Cheesecake files: A stadium full

As we’ve been enjoying the World Cup and the Commonwealth Games my latest cheesecake appeared in print online. The topic once more is Kidney Attack biomarkers – those pesky little proteins in the urine that appear when your kidney is injured.  This time I have been getting stuck into some math (sorry) to try and understand what it is that affects when these biomarkers appear in the urine after injury.  I call this a biomarker time-course.  A “Pee Profile” may be a better term but it would never get past the editor.  What I care about is whether the type of biomarker and/or extent of injury, affects the pee profiles.

There are three basic types of biomarkers.  First are those that are filtered from the blood by the two million odd filters in the kidney.  Often they are then reabsorbed back into the blood in the little tubules where the pee is produced – that is, they don’t appear in the urine.  Think of it like a stadium with many entrances.  People (biomarkers) come in and sit down (are reabsorbed).  If, though, a section of the stadium has been fenced off because of broken seating from the previous game (the injury), then some of those entering the stadium may end up exiting it again (the pee biomarkers).  The numbers being reabsorbed and exiting will also depend on whether all the entrances are open – if some are closed then this will have a flow on affect on the rate of people leaving the stadium.

The second are preformed biomarkers.  If we change the analogy slightly, imagine these as people already in the stadium (if the analogy was accurate they would have been born there!).  If some terrible injury happens (like the 4th, 5th, 6th and 7th goals of a now famous football match) some of those people would get up and exit quickly.  The overall rate of exit would reflect on the extent of the injury.

The third, are induced biomarkers.  These are ones that don’t already exist, but are produced in response to an “injury.”  Instead of being biomarkers, let us think of the spectators as produces of these biomarkers and let noise be the biomarker.  There is some background noise of course, but when an “injury” (goal, gold medal performance etc) occurs there is a sudden increase in noise which slowly dies down.  Depending on the team and the number of supporters this will be softer or loader and will carry on for shorter or longer periods (Goooooooooooaaaaaaaaaaaa……lllllllllllll).

The upshot of it all were many coloured graphs and a step towards understanding how we may better make use of the various types of novel biomarkers of kidney injury that have been recently discovered.

PlosOneFigs

_____________

Pickering, J. W., & Endre, Z. H. (2014). Acute kidney injury urinary biomarker time-courses. PloS One. doi:10.1371/journal.pone.0101288

 

 

A new entity is born: CDaR

Have you ever been told the blood test is positive and the disease in question is shocking – Cancer, an STD (but you don’t sleep around!), MS?  Have you every wondered why it is that some drugs get withdrawn years after, and millions of prescriptions after, they were first approved?  Surely, you’ve read a headline that coffee is good for you and chocolate bad, or was that chocolate good and coffee bad or were they both good, or both bad? Probably you’ve read all those headlines.  What does it all mean?  Am I sick or not (I heard some tests falsely give positive results)? Does it matter if I’ve been taking that drug or drinking three cups a day?  The answer to all those questions depends on one thing – clinical data research.  That is, it depends on how we collect the numbers, and what story those numbers are telling us.  Today, I am thrilled to announce that I have had my department’s (Department of Medicine, University of Otago Christchurch) endorsement to establish a new group, Clinical Data Research (CDaR), which will focus on the stories numbers in medicine tell us.

Source: Pickering et al http://ccforum.com/content/17/1/R7

Source: Pickering et al http://ccforum.com/content/17/1/R7

My recent expertise, as readers of this blog may have picked up, is in Kidney Attack (or Acute Kidney Injury). My contribution, as someone with a physics background, has been in data analysis and mathematical modeling.  It has been a privilege to have been involved with many discoveries and helping bring to light the stories of the biomarkers of that disease and the results of a unique randomised controlled trial.  Kidney Attack is a notoriously difficult to detect, and, partly because of that, one that has no effective treatment.  I’m currently working on the story of the association of Kidney Attack with death following surgery with cardiopulmonary bypass.  I am now looking to take those skills and work with other researchers in other medical specialties who generate data and are looking to tell its story (although I will still work on the kidney data!).  I’m particularly keen to engage with more students and pass on some of the data analysis skills I have acquired.  Moves towards open data as well as collection of data in large databases is providing more opportunities to assess the efficacy of health interventions and detect disease risk factors. The prospect of personalised medicine is one of both hope and hype. To sort fact from fantasy in all these areas will require development of new analytical techniques and careful assessment of evidence. This is what I wish to devote the rest of my career to, and to inspire others along the way. John Ioannidis, a highly respected biostatistician, once wrote an essay entitled “Why most research findings are false”  It is a scary thought that many interventions and diagnostic techniques in medicine may be based on biased studies (usually inadvertently biased!). More data will help reduce the bias, if it is treated nicely.  I promise to do my best to treat my data nicely, after all it is your and my health that is at stake.

I posted a few weeks ago my ten commandments of a data culture.  This is the ethos of CDaR.  Below is the lay summary of the new entity.

Group Name:            Clinical Data Research (CDaR)

Department:             Department of Medicine

Institution:                University of Otago Christchurch

Aim: To provide transparent evidence, with the lowest possible risk of bias, of the utility of biomarkers and efficacy of treatments in health or disease.

Lay summary of our aim: We aim to save lives and reduce the burden of disease by applying new ways to collect and analyse clinical data to better diagnose diseases, to predict the course and outcomes of diseases, and to assess how well treatments work.  We do this because we all want the best possible health outcomes for our communities, our families, and ourselves, with the least possible harm done along the way.  We are excited by the new ways scientists, including those at the University of Otago Christchurch, have come up with to measure disease, disease risk, and treatment outcomes. We are also living in an age of unprecedented data generation. To discover both benefits and harm in all this data and to make those discoveries available to all those making clinical decisions requires people dedicated to analysing this data in a transparent and open fashion that exposes both the good and the bad. That is who we want to be and who we want our students to become.

Definition:  A biomarker is any measureable quantity related to disease risk or diagnosis, or disease or health outcomes.

Nelson Mandela is on dialysis

CNN is reporting Nelson Mandela is on dialysis. http://t.co/HZTIlmGrtO.  This means he is suffering from Acute Kidney Injury, the disease I study.  Having to have dialysis is very serious. Unfortunately, survival rates are only about 50% by this stage, less in the very elderly.  Dialysis is not a treatment, merely a support for the kidney to try and give them time to recover  function on their own and  a means to remove toxins from the body.

 

Cheesecake files: Too little pee

This week’s post is really about the coloured stuff & why too little of it is dangerous.  Note, I say coloured stuff because it aint just yellow – check out this herald article if you don’t believe me (or just admire this beautiful photo).

 A rainbow of urine from a hospital lab. Credit:  laboratory scientist Heather West.

A rainbow of urine from a hospital lab.
Credit: laboratory scientist Heather West.

Story time

A long time ago, when Greeks wore togas, and not because they couldn’t afford shirts, a chap named Galen* noted that if you didn’t pee you’re in big trouble.  It took 1800 more years before the nephrologists and critical care physicians got together to try and decide just how much pee was too little.  This was at some exotic location in 2003 where these medics sat around for a few days talking and drinking (I’m guessing at the latter, but I have good reason to believe…) until they came up with the first consensus definition for Kidney Attack (then called Acute Renal Failure, now called Acute Kidney Injury)1.  It was a brilliant start and has revolutionised our understanding of just how prevalent Kidney Attack is.  It was, though, a consensus rather than strictly evidence based (that is not to say people didn’t have some evidence for their opinions, but the evidence was not based on systematic scientific discovery).  Since then various research has built up the evidence for or against the definitions they came up with (including some of mine which pointed out a mathematical error2 and the failings of a recommendation of what to do when you don’t have information about the patient before they enter hospital3).  One way they came up with to define Kidney Attack was to define it as too little pee.  Too little pee was defined as a urine flow rate of less than half a millilitre per kiliogram of body weight per hour over six hours (< 0.5ml/kg/h over 6h).  Our groups latest contribution to the literature shows that this is too liberal a definition.

The story of our research is that as part of a PhD program Dr Azrina Md Ralib (an anaesthesist from Malaysia) conduct an audit of pee of all patients entering Christchurch’s ICU for a year.  She did an absolutely fantastic job because this meant collecting information on how much every patient peed for every hour during the first 48 hours as well as lots of demographic data etc etc etc. Probably 60-80,000 data points in all!  She then began to analyse the data.  We decided to compare the urine output data against  meaningful clinical outcomes – namely death or need for emergency dialysis.  We discovered that if patients had a flow rate of between 0.3 to 0.5 ml/kg/h for six hours it made no difference to the rates of death or dialysis compared to those with a flow rate greater than 0.5.  Less than 0.3, though, was associated with greater mortality (see figure).  For the clinician this means they can relax a little if the urine output is at 0.4 ml/kg/h.  Importantly, they may not give as much fluid to patients. Given that in recent times a phenomenon called “fluid overload” has been associated with poor outcomes, this is good news.

The full paper can be read for free here.

Proportion of mortality or dialysis in each group. Error bars represent 95% confidence intervals.From Ralib et al Crit Care 2012.

Proportion of mortality or dialysis in each group. Error bars represent 95% confidence intervals.From Ralib et al Crit Care 2013.

———————————————————

*Galen 131-201 CE.  He came up with one of the best quotes ever: “All who drink of this remedy recover in a short time, except those whom it does not help, who all die.”

1.     Bellomo R, Ronco C, Kellum JA, Mehta RL, Palevsky PM, Acute Dialysis Quality Initiative workgroup. Acute renal failure – definition, outcome measures, animal models, fluid therapy and information technology needs: the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Crit Care 2004;8(4):R204–12.

2.     Pickering JW, Endre ZH. GFR shot by RIFLE: errors in staging acute kidney injury. Lancet 2009;373(9672):1318–9.

3.     Pickering JW, Endre ZH. Back-calculating baseline creatinine with MDRD misclassifies acute kidney injury in the intensive care unit. Clin J Am Soc Nephro 2010;5(7):1165–73.

Cheesecake files: Injury, function, and death

When they say your tests are positive for a disease just what do they mean?  If it is a simple blood or urine test often they mean that the concentration measured is outside (above or below) some  reference range.  In my field of Kidney Attack (a.k.a. acute kidney injury: AKI) two tests of the same substance (plasma/serum creatinine) are needed a day or two apart . The difference in the concentrations is what is important.  If the creatinine concentration has increased by >0.3 mg/dl within 48 hours or by more than 50% within a week then the diagnosis of AKI is made.  What happens, though, when someone comes along with a new test?  How do we know it is any better (or worse) than the original test? In my view what is required is that both the old and the new tests should be compared to a third, clinically relevant, variable.  For example, a new prostate cancer test may be compared to the present (poor) PSA test  by referencing both to the more definitive biopsy results.

In AKI the reason the creatinine threshold of 0.3 mg/dl was included as diagnostic was because research(1) had shown this level of increase to be associated with a four fold increase in the likelihood of premature death. If you’ve seen any of my previous posts on my research you will know that I am interested in new biomarkers (plasma and urine proteins mainly) that could be used to diagnose AKI earlier than creatinine.  While creatinine is a marker of changes in kidney filtration function, most of these new biomarkers reflect structural injury itself.  An analogy is that movement of a finger hurt in a rugby tackle tells us if the finger is functioning, whereas an x-ray is needed to tell us if it is broken or not.

Sam Whitelock damaged his finger during a game.  It had enough function to let him continue to play.  X-rays later showed it was broken. Picture: TV3

Sam Whitelock damaged his finger during a game. It had enough function to let him continue to play. X-rays later showed it was broken.
Picture: TV3

 

My latest publication(2) describes a method to determine appropriate biomarker thresholds.  It is quite simple.  First, I determine the sensitivity of the creatinine threshold to predict a meaningful clinical outcome – the need for dialysis or death within 30 days. The sensitivity is simply the proportion of all those who end up having the outcome who had a measure above the threshold.  I then take that sensitivity and work out what the biomarker threshold needs to be in order to yield that same sensitivity.

An early sketch of mine as I worked out how to determine structural biomarker thresholds

An early sketch of mine as I worked out how to determine structural biomarker thresholds

(1) Chertow, G. M., Burdick, E., Honour, M., Bonventre, J. V., & Bates, D. (2005). Acute kidney injury, mortality, length of stay, and costs in hospitalized patients. Journal of the American Society of Nephrology : JASN, 16, 3365–3370. doi:10.1681/ASN.2004090740

(2) Pickering, J. W., & Endre, Z. H. (2013). Linking Injury to Outcome in Acute Kidney Injury: A Matter of Sensitivity. PloS one, 8(4), e62691. doi:10.1371/journal.pone.0062691.t001

H7N9 kills and attacks kidneys

27% of patients with H7N9 Influenza A died.  This is the finding of a report just released  in the New England Journal of Medicine is a study of 111 of the 132 confirmed cases of H7N9 Influenza A*.

Acute Kidney Injury or “Kidney Attack” was amongst the most common complications.

Of the 111 patients we evaluated, 85 (76.6%) were admitted to an intensive care unit (ICU); of these patients, 54 were directly admitted to the ICU, and 31 were admitted during hospitalization. Moderate-to-severe ARDS [Acute Respiratory Disease Syndrome] was the most common complication (in 79 patients), followed by shock (in 29 patients), acute kidney injury (in 18 patients), and rhabdomyolysis (in 11 patients).

In an analysis in the Appendix to the paper a comparison was made between the 30 patients who had died and 49 who had recovered (others were still in hospital).  100% of those who died had had ARDS compared with 40% of those who recovered.  One third of those who died had Acute Kidney Injury compared with 4% of those who recovered.  From a statistical perspective these numbers illustrate a real difference with a low probability (~ 1-2 out of 1000) of observing such a difference by chance.**

Note, all patients had been in close contact withe live chickens or pigeons within 2 weeks of hospitalisaton.

NEJM 23 May 2013

NEJM 23 May 2013

* Clinical Findings in 111 Cases of Influenza A (H7N9) Virus Infection

Hai-Nv Gao, M.D., Hong-Zhou Lu, M.D., Ph.D., Bin Cao, M.D., Bin Du, M.D., Hong Shang, M.D., Jian-He Gan, M.D., Shui-Hua Lu, M.D., Yi-Da Yang, M.D., Qiang Fang, M.D., Yin-Zhong Shen, M.D., Xiu-Ming Xi, M.D., Qin Gu, M.D., Xian-Mei Zhou, M.D., Hong-Ping Qu, M.D., Zheng Yan, M.D., Fang-Ming Li, M.D., Wei Zhao, M.D., Zhan-Cheng Gao, M.D., Guang-Fa Wang, M.D., Ling-Xiang Ruan, M.D., Wei-Hong Wang, M.D., Jun Ye, M.D., Hui-Fang Cao, M.D., Xing-Wang Li, M.D., Wen-Hong Zhang, M.D., Xu-Chen Fang, M.D., Jian He, M.D., Wei-Feng Liang, M.D., Juan Xie, M.D., Mei Zeng, M.D., Xian-Zheng Wu, M.D., Jun Li, M.D., Qi Xia, M.D., Zhao-Chen Jin, M.D., Qi Chen, M.D., Chao Tang, M.D., Zhi-Yong Zhang, M.D., Bao-Min Hou, M.D., Zhi-Xian Feng, M.D., Ji-Fang Sheng, M.D., Nan-Shan Zhong, M.D., and Lan-Juan Li, M.D.New England Journal of Medicine Online May 22, 2013 DOI: 10.1056/NEJMoa1305584

** something called a multivariate analysis was attempted which trys to take into account correlations between diseases to see which diseases are the major factors.  However, with “only” 30 deaths such an analysis is very limited and I do not think of value in this situation.

Two new Health Research Council grants worth crowing about

This week’s announcement by the HRC of Feasibility Study and Emerging Researcher grants have many great projects.  Two in particular are worth crowing about (because they have some relationship to kidneys and they involve two excellent people).  I have put summaries in their own words below, but first my comments.

Dr Palmer (Department of Medicine, University of Otago Christchurch), who has appeared on this blog site before, conducts what in the trade are called “meta-analyses” and “systematic reviews.”  Simply put, these are methods to extract the best possible evidence from all the studies that have been done for the effectiveness of a treatment.  Just as one person may toss a coin 4 times in a row and get 4 heads, so too can any one trial give a mistaken impression that a treatment is efficacious (or not) when it really isn’t (or is).  By pooling together many treatments Suetonia provides the very best quality evidence available.  Given that Chronic Kidney Disease affects a large and growing proportion of us, knowing which treatments have the best outcomes is of national significance, not merely to our health but also to the national budget.  A particular problem is that after a trial it can be many many years until meaningful health outcomes are know (e.g. if the treatment delays dialysis need or reduces mortality).  Suetonia’s study will assess the effectiveness of surrogate endpoints for clinical trials.  Surrogate endpoints, such as plasma creatinine which I’ve discussed many time in this blog, are physiologically related to the functioning of an organ or to a disease state as well as statistically associated with future hard outcomes.  However, their use in trials is limited by how well they are associated and how they are used.  I look forward to finding out what Suetonia discovers.

Mrs Rachael Parke (Auckland DHB) is an experienced nurse undertaking a PhD. Ensuring patients have adequate fluids on board is particularly crucial to the kidneys and other organs. Obviously with surgery any blood loss needs to be compensated for. However, there are also physiological changes in where fluid is distributed throughout the body.  Cardiopulmonary bypass, used in cardiac surgery, is a particular risk factor for Acute Kidney Injury. In the past the practice has been to give large amounts of fluid in order to ensure adequate fluid is given.  However, recent research has shown that too much fluid can have a negative impact (increased mortality).  A more restrictive fluid regime may have very meaningful outcomes.  Rachael is investigating, in a randomised controlled trial, if restricting fluid improves outcomes.  The outcome she is most interested in is how long patients stay in the hospital.  This is a very practical outcome for both patient and budget.  I am particularly pleased that this study is nurse-led.  Nurses play an incredibly important role in research as well as patient management.

In their own words:

Dr Suetonia Palmer: Making better clinical decisions to prevent kidney disease

More than ten percent of adults will develop chronic kidney disease. The effectiveness of many treatments used to improve outcomes in kidney disease is tested against surrogate (indirect) markers of health (e.g., cholesterol levels or blood pressure).

Unexpectedly, subsequent systematic analysis has identified little evidence to show that treatment strategies based on these surrogate markers translate to improved health for patients. Serum creatinine and proteinuria levels are commonly-used markers of kidney function to guide treatment.

The research involves using systematic review methods to summarise the quality of evidence for using proteinuria and serum creatinine as markers of treatment effectiveness in clinical trials. It will be determined whether using these markers to guide clinical care improves patient health or, conversely, leads to treatment-related harm or excessive use of ineffective medication.

These summaries will help clinicians and patients make better shared decisions about which therapeutic strategies actually improve clinical outcomes in kidney disease.

Mrs Rachel Parke: Fluid therapy after cardiac surgery – A feasibility study

Following cardiac surgery, patients receive large amounts of fluid in the intensive care unit. This may cause problems with wound healing and delay hospital discharge. A planned randomised controlled trial of a restrictive fluid regime as compared to a more liberal approach utilising advance hemodynamic monitoring, aims to reduce the amount of fluid patients receive and reduce hospital length of stay. This feasibility study aims to determine whether this nurse-led protocol is practicable and feasible and will help answer the research question. This study is simple and inexpensive and if it demonstrates a decreased length of hospital stay then this will represent a significant benefit for both individual patients and the health system.