannot be drawn from the results of this study:

  • The study does not account for, or report, disease (infection) severity and this information is crucial to understanding whether development of AKI was from the choice of antibiotics and not more severe infection/sepsis.
  • Whilst all the patients in the study had a defined infection, there were differences in the source of infection between groups, so we are not really able to definitively compare like-for-like.
  • The study design was inherently bias as physicians were permitted to dictate antibiotic therapy from a guideline that permitted them this choice. Given that twice as much pip/tazo was prescribed than cefepime or meropenem the influence of selection bias on these results cannot be excluded.
  • The trial does not report the severity of AKI, or look at any outcomes beyond discharge.
  • The numbers are small and significantly below the numbers that the authors suggested would be needed to confidently suggest a difference between the groups (even though details of the power calculation were lacking).

Although the observed effect appears to be quite large, a causal relationship between treatment with pip/tazo + vancomycin and development of AKI cannot be drawn from this study. The authors acknowledge this and we must acknowledge that the study did not seek to show causation, clearly stating it’s hypothesis was ‘to determine whether or not there is an increased incidence of AKI in patients receiving the combination of piperacillin-tazobactam and vancomycin versus the combination of other anti-pseudomonal beta-lactams.’

These results are consistent with previously published retrospective studies and additionally suggest that when used in combination with anti-pseudomonal drugs, it may not be vancomycin that is responsible development of renal dysfunction as trough levels were not significantly different between the groups. This finding is, of course, subject to the same bias as already discussed.

To define a causal relationship between Tazocin and AKI a significantly different trial design would be required, perhaps one that would randomly allocate patients to treatment – ethics for which would be tricky (for a number of reasons). Defining appropriate end-points would also be key, as ‘development of AKI’ isn’t sensitive enough to pick up harm associated with antibiotic use. Endpoints including severity/stage of AKI, follow-up beyond discharge, as well as collecting acute and long-term sequelae related to AKI, including mortality would have been nice in this study and would be essential in future studies. Studies that have looked at relationships between gentamicin and AKI have done so in studies conducted in peri-operative patients who receive prophylactic antibiotics. Again, based on the quality of the available observational data, you have to question the ethics of carrying out such studies with pip/tazo given our responsibilities as antibiotic stewards. The development of multi-drug resistant organisms is a terrifying threat to us all and significantly more high quality observational research is needed before such trials could even be considered.

Whilst the authors raise an important point and highlight the need for investigation, the flaws in the design and methodology of this study (which to their credit they acknowledge) mean the results should be viewed as hypothesis generating only. The nephrology and microbiology communities need to put their heads together and see if definitive trials are required and if so a great deal of thought and planning will be needed to unpick the role pip/tazo + vancomycin may play in the development of ‘AKI’ in patients who are already acutely unwell and at risk of AKI. Though not impossible, it will be like trying to unravel a knot of knots!

Summary by Matt Graham-Brown with contributions from Micro/ID Doctor George Trafford (@george_trafford)