Incidence and risk of dialysis therapy within 30 days after contrast enhanced computed tomography in patients coded with chronic kidney disease: a nation-wide, population-based study

Introduction

Administration of intravenous iodinated contrast media (ICM) during computed tomography (CT) scans plays an indispensable role in improving diagnostic performance in numerous settings and is thereby widely employed in clinical practice. The healthcare industry has seen a steady increase in demand for CT scans ordered by clinicians, which inevitably leads to a corresponding increase in use of ICM (Hu et al., 2016). However, intravenous ICM also raises major concerns of the development of contrast-induced acute kidney injury (CI-AKI). The exact pathophysiology leading to this condition is poorly understood, but could possibly result from contrast-induced vasoconstriction, a decrease in filtration via tubuloglomerular feedback, direct renal epithelial toxicity, and increased free radial formation (Pasternak & Williamson, 2012). The elevations of blood osmolality and tubular fluid viscosity may also cause microvascular thrombosis and renal tubular obstruction, respectively (Mehran, Dangas & Weisbord, 2019). After excluding other etiologies of renal impairment, the diagnostic criteria for CI-AKI are serum creatinine elevation varying from 0.3 to 0.5 mg/dL in an absolute increase, or a 25–50% relative increase within the time period of 24–72 h after contrast administration (McDonald et al., 2013; Mehran & Nikolsky, 2006; Meinel et al., 2014).

It is known that patients with chronic kidney disease (CKD) are particularity at risk of developing CI-AKI due to diminished renal function reserve (Davenport et al., 2013; Kim et al., 2010; ACR Committee on Drugs & Contrast Media, 2018; Mehran & Nikolsky, 2006; Scharnweber, Alhilali & Fakhran, 2017). However, there is also emerging evidence suggesting that the risk of CI-AKI in CKD patients may have been overestimated in the past (Garfinkle, Stewart & Basi, 2015; Hinson et al., 2017; McDonald et al., 2013; Mehran, Dangas & Weisbord, 2019; Van Der Molen et al., 2018). The type, amount and route of ICM given to a patient are also crucial parameters. Patients receiving high-osmolar contrast medium (HOCM), i.e., ionic monomers, large volumes of ICM administration, or intraarterial ICM injections are at high risks of developing CI-AKI (Pasternak & Williamson, 2012), and these findings are mostly based on studies focusing on coronary interventional procedures. However, the doses given during contrast-enhanced CT scans are intravenous and usually smaller in volumes as compared with coronary interventional procedures. In addition, the majority of ICM now used in modern clinical practices are low-osmolar (LOCM) and iso-osmolar contrast media (IOCM). Therefore, many are now questioning the existence of clinically significant CI-AKI in this setting (Mehran, Dangas & Weisbord, 2019; Pasternak & Williamson, 2012). Baseline fluctuations in renal function indices may also cause spurious indications of CI-AKI, which are unrelated to ICM administration (Bruce et al., 2009). Nevertheless, it is the current consensus amongst practicing radiologists that pre-existing CKD is still the most important risk factor in the development of CI-AKI (ACR Committee on Drugs & Contrast Media, 2018; Van Der Molen et al., 2018). In addition, the prevalence of CKD is high in many countries (Hill et al., 2016). If the risk is truly overestimated, contrast-enhanced CT scans may be inappropriately withheld from patients with legitimate clinical indications, resulting in suboptimal or delayed diagnoses of serious conditions and unfavorable clinical outcomes.

Contrast-induced acute kidney injury is often but not always a reversible process (Scharnweber, Alhilali & Fakhran, 2017). One of the adverse events after CI-AKI causing the greatest concern is the possible necessity for dialysis therapy due to sustained renal function impairment. The exact figure for CKD patients who require dialysis shortly after undergoing CT with ICM is not well known, and it possess a high level of clinical interest. The necessity for dialysis in the short-term is defined within a time-frame of 30 days after undergoing CT scan, in concordance with other studies published in the literature (McDonald et al., 2014, 2015, 2017a). To directly identify CKD patients, it is essential to obtain laboratory information. An alternative method is to select patients with CKD based on diagnostic codes within an administrative dataset. The usefulness for administrative data regarding CKD status has been validated in various research articles, which reveal inconsistent sensitivities but very high specificities (Grams et al., 2011; Navaneethan et al., 2011; Ronksley et al., 2012; Tonelli et al., 2015). In the published literature, numerous studies have been conducted implementing diagnostic codes to identify CKD patients (Chang et al., 2014; Chen et al., 2016; Chung et al., 2017; Yu et al., 2017). In this study, a longitudinal, nation-wide, population-based data is used to evaluate the incidence and risk of the necessity for dialysis within 30 days of undergoing contrast-enhancing CT scans amongst CKD patients. It is also our imperative to determine whether the administrative data is able to successfully identify CKD patients who are at risk.

Materials and Methods

Patient identification and selection

Within the LHID2010, CKD patients were identified by using relevant International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes with at least two outpatient claims or one inpatient claim documented in the diagnosis codes section. Patients receiving CT scans were selected by using the Taiwan NHI codes for procedures, including information on study date and the presence or absence of ICM administration. The specific ICM injected was unknown, however, the majority of ICM used in Taiwan is LOCM, i.e., non-ionic monomers. CKD patients receiving CT without ICM were assigned to the control group. The specific dates of CT scans were not available before year 2012 within LHID2010, therefore only CKD patients receiving CT during January 2012 to December 2013 were included for analysis. The following patients were excluded: ages below 20 or over 100 years old; patients who were already receiving regular dialysis before the date of the CT scan; patients receiving CT scans for two or more times within 30 days; and patients who had undergone angiography or transarterial embolization within 30 days of the CT scan. The flow diagram for patient selection was shown in Fig. 1. All included patients were followed up for a 30-day period, to the occurrence of dialysis or death events, until the end of December 2013 or when leaving the NHI program. The follow-up period was not limited to 30 days when evaluating the time interval from CT to dialysis event as a secondary outcome described subsequently. Furthermore, since laboratory data was not available in LHID2010, the severity of CKD could not be directly assessed. An indirect approach was to select CKD patients receiving erythropoietin-stimulating agents (ESA) as the advanced CKD subgroup in this study. The NHI reimbursed ESA only in CKD patients with serum creatinine above 6 mg/dL, approximately equivalent to an estimated glomerular filtration rate (eGFR) <15 mL/min/1.73 m², and a hematocrit level below 28% observed concurrently. This approach to subdivide patients with advanced CKD had also been used in previous research articles (Chung et al., 2017; Hsieh et al., 2016). Another subgroup was the status of admission to the intensive care unit (ICU) during the time the CT scan was performed.

Results

Necessity for dialysis and death within 30 days of CT

There were 20 patients (4.11%) in the CT with ICM group and 66 patients (7.14%) in the CT without ICM group requiring dialysis within 30 days of CT scans. The cumulative incidences of dialysis events were drawn in Fig. 2. A greater number of patients underwent dialysis therapy in the CT without ICM group (p = 0.0295). However, no statistically significant difference was found after the adjustment of age, sex, and the specified comorbid conditions (aHR 0.84, 95% CI [0.46–1.54], p = 0.5700). A total of 16 patients (3.29%) died in the CT with ICM group, which was also not significantly different from the 61 patients (6.60%) who died in the CT without ICM group (aHR 0.62, 95% CI [0.35–1.10], p = 0.1012). In the advanced CKD and ICU admission subgroups, neither dialysis within 30 days after CT nor death rates differ significantly between the CT with or without ICM groups. After excluding the critically-ill patients in the ICU, the percentages of patients requiring dialysis in 30 days was 1.61% in the CT with ICM and 2.81% in the CT without ICM groups, respectively (aHR 1.11, 95% CI [0.40–3.06], p = 0.8371). The events, percentages and aHRs of dialysis and death are shown in Table 2.

Table 2:

Adjusted hazard ratios (aHR) for dialysis and death within 30 days after contrast enhanced CT scans in CKD patients, advanced CKD subgroup, ICU admission subgroup, and CKD patients excluding ICU admissions.

	Outcome	CT with ICM N = 487 n/N (%)	CT without ICM N = 924 n/N (%)	Adjusted HR* (95% CI)	p-Value
CKD patients	30-day dialysis	20 (4.11)	66 (7.14)	0.84 [0.46–1.54]	0.5700
	30-day mortality	16 (3.29)	61 (6.60)	0.62 [0.35–1.10]	0.1012
Advanced CKD subgroup^	30-day dialysis	7 (46.67)	30 (35.71)	1.12 [0.38–3.33]	0.8333
	30-day mortality	2 (13.33)	4 (4.76)	3.87 [0.16–92.07]	0.4032
ICU admission subgroup	30-day dialysis	13 (25.49)	45 (25.57)	0.95 [0.44–2.05]	0.8973
	30-day mortality	6 (11.76)	42 (23.86)	0.84 [0.22–1.33]	0.1788
CKD patients excluding ICU admission	30-day dialysis	7 (1.61)	21 (2.81)	1.11 [0.40–3.06]	0.8371
	30-day mortality	10 (2.29)	19 (2.54)	1.24 [0.54–2.85]	0.6158

DOI: 10.7717/peerj.7757/table-2

Notes:

n/N, event numbers/total numbers; ICM, iodinated contrast media; ICU, intensive care unit.

^ Advanced CKD subgroup is defined as CKD patients with concurrent usage of erythropoietin-stimulating agents.

* The HRs were adjusted by the age, sex, and comorbidities listed in Table 1.

Discussion

In our study, the adjusted risks of dialysis and death within 30 days after CT scans in CKD patients were not increased, regardless of whether the patient received ICM or not. Similar results were reported in three retrospective studies from the same institution (McDonald et al., 2014, 2015, 2017a). Regardless of CKD stage 3, 4, or 5, the incidences of dialysis and death rates were also not affected by ICM (Garfinkle, Stewart & Basi, 2015; McDonald et al., 2015, 2017a). Therefore, CT with ICM is probably not an independent risk factor for the necessity of dialysis and event of death in CKD patients, even in advanced stages, in the short-term observation. The crude dialysis rate was lower in our CT with ICM group, although not statistically significant after adjustment. One reason could be the younger age and the existence of fewer comorbid conditions in this group. The crude dialysis rate elevation in CKD patients receiving CT without ICM was observed by others as well (McDonald et al., 2017a). A recent meta-analysis consisted of patients with acute ischemic stroke also revealed a lower CI-AKI risk in those receiving contrast enhanced CTs as compared with those who received non-enhanced ones (Brinjikji et al., 2017). Consistent with our results, they became non-significant after the adjustment of confounding factors. It is possible that various predisposing conditions lead to concerns of developing CI-AKI, and clinicians may have ordered non-contrast CTs instead. The underlying causes may have resulted in renal failure and subsequent elevation in the crude dialysis rate in our studied population.

Reported post-CT dialyses within 30 days were few in the reviewed studies, with an overall percentage of 0.2–0.3% in CKD patients and only 5.1% in CKD stages four to five patients (McDonald et al., 2014, 2017a). However, our study revealed higher rates of post-CT dialyses within 30 days at 4.1–7.1%. When excluding the critically-ill ICU patients, the percentages were still high at 1.6–2.8%. This finding could be related to the fact that Taiwan has the highest incidence of end-stage renal disease (ESRD) in the world (United States Renal Data System, 2017). Abundant dialysis-related resources are available, and the NHI provides full coverage of incurred expenses. Due to this relatively common practice, doctors, patients, and their family members would probably prefer dialysis therapy in the occurrence of renal failure in most situations. Another possible cause is that some CKD patients were unlabeled from ICD-9-CM, especially those with less severe disease (Ferris et al., 2009). Therefore, the reported incidences of dialyses could be overestimated in our study.

The importance of critical illness was mentioned in a previous study, stating that ICU patients with eGFR <45 mL/min/1.73 m² demonstrated an increased risk of emergent dialysis within 7 days after contrast enhanced CT scans (McDonald et al., 2017b). An earlier article also demonstrated that the development of CI-AKI may result in increased need for dialysis in the ICU (Valette et al., 2012). However, in our results, the necessity for post-CT dialysis within 30 days and death rates did not differ in the ICU subgroup. A more detailed stratification of patient condition, especially any acute illness, might be required to reveal any undesirable effect of ICM injected during CT particularly in the critically-ill setting.

The advanced CKD subgroup appears interesting since they were presumed to be most susceptible to CI-AKI and the possibility of dialysis. It is noteworthy that one previous study mentioned that CKD patients who underwent exposure to ICM during CT scans more than once per year may develop ESRD earlier in the time scale of years (Hsieh et al., 2016). However, CKD patients with concurrent usage of ESA were excluded from their study due to proposed short-interval deterioration of residual renal function. Therefore, these pre-dialysis patients are the focus of our scrutiny in determining whether the time-interval to dialysis requirement after intravenous ICM exposure shortens. The result, however, included only small numbers of patients as the median time decreased from 38 to 12 days. It was inconclusive while containing wide interquartile ranges. Further investigation is warranted to clarify whether contrast exposure truly shortened the time to dialysis requirement in advanced CKD patients.

This is the first national database study to describe CKD patients receiving CT with ICM is not associated with higher risk of dialysis in the short-term. The advantage of this study is the conduction of a nation-wide population-based design to minimize potential single institution related biases. Whether the risk of CI-AKI varies among different races remains unclear (Chawla et al., 2017; Powell et al., 2018), yet there are no previous articles aimed specifically at Asians before this current study. In addition, it is unknown if CKD patients identified from administrative data can be viewed as a risk factor in terms of dialysis requirement within 30 days before our present study.

Considering that the information was extracted from a national research database, the cohort size appeared smaller than expected; but it was also a result of our rigorous selection criteria. Even if, our results offered valuable information in clinical practice. Although the relatively small cohort size may decrease the statistical power and increase the risk of a false negative result, our case numbers of dialysis events to be analyzed were larger than previous comparable studies showing the same results (McDonald et al., 2014, 2015).

This study still has several other limitations. The recognition of diagnoses including CKD and comorbid conditions were based on ICD-9-CM diagnostic codes, and cases of mis-registration cannot be excluded. The indications for CT scans were unknown, and acute medical illnesses were also not analyzed due to database restrictions. Details on the volume and specific ICM injected were not retrievable in our dataset, but the majority of ICM used were LOCM among routine clinical practice in Taiwan. HOCM has now become obsolete, and IOCM is rarely used. There is no sufficient evidence to suggest a significant CI-AKI risk difference between LOCM and IOCM (Eng et al., 2016), and our results will unlikely change when considering that few patients actually received IOCM. And lastly, this is a retrospective cohort study, which inevitably contains biases inherently related to the study design. For further investigation, a well-designed prospective cohort study in CKD patients comparing those receiving CT with or without ICM can be conducted. The type and volume of ICM, acute medical conditions, nephrotoxic medications, and preventive measurements should be controlled. Novel biomarkers to predict CI-AKI and dialysis therapy requirement could also be studied in the future.

Conclusions

In our population-based study, the percentage of CKD patients who require dialysis within 30 days of undergoing contrast enhanced CT scans is 4.1%, a lower incidence compared with subjects who receive non-contrast enhanced scans. However, no statistically significant difference is observed after adjustments are made for other baseline conditions. Patients with CKD identified via the application of administrative data, therefore, do not appear to be at risk of dialysis within 30 days after CT with ICM studies.