Pharmacological prevention of serious anaphylactic reactions due to iodinated contrast media: systematic review

Search strategy

We searched for reports of trials (without language restrictions) that tested premedication in patients who received iodinated contrast media. We searched in Medline, Oldmedline, Embase, HealthSTAR, and CINAHL by using a comprehensive list of generic names and brand names of contrast media, steroids, antihistamines, and ephedrine and key words for adverse drug reactions (for instance, allerg$ or hypersensitiv$) and for prevention and prophylaxis (for instance, prevent$ or protect$). The search period was 1950 to October 2005. We searched the Cochrane controlled trials register and reviewed bibliographies of relevant reviews³⁵⁷ and included articles. Our prespecified inclusion criteria were random allocation of patients, use of premedication alone or in combination, presence of a placebo or a no treatment control group, and reporting of presence or absence of allergic reactions.

Outcomes

These trials reported on a large variety of outcomes and combinations of outcomes. We classified them as “distinct allergy related symptoms,” “symptom categories” (grades), “non-specific symptoms,” and “adverse drug reactions.” Two large trials reported exclusively on arbitrary categories of symptom combinations (grades) according to the presumed degree of severity of the symptoms.^{w1 w2} Grade 1 included a single episode of emesis, nausea, sneezing, and vertigo. Grade 2 consisted of hives, erythema, emesis more than once, and fever or chills (or both). Grade 3 comprised clinical shock, bronchospasm, laryngospasm, laryngeal oedema, loss of consciousness, convulsions, decrease or increase in blood pressure, cardiac arrhythmia, angina, angiooedema, or pulmonary oedema.

Data extraction

One investigator (EvE) abstracted the articles; the others independently validated the extracted data. Two investigators, an allergologist (CH) and an anaesthetist (MRT), decided together for each symptom whether it was most likely allergy related, on the basis of the description of typical features reported in the original trials. They also decided for each symptom whether it was serious (that is, potentially life threatening). We regarded symptoms as potentially life threatening when they had the potential to deteriorate towards a status in which ventilatory or haemodynamic support would become necessary (for instance, laryngeal oedema needing tracheal intubation or arterial hypotension needing catecholamines). When symptom localisation was unclear (angio-oedema may be cutaneous or respiratory), we assumed that the more serious (in this case, laryngeal oedema) had occurred. All investigators read all included reports independently to assess the quality of randomisation, concealment of treatment allocation, blinding (patient, caregiver, observer), and patient follow-up. We compared extracted data and resolved disagreements by discussion.

Data synthesis

We calculated dichotomous data as Peto odds ratios (as many zero cells existed) with 95% confidence intervals; an odds ratio < 1 suggested efficacy with premedication. When we combined data from more than two trials, we did formal heterogeneity testing (we considered P > 0.1 to be homogeneous) and calculated the I² value as a measure of the degree of inconsistency between the trial results.8

Some trials reported on several similar outcomes. As more than one symptom could have occurred in the same patient, we had to select one for analysis. We knowingly selected the outcome with the most pronounced treatment effect (that is, with the lowest odds ratio). We assumed that this would create a bias towards an overestimation of the efficacy of premedication; this overestimation would have to be taken into account when interpreting the results of the analyses. We used Stata version 7 for meta-analyses.

Search results

Of 64 potentially relevant reports, we excluded 55; most compared different contrast media (fig 1). We identified one cluster of two reports that originated from a single trial; we regarded the older report as the main article,^w1 and excluded the copy.9 Nine trials, published between 1975 and 1996, met our inclusion criteria (table 1).^w1-w9 The total number of participants was 10 011; 6108 received premedication. All but two reports were in English.^{w4 w9} The median trial size was 299 patients (range 66-6763). In no report was an adequate randomisation method described, and only in one was treatment allocation concealed.^w1 In four reports, evidence existed of adequate blinding of patients, caregivers, and observers.^{w1-w3 w8} No report described a complete patient follow-up that enabled an intention to treat analysis.

Fig 1

Flowchart of retrieved, excluded, and analysed reports

• Download figure
• Open in new tab
• Download powerpoint

Patients

All trials included adults. In three, patients with a history of reaction to iodinated contrast media were excluded.^{w1 w3 w6} Bertrand et al also excluded patients with a history of allergy, atopy, or drug hypersensitivity.^w3 The other six did not specify exclusion criteria and only inconsistently specified the percentage of patients who had a history of at least one previous serious reaction or of those who had never had iodinated contrast media.

Premedications

Five trials tested H₁ antihistamines (hydroxyzine,^w3 clemastine,^{w6 w9} chlorpheniramine,^w7 dimenhydrinate^w8), five tested corticosteroids (betamethasone,^w4 dexamethasone,^w5 methylprednisolone,^{w1 w2} prednisolone^w6), and one tested an H₁-H₂ combination (clemastine-cimetidine)^w6 (table 1). None tested a steroid-antihistamine combination, and we found no reports on ephedrine.

Contrast media

In eight trials, iodinated contrast media were given intravenously, and in one it was given intrathecally (table 1). Three trials used a non-ionic low osmolar medium (ioxaglate,^w8 iohexol or ioversol^w2 w5). Two trials used an ionic high osmolar medium (amidotrizoat sodium or meglumine^{w6 w9}). Chevrot et al used four different ionic media,^w4 but 92% of the patients received ioxithalamate, an ionic high osmolar medium. In one trial, both an ionic high osmolar medium (meglumine) and a non-ionic low osmolar medium (ioxaglate) were injected.^w3 In one trial, media were specified as ionic only.^w1 Finally, one trial did not specify the type of medium.^w7

Radiological interventions

Radiological interventions were urography in five trials,^{w2-w4 w6 w9} computed tomography scan in three trials,^w2-w4 venography or arteriography in three trials,^{w3 w4 w8} and myelography,^w5 pyelography,^w7 or cholangiography^w9 in one trial each (table 1). In one trial, the type of imaging procedure was not specified.^w1

Outcomes

Table 2 presents the large variety of outcomes that were reported in these trials.

Distinct allergy related symptoms

Three trials reported on arterial hypotension. We regarded hypotension as potentially life threatening, although none of the reports provided a definition of hypotension. Intravenous dimenhydrinate 25 mg,^w8 intravenous betamethasone 8 mg,^w4 and oral methylprednisolone 2×32 mg were tested.^w2 None of the 689 patients who received steroids had hypotension compared with 3/687 (0.4%) controls (odds ratio 0.14, 95% confidence interval 0.01 to 1.30) (fig 2).

Fig 2

Distinct haemodynamic, respiratory, and cutaneous symptoms. Hypotension, bronchospasm, angio-oedema, and laryngeal oedema were considered to be potentially life threatening.

Anti-H=antihistamine. *P for heterogeneity=0.03, I²=62%

• Download figure
• Open in new tab
• Download powerpoint

Three trials reported on respiratory symptoms. Oral hydroxyzine 100 mg,^w3 intravenous clemastine 0.03 mg/kg and cimetidine 2-5 mg/kg,^w6 oral methylprednisolone 2×32 mg,^w2 and intravenous prednisolone 250 mg were tested.^w6 The outcomes were bronchospasm, angio-oedema (with urticaria), and laryngeal oedema; we regarded all as potentially life threatening. Four of 391 (1%) patients who received an antihistamine had a respiratory symptom compared with 9/394 (2.3%) controls (odds ratio 0.46, 0.15 to 1.39) (fig 2). Of patients who received a steroid, 3/778 (0.4%) had a respiratory symptom compared with 11/769 (1.4%) controls (odds ratio 0.31, 0.11 to 0.88). Finally, in one trial, 1/196 (0.5%) patients who received the clemastine-cimetidine combination had angio-oedema compared with 8/194 (4.1%) controls (odds ratio 0.20, 0.05 to 0.76).^w6

Six trials reported on cutaneous symptoms. Oral hydroxyzine 100 mg,^w3 intravenous dimenhydrinate 25 mg,^w8 subcutaneous chlorpheniramine 10 mg,^w7 intravenous clemastine 2 mg^w9 or 0.03 mg/kg,^w6 intravenous prednisolone 250 mg,^w6 and oral methylprednisolone 2×32 mg were tested.^w2 The outcomes were urticaria, erythema, hives, pruritus, and flushing; we regarded none of these as life threatening. Fourteen of 711 (2%) patients who received an antihistamine had cutaneous symptoms compared with 49/801 (6.1%) controls (odds ratio 0.36, 0.22 to 0.60; P for heterogeneity = 0.03, I² = 62%) (fig 2). With steroids, 5/778 (0.6%) patients had cutaneous symptoms compared with 15/769 (2%) controls (odds ratio 0.36, 0.15 to 0.87).

Symptom categories (grades)

Among the three grades of symptoms, we regarded only grade 3 reactions as potentially life threatening (table 2, fig 3).^{w1 w2} Grade 1 reactions were significantly reduced with the double dose but not with the single dose methylprednisolone regimen: 87/3093 (2.8%) patients who received the double dose regimen had a grade 1 reaction compared with 89/2178 (4.1%) controls (odds ratio 0.62, 0.46 to 0.84) (fig 3). Grade 2 reactions were not significantly reduced, either with the single dose or with the double dose methylprednisolone regimen. Grade 3 reactions were significantly reduced only with the double dose regimen: 7/3093 (0.2%) patients who received the double dose regimen had a grade 3 reaction compared with 20/2178 (0.9%) controls (odds ratio 0.28, 0.13 to 0.60).

Fig 3

Arbitrary symptom combinations (“grades”) as defined in the original reports.^{w1 w2} Grade 1=single episode of emesis, nausea, sneezing, or vertigo; grade 2=hives, erythema, emesis more than once, or fever or chills (or both); grade 3=shock, bronchospasm, laryngospasm or laryngeal oedema, loss of consciousness, convulsions, fall or rise in blood pressure, cardiac arrhythmia, angina, angio-oedema, or pulmonary oedema. Grade 3 was considered to be potentially life threatening.

MP=methylprednisolone (oral)

• Download figure
• Open in new tab
• Download powerpoint

Non-specific symptoms

The trials reported on a large number of symptoms that we did not consider to be allergy related or life threatening (table 2). We did not analyse these any further.

Adverse drug reactions

Patients who received dexamethasone reported dyspepsia, insomnia, bitter taste, nausea, and headache (table 2).^{w2 w5} Patients who received hydroxyzine or clemastine reported bad taste, local skin reaction, and somnolence.^{w3 w9}

Potential limitations

Our analysis has some limitations. Firstly, none of the trials included only patients with a history of allergic reactions. It may be argued that premedication is given exclusively to high risk patients, such as those with a history of allergic reaction to iodinated contrast media, and that in these selected patients premedication will certainly be beneficial. However, although three trials excluded patients with a history of a reaction to iodinated contrast media, six did not; we may thus assume that these trials represent daily clinical practice. It is unlikely that conscious selection of low risk patients in the original trials was responsible for the disappointing degree of efficacy of premedication. Also, premedication may not necessarily be efficacious in a patient with a positive history. Indeed, breakthrough reactions after steroid premedication have been observed.12 In a retrospective survey, breakthrough reactions were similar to the patients' initial reactions in 85% of cases and severe or life threatening reactions occurred in 24% of the cases.13

Secondly, a large variety of symptoms were reported. The decision as to whether a symptom was allergy related, and whether a symptom could be regarded as potentially life threatening, had to be made on the basis of clinical features that were described in the original reports; we cannot rule out selection bias. Six trials reported on symptoms that could be regarded as potentially life threatening.^{w1-w4 w6 w8} Although bronchospasm or laryngeal oedema may represent a certain threat to the patient, most of the reported symptoms were clinically of minor importance. Curiously, grade 1 and grade 3 reactions were significantly reduced but grade 2 reactions were not. However, for composite outcomes to be appropriate the individual symptoms need to be well defined, to be of equal importance, and to occur with similar frequencies; additionally, the active treatment needs to lead to a similar reduction in the risk of all components.14 No evidence exists to show that this is the case for the three arbitrarily defined gradings that were reported in the two largest trials.^{w1 w2}

Thirdly, contrast media that are used today may have a more favourable risk profile than the tested iodinated contrast media. In a large scale Japanese survey,11 severe reactions occurred in about 0.2% of patients with high osmolar iodinated contrast media and in only about 0.04% with low osmolar non-ionic contrast media. However, when the baseline risk is very low, as with low osmolar non-ionic contrast media, prevention of allergic reactions is likely to become even less efficacious.

Finally, the average quality of these trials was limited. We do not know whether this was because of inadequate conduct or reporting of the trials. All were published before the first CONSORT checklist.15 Low quality trials are prone to bias. Empirical evidence shows that, for instance, selection bias due to inadequate randomisation or observer bias due to lack of blinding may lead to an overestimation of the effect of a treatment.16 If such biases existed in these trials, we may infer that our effect estimates even overestimated the true degree of efficacy.

Arguments for and against premedication

An argument in favour of premedication is that serious non-fatal anaphylactic reactions may contribute to major morbidity, prolonged hospital stays, and excess cost. In hospitals with limited resources and personnel who infrequently manage reactions to iodinated contrast media, an even larger number of patients that need to be treated for one to benefit may justify the use of premedication. Some may argue that our analyses are simplistic as they do not consider the potentially additive or synergistic effect of drug combinations. Indeed, drug combinations have been widely advocated³⁷ and have been recommended in guidelines.⁴⁶ H₁ antihistamine mainly showed efficacy against cutaneous symptoms, and steroids prevented respiratory symptoms. Thus, combining these drug may be justified to further improve efficacy. This, however, needs to be formally tested in randomised trials as none of the retrieved trials tested a steroid-antihistamine combination.

Arguments against premedication include the cost and the risk of doing more harm than good to the patients. Although an oral double dose steroid regimen may not be excessively expensive, a large number of patients need to be treated for one to benefit. Potentially life saving radiological interventions may be delayed by prolonged drug prophylaxis with doubtful efficacy. Pretreatment may create a sense of security among people who inject contrast media. Healthcare providers may neglect appropriate measures to survey patients and to treat anaphylaxis. Finally, the drugs used may cause harm. No serious problems were reported with steroids. However, in susceptible patients, even a single steroid dose may lead to considerable adverse effects.17 Incremental cost effectiveness ratios, taking into account all potential costs (drug cost, prolonged hospital stay, delay in diagnosis, unplanned admission to the intensive care unit) would be needed for rational decision making.

Conclusions

The existing evidence suggests that a large number of unselected patients need to receive an oral double dose of methylprednisolone to prevent a potentially life threatening, iodinated contrast medium related reaction in one of them. For antihistamines, limited evidence shows that they may prevent some reactions. Valid data supporting the efficacy of drug combinations or the use of premedication in patients with a history of allergic reactions are completely lacking. Severe allergic reactions due to contrast media seem to be rare; this may explain why no reports of disastrous reactions exist. Allergologic evaluation including cutaneous tests and biological assessment to identify and eliminate the culpable agent has been proposed in all patients with a history of an anaphylactic reaction to a contrast medium.18 However, only a fraction of patients with severe and life threatening reactions have a positive skin test for the administered contrast material.2 Another approach would be to train physicians who are using iodinated contrast media to recognise and treat anaphylactic reactions appropriately.¹⁹²⁰ Accordingly, radiology departments should be staffed with the necessary equipment for resuscitation.^w4 Physicians dealing with patients receiving contrast media should not rely on the efficacy of premedication; routine prophylaxis should be abandoned.