**4.1 Epidemiology**

Most epidemiological studies are retrospective and face the difficulty of missing data, particularly baseline creatinine data. This was discussed in section 3.4. It is important to quantify the severity of AKI, as more severe AKI is likely to have greater long term impact on health resources. Where possible, data on the duration of AKI as well as severity should be captured as this is an independent predictor of outcome. Three areas of epidemiology lack data. First, there are comparatively few good epidemiology studies in countries other than in Europe, North America or Australasia (Cerda et al. (2008)). The incidence of AKI in countries with large populations such as China, India, Indonesia, Nigeria has significance beyond their own borders. Some countries have numbers of particular AKI aetiologies which, if well studied, could provide useful data world-wide. For example, with anti-retroviral therapy

where *Z<sup>α</sup>* is the Z coefficient for a Type I error and *Z<sup>β</sup>* for a Type II error. In order to detect a true AUC of 0.7 with p<0.05 (*Z<sup>α</sup>* = 1.96 (two sided)) and at 80% power (*Z<sup>β</sup>* = 0.84) in a population with a 30% incidence 212 participants would be needed. A population with a much lower incidence, say 5%, would need a much greater sample (*N* = 1269). If we were comparing a known biomarker with an AUC of 0.7 and wanting to know if another biomarker was better (AUC of 0.8 or more), then at p<0.05, 80% power and 30% incidence we would need 676 participants. This highlights the importance of knowing *a priori* the AKI incidence, what

The Metamorphosis of Acute Renal Failure to Acute Kidney Injury 141

In larger studies it is possible to assess the biomarker's ability to predict premature death or need for renal replacement therapy. In both cases, it is important to assess the risk relative to known risk factors. A useful technique is to use logistic regression models and compare a model of known risk factors with one with the same risk factors plus the biomarker. The integrated discrimination improvement (IDI) is a useful technique which, along with the AUC, allows for analysis of biomarker performance across the whole range of biomarker

Studies in heterogeneous populations with multiple aetiologies and timing of injury are particularly difficult. Given the recent findings with respect to biomarker dependency on pre-existing renal function it is important that study size is sufficient to allow for cohort analysis, in particular cohorts of CKD and sepsis. In studies with CKD patients it is important to measure urinary albumin. Albumin is a possible AKI biomarker in its own right, but also competes for reabsorption in the proximal tubules with cystatin C and NGAL, IL-18 and

The quality of reporting of biomarker studies varies widely. The Standards of Reporting of Diagnostic Accuracy (STARD) statement provides minimum standards and is worth referring to in the planning of the study (Bossuyt et al. (2003)). Biomarker concentrations should almost always be presented as median with inter-quartile range as they are usually non-normal distributions. With the jury still out on "normalising" urinary biomarkers to urinary creatinine we have recommended that both normalised and non-normalised results be presented.

Clinical trial design follows one of three paradigms: (i) Prevention, namely treatment prior to a procedure which may cause injury; (ii) Early intervention following following a raised biomarker, but prior to observation of functional change (eg Endre et al. (2010)); (iii) Late intervention following an elevation of creatinine (Pickering & Endre (2009c); Pickering et al. (2011)). Whilst it may be anticipated that if successful prevention or early intervention treatments are developed there will be less need for late intervention treatments, late intervention treatments will still be needed either where prevention or early intervention fails, or where it is not possible to administer it. Injury biomarkers will play a role in all three paradigms. Some are being identified as risk factors, prior to a procedure (Bennett et al. (2008)), whilst others will be elevated early enough after injury to allow early intervention, finally others will serve as outcome variables. As with plasma creatinine as an outcome variable, analogous to the RAVC, it is likely that a continuous measurement of the biomarker

over a pre-specified duration will serve best as an outcome variable.

a clinically relevant AUC would be, and choosing a sample size appropriately.

concentrations (Pencina et al. (2008)).

possibly other markers (Nejat et al (2011a)).

**4.3 Clinical trials**

reducing mortality in HIV patients, the incidence of AKI in this population may be increasing (Lopes et al. (2011)). In South-East Asia, AKI is a prominent complication of paraquat (a contact herbicide) self-poisoning (Roberts et al. (2011)). In both these cases epidemiology in countries with relevant populations would help identify the extent of risk in other countries where incidence of HIV or paraquat poisoning is relatively low. Second, there are few studies that have investigated AKI on CKD (Ali et al. (2007)). The growing world wide epidemic of CKD demands epidemiology that quantifies the additional risk CKD has for AKI and the effects of AKI on the progression of CKD. Third, there are few studies on CKD induced by AKI that have comparable control groups of hospitalised non-AKI patients (Coca et al. (2009)).

#### **4.2 Biomarker studies**

It is important to report the timing of sampling for biomarker measurements in relation to putative onset of injury. As we have seen, different biomarkers are likely to have different time profiles. A weakness of many studies is that samples are taken only at one or two time points, and even then they may be outside the temporal "window of opportunity" of the biomarker, possibly rendering the study results misleading. Temporal profiles are available for many biomarkers following the time of injury which can be used to plan sampling (eg Endre et al. (2011)) . For novel biomarkers, frequent sampling will be necessary to establish their time course. The first 12 hours following injury are crucial as this is the window for early intervention. There is also a need to discover and assess injury biomarkers with slightly longer (∼24 to 36-h) time courses as samples may not be able to be taken earlier in many critically ill patients. Very few studies have measured biomarker profiles beyond a two or three days, yet there is tantilising evidence that some biomarkers (eg KIM-1) may become elevated during a repair phase which in some patients may not begin until several days following injury.

The choice of outcome variable is particularly important. Despite the caveats discussed with respect to using creatinine changes as a surrogate for change in GFR and the issue of comparing injury to function, we are still in a period where a functional definition of AKI is likely to be the outcome of choice for biomarker studies. Use of non-standard AKI definitions are unhelpful and should be avoided in preference to the AKIN, RIFLE and, potentially, new KDIGO AKI definitions. The duration as well as severity of creatinine increase is also of importance. The AKIN definition requires only one time point at which plasma creatinine may be elevated. This allows for mild and transient changes in function to equally be recorded as AKI. We have shown that transient increases in creatinine are associated with injury biomarkers (Nejat et al (2011b)), although not to the extent of more sustained increases. We recommend evaluating biomarkers in relation to duration of increase of creatinine as well as the increase itself.

Most biomarker studies are underpowered. The number of participants are rarely calculated *a priori*. Typically we want to ascertain if a biomarker is clinically useful. At a minimum we would want an AUC of 0.7, more likely 0.85 or greater. Given an expected incidence (proportion), *I*, in the population, how many participants (*N*) will we need in the study? Hanley and McNeil provide a general equation which we may adapt as we are interested in the difference between the true AUC and 0.5 (Hanley & McNeil (1982)):

$$N = \frac{1}{I} \left[ \frac{0.5773Z\_{\text{\tiny{\tiny{\tiny{\tiny{\tiny{\tiny{\tiny{\tiny{\tiny{\text{H}\_{\text{II}}}}}}}}} + Z\_{\text{\tiny{\beta}}}\sqrt{0.1667 + \frac{A\text{U}\text{C}}{2-A\text{I}\text{I}\text{C}} + \frac{2A\text{U}\text{C}^2}{1+A\text{I}\text{I}\text{I}\text{C}^2} - 2A\text{I}\text{I}\text{C}^2}}{A\text{I}\text{I}\text{C} - 0.5} \right]^2\tag{12}$$

where *Z<sup>α</sup>* is the Z coefficient for a Type I error and *Z<sup>β</sup>* for a Type II error. In order to detect a true AUC of 0.7 with p<0.05 (*Z<sup>α</sup>* = 1.96 (two sided)) and at 80% power (*Z<sup>β</sup>* = 0.84) in a population with a 30% incidence 212 participants would be needed. A population with a much lower incidence, say 5%, would need a much greater sample (*N* = 1269). If we were comparing a known biomarker with an AUC of 0.7 and wanting to know if another biomarker was better (AUC of 0.8 or more), then at p<0.05, 80% power and 30% incidence we would need 676 participants. This highlights the importance of knowing *a priori* the AKI incidence, what a clinically relevant AUC would be, and choosing a sample size appropriately.

In larger studies it is possible to assess the biomarker's ability to predict premature death or need for renal replacement therapy. In both cases, it is important to assess the risk relative to known risk factors. A useful technique is to use logistic regression models and compare a model of known risk factors with one with the same risk factors plus the biomarker. The integrated discrimination improvement (IDI) is a useful technique which, along with the AUC, allows for analysis of biomarker performance across the whole range of biomarker concentrations (Pencina et al. (2008)).

Studies in heterogeneous populations with multiple aetiologies and timing of injury are particularly difficult. Given the recent findings with respect to biomarker dependency on pre-existing renal function it is important that study size is sufficient to allow for cohort analysis, in particular cohorts of CKD and sepsis. In studies with CKD patients it is important to measure urinary albumin. Albumin is a possible AKI biomarker in its own right, but also competes for reabsorption in the proximal tubules with cystatin C and NGAL, IL-18 and possibly other markers (Nejat et al (2011a)).

The quality of reporting of biomarker studies varies widely. The Standards of Reporting of Diagnostic Accuracy (STARD) statement provides minimum standards and is worth referring to in the planning of the study (Bossuyt et al. (2003)). Biomarker concentrations should almost always be presented as median with inter-quartile range as they are usually non-normal distributions. With the jury still out on "normalising" urinary biomarkers to urinary creatinine we have recommended that both normalised and non-normalised results be presented.

#### **4.3 Clinical trials**

16 Will-be-set-by-IN-TECH

reducing mortality in HIV patients, the incidence of AKI in this population may be increasing (Lopes et al. (2011)). In South-East Asia, AKI is a prominent complication of paraquat (a contact herbicide) self-poisoning (Roberts et al. (2011)). In both these cases epidemiology in countries with relevant populations would help identify the extent of risk in other countries where incidence of HIV or paraquat poisoning is relatively low. Second, there are few studies that have investigated AKI on CKD (Ali et al. (2007)). The growing world wide epidemic of CKD demands epidemiology that quantifies the additional risk CKD has for AKI and the effects of AKI on the progression of CKD. Third, there are few studies on CKD induced by AKI that have comparable control groups of hospitalised non-AKI patients (Coca et al. (2009)).

It is important to report the timing of sampling for biomarker measurements in relation to putative onset of injury. As we have seen, different biomarkers are likely to have different time profiles. A weakness of many studies is that samples are taken only at one or two time points, and even then they may be outside the temporal "window of opportunity" of the biomarker, possibly rendering the study results misleading. Temporal profiles are available for many biomarkers following the time of injury which can be used to plan sampling (eg Endre et al. (2011)) . For novel biomarkers, frequent sampling will be necessary to establish their time course. The first 12 hours following injury are crucial as this is the window for early intervention. There is also a need to discover and assess injury biomarkers with slightly longer (∼24 to 36-h) time courses as samples may not be able to be taken earlier in many critically ill patients. Very few studies have measured biomarker profiles beyond a two or three days, yet there is tantilising evidence that some biomarkers (eg KIM-1) may become elevated during a repair phase which in some patients may not begin until several days following injury. The choice of outcome variable is particularly important. Despite the caveats discussed with respect to using creatinine changes as a surrogate for change in GFR and the issue of comparing injury to function, we are still in a period where a functional definition of AKI is likely to be the outcome of choice for biomarker studies. Use of non-standard AKI definitions are unhelpful and should be avoided in preference to the AKIN, RIFLE and, potentially, new KDIGO AKI definitions. The duration as well as severity of creatinine increase is also of importance. The AKIN definition requires only one time point at which plasma creatinine may be elevated. This allows for mild and transient changes in function to equally be recorded as AKI. We have shown that transient increases in creatinine are associated with injury biomarkers (Nejat et al (2011b)), although not to the extent of more sustained increases. We recommend evaluating biomarkers in relation to duration of increase of creatinine as well

Most biomarker studies are underpowered. The number of participants are rarely calculated *a priori*. Typically we want to ascertain if a biomarker is clinically useful. At a minimum we would want an AUC of 0.7, more likely 0.85 or greater. Given an expected incidence (proportion), *I*, in the population, how many participants (*N*) will we need in the study? Hanley and McNeil provide a general equation which we may adapt as we are interested in

0.1667 + *AUC*

*AUC* − 0.5

<sup>2</sup>−*AUC* <sup>+</sup> <sup>2</sup>*AUC*<sup>2</sup>

<sup>1</sup>+*AUC* − <sup>2</sup>*AUC*<sup>2</sup>

2

(12)

the difference between the true AUC and 0.5 (Hanley & McNeil (1982)):

0.5773*Z<sup>α</sup>* + *Z<sup>β</sup>*

**4.2 Biomarker studies**

as the increase itself.

*<sup>N</sup>* <sup>=</sup> <sup>1</sup> *I*  Clinical trial design follows one of three paradigms: (i) Prevention, namely treatment prior to a procedure which may cause injury; (ii) Early intervention following following a raised biomarker, but prior to observation of functional change (eg Endre et al. (2010)); (iii) Late intervention following an elevation of creatinine (Pickering & Endre (2009c); Pickering et al. (2011)). Whilst it may be anticipated that if successful prevention or early intervention treatments are developed there will be less need for late intervention treatments, late intervention treatments will still be needed either where prevention or early intervention fails, or where it is not possible to administer it. Injury biomarkers will play a role in all three paradigms. Some are being identified as risk factors, prior to a procedure (Bennett et al. (2008)), whilst others will be elevated early enough after injury to allow early intervention, finally others will serve as outcome variables. As with plasma creatinine as an outcome variable, analogous to the RAVC, it is likely that a continuous measurement of the biomarker over a pre-specified duration will serve best as an outcome variable.

**6. Acknowledgements**

**7. References**

administered by the Royal Society of New Zealand.

injury, *Mediat Inflamm* 2009: 137072.

study, *J Am Soc Nephrol* 18(4): 1292–1298.

*Transpl* 23(4): 1203–1210.

*Care* 8(4): R204–12.

5(12): 1988–1992.

*Pharmacokinet* 4(3): 200–222.

in the critically ill, *Am J Kid Dis* 48(2): 262–268.

Dysfunction, *Nephron Clin Pract* 118(2): c173–c181.

Dr Pickering is supported by an Australia and New Zealand Society of Nephrologists enabling and infrastructure grant and the Marsden Fund Council from government funding,

The Metamorphosis of Acute Renal Failure to Acute Kidney Injury 143

Ahlstrom, A., Kuitunen, A., Peltonen, S., Hynninen, M., Tallgren, M., Aaltonen, J. & Pettila, V.

Akcay, A., Nguyen, Q. & Edelstein, C. L. (2009). Mediators of inflammation in acute kidney

Alachkar, N., Rabb, H. & Jaar, B. G. (2010). Urinary Biomarkers in Acute Kidney Transplant

Alejandro, V., Scandling, J., Sibley, R., Dafoe, D., Alfrey, E., Deen, W. & Myers, B. D. (1995).

Ali, T., Khan, I., Simpson, W., Prescott, G., Townend, J., Smith, W. & Macleod, A. (2007).

Bagshaw, S. M., George, C., Dinu, I. & Bellomo, R. (2008). A multi-centre evaluation of the

Bagshaw, S. M., Langenberg, C., Haase, M., Wan, L., May, C. N. & Bellomo, R. (2007). Urinary biomarkers in septic acute kidney injury, *Intens Care Med* 33(7): 1285–1296. Bagshaw, S. M., Uchino, S., Cruz, D., Bellomo, R., Morimatsu, H., Morgera, S., Schetz, M.,

Bellomo, R., Ronco, C., Kellum, J. A., Mehta, R. L., Palevsky, P. M. & workgroup,

Bennett, M. R. & Devarajan, P. (2011). Proteomic analysis of acute kidney injury: Biomarkers

Bennett, M. R., Ravipati, N., Ross, G., Nguyen, M. T., Hirsch, R., Beekman, R. H., Rovner, L.

Berliner, R. (1995). Homer Smith: His contribution to physiology, *J Am Soc Nephrol*

Bjornsson, T. (1979). Use of serum creatinine concentrations to determine renal-function, *Clin*

contrast induced nephropathy, *Proteom Clin Appl* 2(7-8): 1058–1064.

study of the reperfused renal allograft, *J Clin Invest* 95(2): 820–831.

*American Society of Nephrology Renal Research Report* (2005). 16(7): 1886–1903.

kidney injury., *Nephrol Dial Transpl* 24(9): 2739–2744.

to mechanisms, *Proteom Clin Appl* 5(1-2): 67–77.

(2006). Comparison of 2 acute renal failure severity scores to general scoring systems

Mechanisms of filtration failure during postischemic injury of the human kidney. A

Incidence and outcomes in acute kidney injury: a comprehensive population-based

RIFLE criteria for early acute kidney injury in critically ill patients., *Nephrol Dial*

Tan, I., Bouman, C., Macedo, E., Gibney, N., Tolwani, A., Oudemans-van Straaten, H. M., Ronco, C., Kellum, J. A. & Beginning and Ending Supportive Therapy for the Kidney (BEST Kidney) Investigators (2009). A comparison of observed versus estimated baseline creatinine for determination of RIFLE class in patients with acute

A. D. Q. I. (2004). 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*

& Devarajan, P. (2008). Using proteomics to identify preprocedural risk factors for


Table 3. Practical considerations
