Dramatic Increase of Amoxicillin-Induced Crystal Nephropathy Found in a Cohort Study of French Pharmacovigilance Centers

ABSTRACT

An increase in amoxicillin-induced crystal nephropathy (AICN) incidence has been recently suggested. The aims of this study were to investigate the trend of AICN incidence through Paris' regional centers of pharmacovigilance (Paris RCPVs) and better describe this rare adverse drug reaction. Forty-five AICN cases were identified between 1985 and 2016. All cases, except one, were reported since 2010. Amoxicillin (AMX) was administered intravenously (65 [interquartile range {IQR}, 43 to 110] mg/kg of body weight/day) in all patients, either for treating infection (n = 15) or as surgical prophylaxis (n = 30). Delay between AMX administration and AICN onset was 1 (IQR, 1 to 3) day; 30, 4, and 11 patients developed KDIGO stage 1, 2, and 3 acute kidney injury, respectively. Delay between AICN onset and kidney function recovery was 4 (IQR, 2 to 6) days. Precipitating factors were identified in only one-third of cases. Twelve patients required intensive care unit admission, and 8 needed renal replacement therapy. Neither chronic kidney disease nor death was observed. We confirmed the recent and dramatic increase of AICN in the Paris RCPVs since 2010. The absence of precipitating factors in the majority of cases and the onset of AICN in apparent routine indications, such as surgical prophylaxis, are alarming and justify a high vigilance from all AMX prescribers.

INTRODUCTION

Amoxicillin (AMX) alone or in combination with clavulanate (AMX-CLA) is widely prescribed as treatment for infections or as antimicrobial surgical prophylaxis in France. If its main adverse drug reaction is hypersensitivity reactions, amoxicillin-induced crystal nephropathy (AICN) is a rarely reported adverse drug reaction not related to hypersensitivity (1).

It is usually described as the onset of sudden, sometimes painful, macroscopic hematuria associated with oligo-anuria and an increase in creatinine level (i.e., acute kidney injury, or AKI) in patients receiving AMX. It has been described in patients receiving daily high dose of AMX for treating current infection and in patients receiving a single dose of AMX for surgical prophylaxis (2, 3). AKI results from the precipitation of AMX crystals either in renal tubules (resulting in tubular damage) or in urinary tract (resulting in kidney obstruction). However, AICN associated with preserved urinary output and/or microscopic hematuria as isolated AMX crystalluria without AKI have also been described (2). Although outcome is usually favorable after AMX discontinuation, renal replacement therapy may be temporarily required (4).

While AICN was sporadically reported until now, Zeller et al. recently observed its dramatic increase in a cohort of adult patients receiving intravenous (i.v.) high doses of AMX for osteoarticular infections. In their center, AICN incidence increased from less than one patient per year from 2004 to 2014 to 10 patients per year during 2014 to 2015, without evident explanation or change in medical practices (5).

The aims of this study were, based on cases reported to the regional centers of pharmacovigilance (RCPVs) of Paris over the past 30 years, to investigate the incidence of AICN and describe this rare adverse drug reaction.

RESULTS

From January 1985 to June 2016, the Paris RCPVs recorded 93,801 adverse drug reactions in the French Pharmacovigilance Database (FPVD). Overall, 204 cases of amoxicillin-induced renal and urinary disorders or genitourinary tract disorders, notified to Paris RCPVs by health professionals (not patients), were extracted from the FPVD (Table 1). One hundred thirty-four cases did not meet the criteria of AICN, and 25 did not enable us to rule in or rule out the diagnosis of AICN because of missing data. Finally, 45 patients (56 years [interquartile range {IQR}, 44 to 73]; male/female sex ratio, 0.45) were included in the analysis (Fig. 1 and Table 2).

TABLE 1.

Diagnosis criteria for AICN^a

Diagnosis of AICNb			Exclusion criteriac
Acute kidney injury (according to the KDIGO criteria)	Microscopic/macroscopic hematuria (confirmed by sediment urine analysis)	Temporal coincidence of acute kidney injury onset with AMX/AMX-CLA administration	Concomitant administration of nephrotoxic drugs	Any underlying condition causing acute kidney injury, macroscopic hematuria by itself, or red-colored urine (shock, sepsis, urinary tract infection, urologic surgery, kidney disease, rhabdomyolysis, hemolysis, rifampin treatment, etc.)	Fever, rash or liver enzyme abnormalities (suggesting hypersensitivity including acute interstitial nephritis)	Purpura, arthralgia, or bowel angina (suggesting amoxicillin-induced leucocytoclastic vasculitis)

^aAICN diagnosis was retained if cases fulfilled all items of AICN diagnosis and none of the exclusion criteria. The diagnosis of AICN ultimately requires the identification of amoxicillin crystals in fresh urine. Such analysis is not a routine practice and is only performed in specialized centers. AICN, amoxicillin-induced crystal nephropathy.

^bDiagnosis requires the presence of all criteria listed.

^cExclusion requires the absence of all criteria listed.

FIG 1.

Process of AICN case selection. FPVD, French National Pharmacovigilance Database; AKI, acute kidney injury; AICN, amoxicillin-induced crystal nephropathy.

TABLE 2.

Patient demographics, amoxicillin treatment characteristics, and AICN description (including AKI)

Parameter	Value(s) for group:			P value
	Total (n = 45)	Medical (n = 15)	Surgical (n = 30)
Median (IQR) age (y; n = 45)	56 (44–73)	74 (70–85)	47 (38–57)	<0.0001
Median (IQR) weight (kg; n = 41)	74.0 (60.0–86.0)	79.0 (68.0–90.0)	73.0 (58.5–84.5)	0.28
BMI
Median (IQR) BMI (kg/m2; n = 38)	26.1 (21.2–29.6)	28.0 (24.5–32.5)	25.8 (20.5–29.4)	0.16
Νο. (%) of patients with BMI between ≥25 and <30	13 (34)	3 (37)	10 (33)	1.0
Νο. (%) of patients with BMI of ≥30	9 (24)	3 (37)	6 (20)	0.36
Amoxicillin [median (IQR)]
Dose, in g/day (n = 45)	4 (3.0–11.0)	12.0 (10.0–15.0)	3.5 (2.8–4.0)	<0.0001
Dose, in mg/kg/day (n = 41)	65 (43–110)	160 (127–200)	54 (33–66)	<0.0001
Time, in days [median (IQR)]
From amoxicillin challenge to AICN onset (n = 45)	1 (1–3)	4 (3–8)	1 (1–1)	<0.0001
From AICN onset to amoxicillin discontinuation (n = 45)		2 (1–3)
From AICN onset to kidney function recovery (n = 41)	4 (2–6)	6 (5–8)	3 (2–5)	<0.01
Median (IQR) baseline creatinine level (μmol/liter; n = 45)	58 (59–72)	65 (59–80)	68 (58–71)	0.9
Median (IQR) baseline GFR (ml/min/1.73 m2; n = 45)	95 (79–114)	83 (77–111)	96 (84–117)	0.1
Median (IQR) maximum creatinine level (μmol/liter; n = 41)	319 (221–498)	425 (250–571)	304 (203–464)	0.1
Median (IQR) minimal GFR (ml/min/1.73 m2; n = 41)	14 (9–26)	12 (7–17)	15 (11–29)	0.2
No. (%) of patients by KDIGO stage (n = 45; in case of missing maximal creatinine, a >26.5-μmol/liter increase was noted in all cases)
1	30 (66)	13 (87)	17 (56)	0.05
2	4 (10)	0 (0)	4 (14)	0.3
3	11 (24)	2 (13)	9 (30)	0.28
No. (%) of patients with life-threatening complications and/or need for emergent dialysis requiring ICU admission	12 (26)	6 (0.4)	6 (0.2)	0.17

AICN represented 0.05% of all adverse drug reactions and 1% of AMX/AMX-CLA-related adverse drug reactions reported to the Paris RCPVs during the study period. AICN reports increased significantly since 2010: only 1 case was reported between 1985 and 2009 versus 44 cases (i.e., 3 [IQR, 0 to 7] cases per year) since 2010 (P < 0.001) (Fig. 2). Similarly, the proportion of AICN cases among AMX/AMX-CLA reports increased significantly since 2010 (0.03% from 1985 to 2009 versus 3.2% since 2010; P < 0.0001), while the reports involving AMX/AMX-CLA decreased significantly (5% from 1985 to 2009 versus 4.5% since 2010; P = 0.0004).

FIG 2.

Number of AICN cases reported per year to the Paris-RCPVs from 1985 to 2016; no case was reported before 2007.

AMX was prescribed i.v. in all cases, alone (n = 17, 38%) or in combination with CLA (n = 28, 62%), either to treat acute infection (considered medical patients; n = 15; 33%) or as antimicrobial surgical prophylaxis (considered surgical patients; n = 30; 66%). AMX-CLA was only prescribed in surgical patients. Medical patients were significantly older (74 [IQR, 70 to 85] years) than surgical patients (47 [IQR, 38 to 57] years) (P < 0.001). Thirteen (34%) patients were overweight (body mass index between 25 and 30 kg/m²), and 9 (24%) were obese (body mass index above 30 kg/m²). The body mass index (BMI) and the proportion of overweight and obese patients were similar between the medical and surgical groups. All patients had baseline Modification of Diet in Renal Disease (MDRD) calculated renal clearance of >60 ml/min/1.73 m². The daily doses were 4.0 (IQR, 3.0 to 11.0) g/day and 65 (IQR, 43 to 110) (n = 41) mg/kg of body weight/day and were higher in medical patients (12.0 g/day [IQR, 10.0 to 15.0] and 160 [IQR, 127 to 200] mg/kg/day) than surgical patients (3.5 [IQR, 2.8 to 4.0] g/day and 54 [IQR, 33 to 66] mg/kg/day) (P < 0.0001) (Table 2).

Delay between AMX/AMX-CLA administration and AICN onset was 1 (IQR, 1 to 3) day. This delay was longer in medical patients (4 [IQR, 3 to 8] days) than in surgical patients who developed AICN less than 24 h after surgery (P < 0.0001). As AMX/AMX-CLA was prescribed only during surgery, the delay between AICN onset and AMX/AMX-CLA discontinuation was not calculated for surgical patients; it was 2 (IQR, 1 to 3) days in medical patients. Fifteen patients (34%) developed KDIGO stage 2 to 3 AKI without differences between the two groups. The delay between AICN onset and kidney function recovery (n = 41) was 4 (IQR, 2 to 6) days longer in medical patients (6 [IQR, 5 to 8] days versus 3 [IQR, 2 to 5] days in surgical patients; P < 0.01). Urine analysis in the search of AMX crystals could be performed in 9 patients, and AMX crystals were identified in all of them. The residual AMX plasma concentration ([AMX]_plasma) was measured in 4 medical patients (314, 100, 52, and 81 mg/liter) within the 48 h following AICN onset. Renal biopsy was performed up to 10 days after the onset of AICN in four cases and showed acute tubular necrosis in all cases.

Rechallenge was performed in two medical patients without new AICN onset; however, in both cases, the duration of AMX/AMX-CLA infusion was lengthened (from 30 min previously to 1 and 4 h), while doses were adapted in cases of persistent AKI.

All enrolled cases were classified as serious (all required prolonged hospital stay), and 12 (26%) developed life-threatening events requiring intensive care unit (ICU) admission for pulmonary edema (n = 5) and/or emergent renal replacement therapy (n = 8). Neither chronic kidney disease nor death was observed. Final outcome (kidney function recovery) was favorable in 43 cases and remained unknown in two cases.

DISCUSSION

We identified and described 45 cases of AICN reported to the RCPVs of the Paris area over a 30-year period. We confirmed its recent and dramatic reporting increase since 2010 in this study, encompassing multiple centers and various settings. Although outcome was favorable in all cases, this increase is alarming. Indeed, all cases were serious, more than one-third of patients developed KDIGO stage 2 to 3 AKI, two-thirds of cases involved patients who received ultrashort treatment with AMX/AMX-CLA as prophylaxis for scheduled surgery, and one-quarter of patients required ICU admission for life-threatening complications or emergent renal replacement therapy. AICN was formerly reported in case reports, and the largest series published by Zeller et al. included only 10 medical patients briefly described in a letter; thus, to date, our study reports the largest description of this adverse drug reaction (5).

The diagnosis criteria of AICN is debatable, as no consensual definition is available. The presence of AMX crystals was confirmed in all urinary samples in which the diagnosis of AICN was suspected, but this analysis was performed in only 20% of cases (urinary analysis was not performed in the remaining cases due to logistic considerations). In addition, biological/clinical data were not always available despite additional information requests sent to RCPVs. However, we chose stringent inclusion criteria probably leading to the exclusion of cases (all notified after 2010) with potential confounding factors or missing data (Fig. 1). Moreover, the real incidence and increase in AICN could be questionable. Given the small number of publications reporting AICN since the beginning of AMX/AMX-CLAV use, AICN seems to be a rare adverse drug reaction (6). Onset of AKI during AMX/AMX-CLAV prescription is usually attributable to other factors, including sepsis or shock. Consequently, AICN is probably underrecognized and underestimated. However, no awareness campaign was launched during the study period, and no change in medical recommendations was published in the meantime. Thus, there is no clear reason why AICN would suddenly be more diagnosed since 2010.

Other contextual factors could be hypothesized to explain the recent increase of AICN reports. An increase in adverse drug reaction reports to Paris RCPVs was noted since 2010. It is a global tendency worldwide due to several factors, including implementation of safety regulation, patient's reports, and media coverage on safety issues. However, the percentage of reports involving AMX/AMX-CLA decreased significantly after 2010, while the percentage of AICN cases among AMX/AMX-CLA reports significantly increased. Moreover, case notification is based on spontaneous reporting; thus, our AICN description probably suffers from underreporting, which would aggravate the real increase in the notification rate (7). An increase in AMX prescription responsible for an increase in AICN is also unlikely. Although the exact AMX consumption during the study period is not available, an antibiotic reduction policy in France since the nineties has led to a decrease in beta-lactam (including AMX/AMX-CLA) prescription in French hospitals by 17% between 2000 and 2014 (http://ansm.sante.fr/content/download/98417/1249747/version/3/file/Resistance+Antibiotiques-nov2016.pdf). Finally, contamination during the manufacturing process or changes in formulation could be questionable. AMX/AMX-CLA was provided by three pharmaceutical companies during the 2010-2016 period, and no change in formulation was noted. The AMX manufacturing process was assessed by the Agence Nationale de Sécurité du Médicament in 2016 and found to be satisfactory [http://ansm.sante.fr/Activites/Syntheses-des-campagnes-d-inspection/Etat-des-lieux-sur-la-substance-active-Amoxicilline/(offset)/25]. We cannot rule out alterations in AMX stability related to storage conditions. However, assessment and strengthening of the rules of drug storage conditions has been implemented in French hospitals since 1999 (date of the first French health care establishments' accreditation procedure), making this hypothesis unlikely. Consistent with this, we probably underestimated this medical complication, which has a clear trend toward increase since 2010.

We identified two patterns of AICN: one occurred promptly (within hours) after AMX/AMX-CLA i.v. administered as antimicrobial surgical prophylaxis, and the second occurred within a couple of days after the first i.v. AMX administration in medical patients. The high rate of cases involving AMX-CLA was explained by the high number of surgical patients included in the study (n = 30), but clavulanate coadministration does not seem to play an additional role in AICN onset (8, 9). AICN severity (ICU admission and renal replacement therapy) was similar in both groups; however, the delay between AICN onset and kidney function recovery was longer in medical patients than in surgical patients. The older age (associated with higher tubular susceptibility) and the discontinuation of AMX within 2 (IQR, 1 to 3) days after AICN onset (suggesting that the causal relationship between AMX administration and AKI is underrecognized by clinicians) in this population may explain this difference (10).

Surprisingly, no case of AICN involved oral AMX/AMX-CLA, while it was reported in children after unintentional AMX oral overdose and in one adult case report (2, 11). Two elements may explain this discrepancy. First, AMX absorption is saturable in humans but higher in children than adults (immaturity of the intestinal barrier). Second, unintentional drug ingestions are reported to regional Poison Centers in France and not to RCPVs (12, 13). The lack of pediatric cases in this study is noteworthy, while cases involving i.v. AMX administration have been published (13). Similarly, no patient had history of chronic kidney disease in this study, while AMX is almost exclusively eliminated by the kidney (glomerular filtration and active tubular secretion) (1). This result is in accordance with previously published cases of AICN that occurred in patients with normal kidney function (2, 3, 6, 9, 14, 15). A dose adjustment recommended in this population may explain this discrepancy (1). However, chronic kidney disease might help as a protection from AICN; despite a probably higher residual AMX plasma concentration, the continuous decrease of AMX urine excretion due to altered GFR may lead to lower AMX concentration in urine and thus less AICN onset (providing sufficient urine output). Finally, we observed acute tubular necrosis in the 4 renal biopsies performed up to 10 days after the onset of AICN. In AICN, renal biopsy specimens have been described in one only case and demonstrated, at day 3, focal edema and leukocytic inflammation of the interstitium, occasional invasion of the tubules, and mitosis of the tubular epithelium, indicating tubular injury and regeneration (6). Acute tubular necrosis has also been observed in other drug-induced crystal nephropathies (16). Acute tubular necrosis may be related to the tubular injury or related to both tubular obstruction and inflammation.

The number of obese patients and its relationship with AICN is questionable. Five obese patients received an i.v. dose of 4 g of AMX (instead of 2 g in nonobese patients) for antimicrobial prophylaxis, as newly recommended by 2010 French guidelines (http://sfar.org/antibioprophylaxie-en-chirurgie-et-medecine-interventionnelle-patients-adultes/). These cases were investigated, and it was hypothesized that the change of AMX dose promoted AICN; anesthetists returned to the standard protocol, and new AICN cases remain sparse in this population. Similarly, two obese medical patients received very high AMX daily doses of 18 and 20 g/day (equivalent to 200 and 179 mg/kg/day, respectively). These data suggest the potential role of AMX overdosing in obese patients, but there is no data regarding AMX pharmacokinetics in obese patients, except the fact that AMX is poorly distributed in fat. Although a dose adjustment may be necessary in obese patients, the doubling of AMX prophylaxis dose and the AMX dose calculation based on real body weight may have contributed to AICN onset (1, 17). Further studies are necessary to better understand AMX pharmacokinetics/pharmacodynamics in obese patients and confirm this hypothesis.

However, the median daily administered dose in medical patients (frankly inferior to the maximal recommended dose of 300 mg/kg in pneumococcal meningitis) and the quasi-immediate AICN onset in surgical patients (in whom the median daily administered dose was lower) suggest that the AMX daily dose is not the sole determinant of AICN onset, and AMX administration modalities may be involved (1). Indeed, to reach the high urine AMX concentration able to induce AICN (minimum 15 g/liter in acidic urine), a large amount of AMX should be excreted. The AMX renal excretion rate follows the [AMX]_plasma curve (peak and then slow decrease) and therefore culminates at the [AMX]_plasma peak whose determinants are the injected dose, the duration of i.v. infusion, and the residual [AMX]_plasma (15). In our study, the probable rapid injection of a single i.v. dose of 4 g AMX to the five obese patients described above may have resulted in a higher [AMX]_plasma peak contributing to AICN onset. Moreover, the role of the duration of i.v. infusion in AICN onset was hypothesized in a cluster of five AICN surgical patients receiving i.v. bolus of AMX-CLA in one hospital before/during surgery. The lack of new AICN onset after prolonging AMX-CLA infusion time in the involved hospitals and the negative rechallenge in another study and in two medical patients in this study (while AMX infusion speed was reduced) strongly support this hypothesis and render a hypersensitivity mechanism unlikely (15). Finally, residual [AMX]_plasma was measured in four medical patients and was found elevated (in terms of MIC of AMX-susceptible bacteria in Europe; http://www.eucast.org/clinical_breakpoints/). Unfortunately, at the time of the AMX plasma sampling, patients had already developed AKI and AMX/AMX-CLA was still administered, making any conclusion difficult regarding whether high residual [AMX]_plasma was a precipitating factor of AICN or a consequence of AKI.

Conversely, no clear precipitating factor was identified in 66% of the remaining cases. Known precipitating factors could be hypothesized. AMX urine solubility decreases with acid urinary pH (pH_urine), changes in urine ion content, and low temperature. AMX solubility in urine and pH_urine follows a U-shaped curve, and it is divided by two when pH_urine decreased from 7 to 6 (15). No coadministration of drug-promoting urine acid release was recorded, but the high proportion of overweight/obese patients, the fast before surgery in surgical patients, and the low-fiber diet associated with the colonoscopy preparation by polyethylene glycol before colic surgery could have contributed to lower pH_urine (18–20). These last two factors also influenced the urine ion content equilibrium and the hydration state, which have been shown to affect AMX urine solubility (21). Finally, AMX solubility variation was observed in vitro (constant pH) from 2.3 g to 1.7 g when temperature decreased from 25°C to 10°C. Given the fact that body temperature of patients undergoing general anesthesia decrease roughly from 0.5°C to 1.5°C, the impact of potential intraoperative hypothermia appears marginal (22, 23). Overall, those known precipitating factors are not new and cannot explain the recent AICN increase. The retrospective design of the study and the fact that cases were disseminated in various hospitals did not allow us to compare patients who developed AICN with those who did not. However, regarding the large prescription of AMX in French hospitals and the scarcity of AICN, a prospective study would be difficult.

Finally, the increase in AICN incidence is currently reported only in France (Zeller et al. [5] and this study), while AICN cases previously have been reported in Belgium, Sweden, Germany, Italy, the Netherlands, and the United States (2, 3, 8, 9, 11, 15, 24–29). AICN increase might be obvious in France because i.v. ampicillin is not available despite similar bacteriostatic and bactericidal activity against AMX. However, ampicillin has also been responsible for crystal nephropathy (30–33). Further studies are necessary in other countries in order to investigate whether an increase in AICN only occurs in France.

Conclusions.

Spontaneous reports of AICN dramatically increased in the Paris area since 2010. Although the outcome is usually favorable, this alarming increase is becoming a public health concern involving mainly patients undergoing scheduled surgery and requiring ICU admission in one-quarter of cases. Optimal i.v. AMX administration and better knowledge of this adverse drug reaction by clinicians (including specialists in infectious disease, intensivists, and anesthesiologists) may avoid the onset of AICN and lead to better care of those patients.

Moreover, precipitating factors were not clearly identified in two-thirds of cases. Further emergent investigations are necessary and include (i) a confirmation of these data by a national investigation survey currently conducted by the Agence Nationale de Sécurité du Médicament (http://ansm.sante.fr/var/ansm_site/storage/original/application/212c52c1acdfd4ea11c7fa6556f413b3.pdf), (ii) a similar approach in other countries and in the European Medicine Agency and Food and Drug Administration databases to investigate whether AICN only occurs in France, and (iii) search for possible changes in the common active substances (raw materials), since several pharmaceutical products are involved.

MATERIALS AND METHODS

Source of data.

In France, all health caregivers and, since 2011, all patients can declare adverse drug reaction to one of the RCPVs located in French university hospitals. The Paris area has a population of 12 million inhabitants. It includes six RCPVs located in the Paris Hospital Network (Assistance Publique des Hôpitaux de Paris). Each adverse drug reaction record includes information regarding the patient's demographics and medical history, the suspected drug exposure and causality (dose, formulation, indication, challenge, withdrawal, rechallenge, and causality score), other nonsuspected drugs concomitantly administered, the adverse drug reaction severity (nonserious or serious according to World Health Organization terminology), and the adverse drug reaction outcome (34, 35). All adverse drug reactions collected by the RCPVs were registered in the French Pharmacovigilance Database (FPVD) since 1985.

Case definition.

The diagnosis of AICN ultimately requires the observation of crystals in fresh urine by phase-contrast microscopy at low and high magnifications and their identification by infrared spectroscopy or high-performance liquid chromatography (2) (Table 1). These analyses require fresh urine samples with strict storage conditions and are only performed in one laboratory in Paris in the morning of working days. As a result, AMX crystal identification is not a routine practice and could not be a systematic inclusion criterion.

Based on previous case reports, AICN diagnosis was retained if cases fulfilled all of the criteria of AICN diagnosis and none of the exclusion criteria (Table 1) (2, 4, 6). AKI was staged according to the Kidney Disease Improving Global Outcomes criteria (KDIGO) (36).

AICN case selection.

All cases of AICN reported to the Paris RCPVs from January 1985 to June 2016 were reviewed by searching in the FPVD database the reports with the suspected active substance amoxicillin combined with either renal and urinary disorders or genitourinary tract disorders (based on the high-level group term from the Medical Dictionary for Drug Regulatory Affairs, or MedDRA) (37). Blinded chart reviewing was performed by one intensivist (D. Vodovar) and one specialist in pharmacovigilance (L. Thomas). Cases of disagreement between the two investigators were ultimately resolved by C. Le Beller, specialist in pharmacovigilance. The FPVD database is declared and approved by the French National Commission on Informatics and Liberty. Informed consent of patients was unnecessary due to the anonymous process of declaration.

Description of AICN cases.

For selected cases, the following data were collected: demographics, medical history (including previous chronic kidney disease), AMX treatment characteristics (indication, i.e., medical for treating current infection or surgical for surgical prophylaxis, and formulation, i.e., AMX or AMX-CLA, dose, duration, and route of administration), AKI characteristics (urine sediment analysis, baseline/maximal creatinine level, baseline/minimal glomerular filtration rate [GFR] calculated from the modification of diet in renal disease equation, AKI KDIGO stage, and AMX crystals identification in urine), outcome (ICU admission, need for renal replacement therapy, and kidney function recovery), and AMX plasma concentration ([AMX]_plasma) when available (36, 38). AICN outcome was also assessed according to the World Health Organization adverse reaction terminology and classified as nonserious or serious (death, life-threatening event, hospitalization, prolonged hospital stay, or persistent disability, i.e., chronic kidney disease in this case) (35). The following periods of time were also collected: AMX/AMX-CLA initiation to the onset of AICN, onset of AICN to AMX/AMX-CLA discontinuation, and improvement of kidney function (defined as the first decrease in creatinine as the beginning of continuous improvement of kidney function).

Missing data.

In the case of unavailable data, a personalized questionnaire was sent to the corresponding RCPV to retrieve additional information. If supplementary data did not enable us to diagnose or rule out AICN, then the case was excluded. Missing data regarding the AICN description are pointed out in Table 2.

Statistical analysis.

Results are expressed as medians (continuous variable) or numbers/percentages (noncontinuous variable). Medians and percentages were compared between the medical and the surgical group using the Mann-Whitney test and the Fisher exact test, respectively (GraphPad Prism, version 5.3, for MacOS; GraphPad Software, La Jolla, CA; www.graphpad.com).