Drug provocation tests
A DPT was conducted to evaluate tolerance to a non-CR alternative ICM, using a volumetric pump (Alaris© GW, BD©), during a 1-day hospital stay. The selection of ICM for the DPT was based on the following criteria: (i) the implicated ICM was avoided, even if ST was negative; (ii) ICM with positive ST were avoided; (iii) ioversol or iobitridol were selected if iohexol or iodixanol were implicated, due to the high C-R between the latter two, unless specifically requested otherwise by radiologists.
Blood pressure, heart rate, and oxygen saturation were monitored for up to 3 hours after the DPT. No premedication was administered. The ICM dose and infusion rate were based on a rapid DPT previously reported37. Specifically, an ICM dose of 100 cc was administered intravenously: 30 cc at 900 cc/h for 2 minutes, followed by 70 cc at 420 cc/h for 10 minutes, completing the DPT in a total of 12 minutes. If any symptoms were observed during or immediately after DPT, the patient’s vital signs were checked, and a physical examination was performed. Cessation of ICM infusion and treatment were performed as necessary.
DPT was considered positive if any of the following occurred within 96 hours after the provocation: a decrease in oxygen saturation below 90%, a drop in blood pressure greater than 30%, or the appearance of skin, gastrointestinal or respiratory symptoms. Patients could report any adverse reactions occurring within the following 7 days.
After a positive DPT, a new DPT was conducted with one of the remaining ICMs after a washout period of at least seven days. Well-tolerated ICM was recommended as alternative for use without premedication. When all ST or DPT were positive, avoiding ICM was recommended. A nephroprotection protocol based on previous recommendations was implemented42.
Statistical analysis was performed using the Minitab© version 18 software package for Windows (Minitab LLC, State College, Pennsylvania, USA). The mean, median, standard deviation, and minimum and maximum values were used to describe quantitative variables, while qualitative variables were represented by frequencies and percentages.
For quantitative variables, the Student’s t-test was used when a parametric test was required for comparing means between independent samples. If the samples did not follow a normal distribution, two types of non-parametric tests were applied: the Kruskal-Wallis test or the Mann-Whitney test when group distributions were similar, and the Mood’s median test when the distribution shapes suggested dissimilarity between groups. In these cases, quantitative variables were expressed as medians. Results were considered statistically significant when p<0.05.
Group analysis and patient segmentation were conducted using unsupervised clustering techniques, utilizing algorithms and functions from the Bayes LCA© Library (Bayesian Latent Class Analysis). This technique divides the population into N groups, aiming to achieve maximum homogeneity within each group and maximum heterogeneity between the groups. The most effective variables were selected to achieve a more homogeneous distribution of patients in the clustering study.
Based on the results obtained, we proposed a mathematical model to predict the outcomes of ST and DPT, considering various variables related to both patient characteristics and the HSR to ICM . A binary logistic regression model was applied, with predictor variables selected based on their highest statistical significance and optimal variance distribution. For ST, the variables were involved iodixanol, involved iohexol, isolated cutaneous symptomatology, previous exposition to ICM and history of atopy . For DPT, the variables were cardiac disease, oncologic disease, isolated cutaneous symptoms and positive ST results.
To evaluate the predictive capacity of the proposed mathematical models, several goodness-of-fit tests were applied, including R-squared (R²), Deviance, Pearson, and Hosmer-Lemeshow tests.
RESULTS
A total of 130 patients with a history of HSR to at least one CM were prospectively enrolled over a six-year period, from January 2014 to July 2019. The median age was 64 years (IR:23–73), with 82 women (64%). Most patients (90%) had comorbidities that likely necessitated radiological examinations involving ICM, with oncological diseases (40%) and cardiovascular diseases (15%) being the most common.
Thirty-eight patients (29%) were atopic, defined as being allergic to aeroallergens (63%), foods (19.5%), or drugs (13.4%), or as having allergic contact dermatitis (4.5%). Additionally, 18.4% of these patients had multiple types of allergies. The prevalence of atopy was higher in women (p<0.05;OR 3.32).
The radiological explorations associated with HSR included computed tomography (CT) in 68.5% of patients, vascular studies in 20.8%, urography in 10.8%, and endoscopic retrograde cholangiopancreatography in one patient. Vascular studies were more associated with moderate delayed reactions, CT with mild immediate HSR, and urography with the most severe episodes (p<0.001; Cramer’s V=0.21). These differences are likely more related to the type of ICM used during the procedure rather than to the inherent risk of the radiological technique itself.
Immediate HSR occurred in 75 patients (58%), with 47 cases (63.5%) classified as mild, 10 cases (13.5%) as moderate, and 17 cases (23%) as severe (anaphylaxis). Among the remaining 55 patients with delayed HSR (42%), 34 cases (61.8%) were mild, and 21 cases (38.2%) were moderate in severity. There were no cases of severe delayed HSR.
Ninety-six patients (74%) exhibited exclusively cutaneous symptoms, with 43.7% presenting with immediate urticaria and 56.3% with delayed maculopapular exanthema.
The ICM involved in HSR were iohexol in 45 patients (34.6%), iodixanol in 26 patients (20%), iopromide in 3 patients (2.3%), ioversol in 2 patients (1.55%), iomeprol in 1 patient (0.7%), and unknown ICM in the remaining 53 patients (40.7%).
The type of HSR was influenced by the specific ICM used: iodixanol, iohexol, ioversol and iomeprol were predominantly associated with exclusively cutaneous symptoms, whereas iopromide was linked to a higher risk of anaphylaxis (p<0.05; Cramer’s V=0.22). The use of unknown ICM was significantly associated with an increased risk of immediate HSR (p<0.05) and a greater severity (p<0.001), with 43.4% of these patients presenting with extracutaneous symptoms and 20% experiencing anaphylaxis. Moreover, iodixanol was more often linked to delayed reactions.
The median time to study initiation was 6 months (IR:2–96), with statistically significant differences depending on the type of HSR (p<0.001): 18 months (IR:2–180) for immediate HSR and 4 months (IR:2–12) for delayed HSR. The delay was also longer when the involved ICM was unknown. No significant differences were observed between the ST results and the delay in initiating the study.
All SPT were negative, but 9 patients (7%) showed positive IDR (4 in the immediate readings and 5 in the delayed readings). The characteristics of the patients with positive ST are described in Table 1.The likelihood of obtaining a positive ST was higher when iodixanol was involved (p<0.05; Cramer’s V=0.33). The calculated negative predictive value (NPV) of ST was 87.6%.
Seventy patients (53.8%) experienced HSR upon their first exposure to ICM and 33% of the patients with positive ST had experienced the HSR during their first contact with ICM.
A total of 141 DPT with an alternative negative skin-tested iodinated ICM were performed on 129 patients, including 17 patients with a history of anaphylaxis. One patient could not undergo a DPT because they tested positive with all the ICM evaluated. The DPT was negative in 85.8% of cases (95% CI: 66.9%–95.9%). However, 20 DPT, involving 16 patients, yielded positive results despite the use of a negative skin-tested ICM. Three patients exhibited positive DPT with more than one ICM. All positive DPT were of a cutaneous type and mild in severity, resolving spontaneously or after treatment with antihistamines and corticosteroids. Two patients with a history of anaphylaxis had positive DPT, with both cases exhibiting mild symptoms (immediate urticaria). Following the DPT, the NPV of ST was calculated to be 87.6%. The characteristics of patients with positive DPT are described in Table 2.
After completing the allergy study, a safe alternative ICM, administered at a high-flow rate without premedication, was recommended for 122 patients (94%): ioversol in 55 patients (45%), iohexol in 48 patients (39.3%), iobitridol in 12 patients (9.8%), and iodixanol in the remaining 8 patients (5.9%). Fifty of these patients subsequently required additional radiological examinations, demonstrating good tolerance to the recommended ICM without premedication. Avoidance of ICM was maintained in only 8 patients: one patient due to positive ST with all tested ICM, and the other seven due to positive DPT with tested ICM.
After the statistical analysis conducted using data mining techniques, the patients were categorized into three well-differentiated clusters (Figure 1). The most common values for the various variables within each cluster, representing the significant characteristics of each, are detailed in Table 3. We compared the percentage of positive DPT in the total population of 129 patients (13.1%) with those obtained in each of our clusters. It was observed that the risk of developing a positive DPT in the allergy study was similar in cluster 3 (13.3% vs. 13.1%), higher in cluster 2 (18.3% vs. 13.1%), and lower in cluster 1 (8% vs. 13.1%)
A binary logistic regression was conducted to identify independent factors predicting the occurrence of a positive ST or a positive DPT. In the regression analysis for ST, the variables with the greatest statistical significance were the type of reaction, symptomatology, the ICM involved, previous exposure to ICM, and history of atopy. The only significant risk factor identified for a positive ST was the involvement of iopromide, with an OR of 13.79 (90% CI: 0.9–194.1). In the regression analysis of DPT, the variables considered included personal history, symptomatology, and ST results. A personal history of cancer was identified as the only risk factor for a positive DPT, with an OR of 2.53 (90% CI: 0.98–6.52).
The ROC curve and regression equation were calculated, resulting in an area under the curve (AUC) of 0.79 for ST and 0.64 for DPT. The goodness-of-fit for these regressions was evaluated using several statistics: the coefficient of determination (R²), and the Pearson, Deviance, and Hosmer-Lemeshow tests, with results displayed in Figure 2 for the ST and Figure 3 for the DPT.