By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Vincent Das, Service de Réanimation Polyvalente Adultes, Centre Hospitalier Intercommunal André Grégoire, 56 Boulevard de la Boissière, 93105 Montreuil Cedex, France. e-mail: email@example.com
Study Type – Therapy (case series)
Level of Evidence 4
What's known on the subject? and What does the study add?
The pathophysiology of post-renal acute kidney injury (PR-AKI), i.e. caused by urinary tract obstruction, has been extensively studied in animal models but clinical studies on this subject are outdated, and/or have focused on the mechanisms of ‘post-obstructive diuresis’ (POD), a potentially life-threatening polyuria that can develop after the release of obstruction.
In severe PR-AKI, the risk of occurrence of POD is high. POD occurrence predicts renal recovery without the persistence of severe chronic kidney failure. In the present study, the occurrence of POD and the persistence of chronic renal sequelae could be predicted early from clinical variables at admission before the release of obstruction.
• To identify predictors of post-obstructive diuresis (POD) occurrence or severe chronic renal failure (CRF) persistence after the release of urinary tract obstruction in the setting of post-renal acute kidney injury (PR-AKI).
PATIENTS AND METHODS
• Bi-centre retrospective observational study of all patients with PR-AKI treated in two intensive care units (ICUs) from 1998 to 2010.
• Clinical, biological and imaging characteristics on admission and after the release of obstruction were analysed with univariate and, if possible, multivariate analysis to search for predictors of (i) occurrence of POD (diuresis >4 L/day) after the release of obstruction; (ii) persistence of severe CRF (estimated glomerular filtration rate <30 mL/min/1.73 m2, including end-stage CRF) at 3 months.
• On admission, median (range) serum creatinine was 866 (247–3119) µmol/L.
• POD occurred in 34 (63%) of the 54 analysable patients. On admission, higher serum creatinine (Odds ratio [OR] 1.002 per 1 µmol/L, 95% confidence interval [CI] 1.000–1.004, P= 0.004), higher serum bicarbonate (OR 1.36 per 1 mmol/L, 95% CI 1.13–1.65, P < 0.001), and urinary retention (OR 6.96, 95% CI 1.34–36.23, P= 0.01) independently predicted POD occurrence.
• Severe CRF persisted in seven (21%) of the 34 analysable patients, including two (6%) cases of end-stage CRF. Predictors of severe CRF persistence after univariate analysis were: lower blood haemoglobin (P < 0.001) and lower serum bicarbonate (P= 0.03) on admission, longer time from admission to the release of obstruction (P= 0.01) and absence of POD (P= 0.04) after the release of obstruction.
• In severe PR-AKI treated in ICU, POD occurrence was a frequent event that predicted renal recovery without severe CRF.
• POD occurrence or severe CRF persistence could be predicted early from clinical and biological variables at admission before the release of obstruction.
Acute kidney injury (AKI) occurs in 100–600 cases per million adults and per year in the community setting . In all, 5–10% of AKI episodes are classified as post-renal (PR-AKI), i.e. caused by urinary tract obstruction [2,3]. In one study performed in the elderly, the proportion of PR-AKI even reached 22% of AKI episodes . The pathophysiology of obstructive nephropathy (the consequences of urinary tract obstruction on renal parenchyma) has been extensively studied in animal models of ureteric obstruction, but clinical studies on PR-AKI are scarce, often outdated, and/or have focused on the mechanisms of ‘post-obstructive diuresis’ (POD), a potentially life-threatening polyuria that can develop after the release of obstruction [5–13]. Clinically relevant questions are still unanswered. For example, can we predict POD occurrence, and select patients who should undergo careful monitoring of volume and electrolytic status, at best in the intensive care unit (ICU)? [10,13–15] Also, can we predict renal recovery after the release of obstruction? [5,12,16,17] Therefore, we conducted a retrospective study in two ICUs to re-assess the outcome of PR-AKI and identify predictors of POD occurrence or severe chronic renal failure (CRF) persistence.
PATIENTS AND METHODS
We conducted a retrospective study of consecutive patients with PR-AKI admitted during the period 1998–2010 in two French medical-surgical ICUs, one being part of a 360-bed general non-teaching hospital and the other of a 500-bed teaching hospital. In both cases, a urological surgeon was available at all times.
The databases of ICU discharge reports were screened to identify the patients admitted for PR-AKI. The diagnosis of PR-AKI was based on the following criteria: (i) criteria for AKI according to the class ‘R’ (Risk) or higher of the RIFLE classification, defined as an increase in serum creatinine × 1,5 above baseline (or above normal value if unknown) or oliguria (urine outflow <0.5 mL/kg/h or 400 mL/day) , and (ii) dilatation of the renal calyces on ultrasonographic examination or CT, or, in case of doubt, a favourable outcome after the release of urinary tract obstruction. The decision to admit a patient to ICU was left to the attending intensivist, based on hyperkalaemia, metabolic acidosis, pulmonary overload, and/or high serum urea and creatinine. It was our policy to admit all severe PR-AKI episodes to the ICU, and not to restrict ICU admission for the patients whose obstruction had been released and in whom POD had already started. Renal transplant recipients and patients with lost charts were excluded.
Urinary tract obstruction was relieved in emergency unless urgent renal replacement therapy was required for severe hyperkalaemia or pulmonary oedema unresponsive to medical treatment or stabilisation in ICU was mandatory in critically ill patients. Obstruction could be released using either a urethral catheter or suprapubic cystotomy in case of urinary retention, and either bilateral insertion of ureteric catheters (JJ stents or transvesico-urethral catheters) or nephrostomy tubes in case of ureteric obstruction.
Our usual policy to treat PR-AKI in ICU was as follows: vital signs and diuresis were monitored hourly, serum electrolytes, serum urea nitrogen and creatinine and fluid balance twice or thrice a day, weight daily; if POD occurred, urinary losses were only partially compensated. No fixed rule for compensation was applied, and fluids prescription was left to the discretion of the attending intensivist. Diuretics use was contraindicated.
The following data were collected from medical and nurse charts:
• General demographic characteristics;
• Baseline estimated GFR (eGFR) was calculated from the last serum creatinine measured before the beginning of urinary symptoms using the Modification of Diet in Renal Diseases (MDRD) equation. If no baseline creatinine value was available, this data was considered as missing, except if serum creatinine regained normal values after the release of obstruction, where the baseline creatinine value was assumed to be the same as the lowest creatinine value during follow-up.
• History of on-going treatment with angiotensin-converting enzyme (ACE) inhibitors just before admission;
• Biological data at ICU admission: serum urea and creatinine, serum electrolytes, blood haemoglobin, arterial blood gases;
• Characteristics of the PR-AKI episode: duration of urinary symptoms before admission, cause of urinary tract obstruction, mean kidneys size (mm), mean diameter of kidney pelvises (mm), presence of urinary infection, type of procedure to release obstruction, elapsed time from hospital admission to the release of obstruction. Mean kidneys size was considered as not available in case of solitary kidney, due to compensatory hypertrophy;
• Outcome after the release of obstruction: levels of electrolytes, urea and creatinine in the first urine sample; course of serum creatinine, urea, electrolytes levels and diuresis during the first 3 days; occurrence of POD and its consequences (changes in natremia, hypotension) during the first 3 days; requirement for renal replacement therapy, renal function, and survival in ICU, at hospital discharge, and 3 months after admission. POD was defined as diuresis of >4 L/day [1,19].
Data are shown as percentage or median (range) values. Predictors of POD occurrence were identified with univariate analysis, using non-parametric tests, followed by multivariate analysis. As shock easily results in oliguric acute tubulopathy, patients admitted with septic shock or after cardiac arrest were excluded from this analysis. The receiver-operating characteristics (ROC) curve of the linear predictor, derived from the multivariate model, to predict POD occurrence was constructed, and the area under the ROC curve (AUC) was calculated. The Youden index (1–[sensitivity + specificity]) was calculated, and the best threshold value with a sensitivity >90% was determined . A high sensitivity was required to insure that, if the score was used to select the patients at risk of POD occurrence for ICU admission, few patients not admitted to the ICU would eventually have POD.
The persistence of chronic end-stage renal disease (ESRD) was too rare an event to search for predictors (see results). Instead, predictors of ‘severe CRF’ persistence (eGFR of <30 mL/min/1.73 m2, including patients with ESRD) at 3 months were identified with univariate analysis using non-parametric tests. Only the data of the patients free of pre-existing severe CRF were analysed. Due to missing data, multivariate analysis could not be performed. The ROC curves of the predictors were constructed. The best threshold values were determined by calculation of the Youden index.
The statistical analysis was performed using the R statistical software (http://www.r-project.org/). A P < 0.05 was considered to indicate statistical significance.
According to the French regulation on research performed on data, usual informed consent was waived for this observational study.
During the study period, 888 patients were admitted in ICU for AKI. In all, 68 patients (8%) had PR-AKI. Six charts were lost. Therefore, we studied 62 PR-AKI episodes. Patients' characteristics on ICU admission are given in Table 1. The reason for ICU admission was pure AKI in 55 (89%) cases. Other reasons included: septic shock (two cases), cardiac arrest complicating hyperkalaemia (two), acute respiratory failure (one), status epilepticus (one), and ketoacidocetosis (one).
Table 1. The patients' characteristics at ICU admission
Characteristics (number of charts with available data)
Median (range) or N (%) value
SAPS II, simplified acute physiology score II; SOFA, sequential organ failure assessment.
Age, years (62)
Gender male/female, (62)
46 (74)/16 (26)
Previous treatment with ACE inhibitors, (62)
Prior eGFR, mL/min/1.73 m2 (35)
Characteristics of the obstruction
Symptoms duration, days (43)
Bilateral/unilateral (solitary kidney), (62)
55 (89)/7 (11)
Mean diameter of kidney pelvises, mm (34)
Urinary retention (62)
Characteristics at ICU admission
SAPS II (62)
Serum urea, mmol/L (62)
Serum creatinine, µmol/L (62)
RIFLE class risk/injury/failure (62)
0/2 (3)/60 (97)
Serum sodium, mmol/L (62)
Serum potassium, mmol/L (62)
Serum bicarbonate, mmol/L (62)
Arterial pH (58)
Serum calcium, mmol/L (60)
Serum phosphate, mg/L (59)
Blood haemoglobin, g/dL (61)
The causes of urinary tract obstruction are given in Table 2. Progressive cancer was the first cause of obstruction, accounting for half of the cases. Most cancers (90%) were of pelvic origin. The most frequent type of cancer was prostate cancer (55%), whereas bladder or uterus cancers were causal in 13% of cases each. Miscellaneous causes of obstruction included : post-radiation ureteritis/bladder dysfunction (two cases), urinary retention after enterocystoplasty (two), retroperitoneal fibrosis (two), urethral stricture (one), other bladder disorders (four) and fecaloma (one).
Urinary tract obstruction was released with urethral catheter or suprapubic cystotomy in 32 (52%) cases, ureteric catheterisation in 18 (29%) cases, and percutaneous insertion of nephrostomy tubes in 12 (19%) cases. Renal replacement therapy was required in ICU in 25 (40%) cases.
After exclusion of the patients admitted for cardiac arrest (two) or septic shock (two) or with missing data (four), 54 PR-AKI episodes were analysed to assess POD incidence and identify predictors of POD occurrence. POD occurred in 34 (63%) cases. It started at a median (range) of 3 (0–27) h after the release of obstruction and lasted 1.8 (0.4–12.5) days. Figure 1 shows patients' diuresis and fluids balance during the first 3 days after the release of obstruction. As displayed, in patients with POD, median (range) diuresis during the first 24 h was 7000 (4000–15 000) mL, and fluid balance was negative. The changes in serum sodium and the incidence of hypotension during the first 3 days were not statistically different whether POD occurred or not (data not shown).
Predictors of POD occurrence with univariate analysis are shown in Table 3. In a multivariate model including blood haemoglobin, serum bicarbonate, serum creatinine, and presence of urinary retention on admission, the independent predictors of POD occurrence were: higher serum bicarbonate (OR 1.36 per 1 mmol/L, 95% CI 1.13–1.65, P < 0.001), higher serum creatinine (OR 1.002 per 1 µmol/L, 95% CI 1.000–1.004, P= 0.004) and urinary retention (OR 6.96, 95% CI 1.34–36.23, P= 0.01). Baseline renal function was not included in the multivariate analysis due to missing data.
Table 3. Predictors of occurrence of POD: results of the univariate analysis
Patients with POD
Patients with no POD
Median (range) age, years (54)
Male gender, n (%) (n= 54)
Median (range) previous eGFR, mL/min/1.73 m2 (32)
Previous treatment with ACE Inhibitors, n (%) (54)
Cause of obstruction: cancer; adenoma; lithiasis; other, n (%) (54)
18 (53); 8 (23); 3 (9); 5 (15)
10 (50); 3 (15); 1 (5); 6 (30)
At ICU admission
Median (range) Symptoms duration, days (43)
Median (range) value of mean kidneys size, mm (38)
Median (range) value of mean diameter of kidney pelvises, mm (29)
Urinary retention, n (%) (54)
Urine infection, n (%) (54)
Median (range) serum sodium, mmol/L (54)
Median (range) serum bicarbonate, mmol/L (54)
Median (range) serum urea, mmol/L (54)
Median (range) serum creatinine, µmol/L (54)
Median (range) serum calcium, mmol/L (53)
Median (range) serum phosphate, mmol/L (52)
Median (range) blood haemoglobin, g/dL (54)
After the release of obstruction
Median (range) time from admission to the release of obstruction, h (31)
Median (range) urine sodium, mmol/L (41)
Median (range) urine urea, mmol/L (35)
Median (range) urine creatinine, mmol/L (33)
The linear predictor derived from the multivariate model was: ‘0.31 × serum bicarbonate (mmol/L) + 0.002 × serum creatinine (µmol/L) + 1.94 if urinary retention −7.15’. This score predicted POD occurrence with an AUC of 0.85 (95% CI 0.749–0.951). The best threshold value was 0. A score higher than 0 predicted POD occurrence with a 91% sensitivity, a 60% specificity, a 79% positive predictive value (PPV) and an 80% negative predictive value (NPP).
The median follow-up duration was 183 (1–2223) days. At 3 months, three patients (5%) had chronic ESRD. After exclusion of patients lost to follow-up and those who died (Fig. 2), renal outcome was analysed in 39 patients. One of the five patients with a pre-existing severe CRF had chronic ESRD at 3 months. Seven (21%) of the 34 patients free of pre-existing severe CRF had severe CRF persisting at 3 months, including two (6%) with chronic ESRD (P= 0.35 for comparison with patients with prior severe CRF).
Predictors of severe CRF persistence with univariate analysis (shown in Table 4) were: on admission: lower blood haemoglobin and lower serum bicarbonate; elapsed time from admission to the release of obstruction; after the release of obstruction: lower diuresis during the first 24 h, the absence of POD, and higher urinary sodium. Renal parenchyma width was not tested due to missing data. Due to the high number of patients with previous severe CRF or missing data, multivariate analysis was not performed.
Table 4. Predictors of persistence of severe CRF at 3 months: results of univariate analysis
Patients with severe CRF
Patients without severe CRF
Median (range) age, years (34)
Male gender, n/N or n (%) (34)
Median (range) previous eGFR, mL/min/1.73 m2 (17)
Previous treatment with ACE inhibitors, n/N or n (%) (34)
Cause of obstruction: cancer; adenoma; lithiasis; other, n/N or n (%) (34)
1/7; 4/7; 1/7; 1/7
15 (56); 6 (22); 4 (15); 2 (7)
At ICU admission:
Median (range) symptoms duration, days (24)
Mean (range) kidneys size, mm (23)
Mean (range) diameter of kidney pelvises, mm (19)
Urinary retention, n/N or n (%) (34)
Urinary infection. n/N or n (%) (34)
Median (range) serum sodium, mmol/L (34)
Median (range) serum bicarbonate, mmol/L (34)
Median (range) serum urea, mmol/L (34)
Median (range) serum creatinine, µmol/L (34)
Median (range) serum calcium, mmol/L (34)
Median (range) serum phosphate, mmol/L (33)
Median (range) blood haemoglobin, g/dL (34)
After the release of obstruction:
Median (range) time from admission to the release of obstruction, h (21)
Median (range) urine sodium, mmol/L (29)
Median (range) urine urea, mmol/L (29)
Median (range) urine creatinine, mmol/L (29)
Median (range) diuresis during the first 24 h, mL (34)
7052 (1500–15 080)
POD occurrence, n/N or n (%) (33)
Need of renal replacement therapy in ICU, n/N or n (%) (34)
The correlation of diuresis (during the first 24 h after the release of obstruction) with renal recovery at 3 months is shown in Fig. 3a,b. Two threshold values of interest were identified: a diuresis of >7000 mL/day predicted renal recovery without severe CRF with a sensitivity of 50%, a specificity of 100%, a PPV of 100% and a NPV of 35%, whereas a diuresis of <4000 mL/day predicted severe CRF persistence with a sensitivity of 57%, a specificity of 85%, a PPV of 50% and a NPV of 88%.
Similarly, a blood haemoglobin concentration of <10 g/dL and serum bicarbonate of <15 mmol/L predicted severe CRF persistence with sensitivities of 100% and 86%, specificities of 78% and 86%, PPVs of 54% and 43%, and NPVs of 100% and 95%, respectively.
In-hospital survival rate was 80%, both for patients with cancer (31) and without cancer (31). The 3-month and 6-month survival rates in survivors at hospital discharge were: 81% (47) and 72% (43), respectively, in the entire population; 92% (24) and 91% (22), respectively, in patients without cancer; and 70% (23) and 58% (19), respectively, in patients with cancer (P= 0.12 and P= 0.03, respectively, for comparison with patients without cancer).
In this population of ICU-admitted patients with severe PR-AKI episodes, we observed that: (i) POD frequently occurred; (ii) at 3 months, in patients free of pre-existing severe CRF, the persistence of chronic ESRD was infrequent but severe CRF persisted in one case out of five. Above all, we identified early predictors of POD occurrence or severe CRF persistence, which could be helpful to physicians dealing with patients with PR-AKI in clinical daily practice.
To our knowledge, the present series is one of the largest of PR-AKI episodes reported [21–23]. As previously reported, the first cause of urinary tract obstruction was cancer, especially prostate cancer . Interestingly, if hospital mortality was low, 6-month mortality was high in patients with obstruction due to cancer, a result which has already been highlighted by others .
The risk of life-threatening consequences of POD (i.e. hypovolemia, hydroelectrolytic disorders) require close monitoring of patients with PR-AKI and compensation, at least partial, of urinary losses [7–10,15]. POD has been the subject of numerous pathophysiological studies [5,8–11,14,25–28]. Yet, in clinical practice, the risk of POD occurrence is unknown and predictors have not been determined [10,13,14]. In the present study, POD frequently occurred and was probably not due to overzealous compensation of urinary losses, since fluid balance was negative in our patients with POD [10,13,29]. Three independent predictors of POD were identified, all available on admission: higher serum creatinine, higher serum bicarbonate, and the presence of urinary retention. One could have expected that the severity of tubular defects in urine acidification would be correlated with defects in sodium and water reabsorption resulting in POD, and, consequently, that POD incidence would actually increase as serum bicarbonate decreased . In the present study, it was the contrary. Actually, the severity of metabolic acidosis correlated with the absence of POD and predicted the presence of chronic renal lesions, as did the severity of anaemia. The absence of POD also directly correlated with severe CRF persistence (see below), which gives consistency to the present results. As we observed that POD incidence correlated with higher serum creatinine and inversely correlated with severe CRF persistence (or its predictor, metabolic acidosis), the present results confirm old observations that POD occurred ‘after a relatively short period of severe obstruction without irreversible renal damage, but always in the presence of advanced renal failure’, and suggest that POD occurrence could be expected if markers of CRF are absent . The correlation that we observed between the presence of urinary retention and POD occurrence has not been explained. The patients with urinary retention could have less severe chronic renal lesions due to shorter duration of obstruction or lower intratubular pressures than patients with ureteric obstruction. Above all, the present results suggest that POD occurrence can be predicted early, before the release of obstruction. Physicians could take predictors of POD occurrence into account, in combination with criteria of AKI severity (hyperkalaemia, pulmonary oedema …), to select patients who should be closely monitored and therefore admitted to the ICU.
Various mechanisms have been identified in the pathogenesis of obstructive nephropathy, leading to renal vasoconstriction and progressive renal fibrosis [5,6,32,33]. While renal vasoconstriction is reversible after the release of obstruction, renal fibrosis may result in irreversible loss of function [5,34]. Yet, the outcome of renal function after the release of obstruction has been rarely studied in humans [12,15,21]. In the present study, the persistence of ESRD was infrequent in patients free of pre-existing severe CRF, but severe CRF frequently persisted (21% of cases). Also, predictors of renal recovery after the release of obstruction have been seldom studied . In animal models, the extent of renal recovery was influenced by the duration and severity of obstruction, but, in humans, it was admitted that renal recovery could not be predicted at the time of diagnosis of PR-AKI [5,12,16,17,34,36–38]. In the present study, the severity of anaemia and metabolic acidosis on admission, a longer delay from admission to the release of obstruction, lower diuresis during the 24 h after the release of obstruction (and, consequently, the absence of POD), and higher urinary sodium after the release of obstruction, predicted severe CRF persistence at 3 months with univariate analysis. Multivariate analysis could not be performed, but these results contribute new evidence to several hypotheses raised in the literature.
First, the present results confirm that POD occurrence is predictive of renal recovery. This result was cited in the literature, but, to the best of our knowledge, our work is the first where the correlation has been statistically established [12,31,39]. We were able to determine a threshold value of 7000 mL for diuresis during the first day after the release of obstruction, which predicted renal recovery without severe CRF with a high specificity, whereas diuresis of 4000–7000 mL per day did not predict renal recovery with certainty.
Second, in the present study, previous treatment with ACE inhibitors was not correlated with renal recovery. In animal models, ACE inhibitors, given before ureteric obstruction, prevented the development of renal fibrosis. Some authors have even proposed giving ACE inhibitors to patients with chronic obstruction . The present results do not support this proposal. However, in the present study, ACE inhibitor treatment was obviously given to patients with prior hypertension or cardiac failure, with probably worse baseline renal function due to nephroangiosclerosis.
Finally, as severe CRF persistence could be predicted early, physicians could select patients at risk of severe CRF persistence for prolonged monitoring of renal function after ICU discharge and early preservation of the venous network.
Renal parenchyma width was not measured in the present study. Also, kidney size was not correlated with renal outcome, but the large amount of missing data makes this result unreliable. The prognostic value of imaging variables in PR-AKI is controversial and remains to be assessed, at best in a prospective study [5,35,40].
The best time to release urinary obstruction in the setting of PR-AKI is unknown, except in cases of sepsis, where it should be performed as an emergency . In the present study, renal outcome inversely correlated with elapsed time from admission to the release of obstruction, which could suggest that the release should be performed as an emergency even in the absence of sepsis. Alternatively, the patients with short time-to-release could be intrinsically different from other patients (differences in cause of obstruction, etc.). A randomised study would be mandatory to answer this question.
The limits of the present study should be highlighted. First, due to its retrospective design, some data were missing and multivariate analysis of predictors of renal recovery could not be performed. Also, the course of renal function at longer follow-up remains to be determined . Second, the sample size was small, and the correlation of some variables could have reached statistical significance if more patients had been included. Third, the present study only included patients with severe AKI treated in ICU, and the extrapolation of our results to patients with less severe PR-AKI should be done cautiously. Finally, a large multicentre prospective study would be mandatory to ensure adequate collection of data to perform multivariate analyses in a larger number of patients, and provide external validation to our results.
In conclusion, in the present series of ICU-treated PR-AKI, the risk of POD occurrence was high, the risk of ESRD persistence was very low, but the risk of severe CRF persistence was relevant. The present study confirmed that POD occurred in the presence of severe renal failure without irreversible chronic renal lesions. POD occurrence or severe CRF persistence could be predicted early from admission, which may help physicians to determine which patients should be admitted to the ICU and adjust monitoring after ICU discharge. However, a large prospective study is mandatory to provide external validation to the present results and assess the value of other variables, such as imaging data, as renal recovery predictors.
The authors wish to thank Dr Naïke Bigé and Mrs Gonzalez for her careful reading of the manuscript.