This is the first large-scale UK study to look at the effects of demographics, medical history, and medication use on the rates of MDIS. The MDIS cohort was compared to patients with one or two documented drug allergies in order to ascertain factors linked to multiple drug intolerances. The majority of patients in the control group were assumed to represent a cohort with true type 1 IgE-mediated hypersensitivity reactions. As such, comparing the MDIS cohort to this group of patients enabled the identification of risk factors specifically for the development of multiple drug intolerance rather than single drug allergies in general.
Among the drug allergies investigated, 18 out of 19 drug groups were shown to be significant risk factors for MDIS, with quinolones, cephalosporins, tetracyclines, and ACEi being identified as the most significant predictors. “Other antibiotic” allergies also formed a significant risk factor for MDIS. This is likely to be a result of the study population we used, all of whom were inpatients tending to be prescribed a wide spectrum of antibiotics during their hospital stay. Interestingly, with the exception of ACEi, the most significant risk factors for MDIS were allergies to broad-spectrum antibiotics prescribed for short-term use. This may suggest that mechanisms underlying MDIS occur with a short latency period.
Penicillin allergy is commonly reported among UK patients [16]. As such, we expected that a penicillin allergy may increase the risk of being intolerant to multiple drugs. Indeed, Smith et al. [17] found that a history of allergy to other drugs was almost three times as common in patients who were penicillin-allergic compared to those who were not [5]. However, we found that the frequency of MDIS cases did not differ significantly between patients with penicillin allergy compared to those without. The frequency of MDIS was greatest in those allergic to broad-spectrum antibiotics and these drugs were also the most significant risk factors for MDIS. Penicillins, however, did not fit this trend. This could potentially be explained by the fact that reported penicillin allergies are likely to represent true IgE-mediated hypersensitivity reactions.
Consistent with previous studies [6, 18], we found that even after adjusting for all other demographics, female gender is a significant risk factor for MDIS. This finding could be linked to gender differences in healthcare use. It is well known that women have higher healthcare utilisation than men [19]. Women are therefore liable to be exposed to a larger range of drugs, making them more likely to report drug allergies and be identified as intolerant.
MDIS patients were found to be significantly older than non-MDIS patients. Older patients are known to have more comorbidities and therefore likely to have greater exposure to drugs, which increases the likelihood of a reported adverse reaction. The fact that age was not significant when included in a multivariable model with the presence of comorbidities, suggests that the presence of increased comorbidities with age, rather than age independently, is the more important predictor of MDIS. These findings support previous work carried out by Onder et al. [20], who highlighted the significance of multiple comorbidities as predictors of adverse drug reactions.
Although De Pasquale et al. [8] found an increased likelihood of somatisation in their MDIS cohort, we showed that rates of psychological comorbidities were similar between MDIS and non-MDIS patients. This disparity is likely to be caused by differences in methodology. We used physician-led diagnoses of psychological disorders, whereas De Pasquale et al. [8] used psycho-diagnostic questionnaires to evaluate patients. Somatisation disorders have recently been identified as key risk factors for MDIS [9]. These disorders are difficult to identify and diagnose in clinical practice. This may explain why the prevalence of psychological comorbidities was not high in our MDIS cohort.
Immunological mechanisms may be more important than psychological factors in governing MDIS. We reported more cases of MDIS in those prescribed any antiallergic agent compared to those without such prescriptions. Such immunological mechanisms seem to differ from those involved in systemic drug sensitisations as we showed, like previous reports [18], that atopic comorbidities were not significant risk factors for MDIS.
In contrast to the work carried out by Macy and Ho [4], we found that patients with MDIS tended to be lighter in body weight. This disparity may be explained by selection bias, because it is possible that the heaviest patients would be the least likely to be weighed due to logistical difficulties. Although estimated weights were provided in some cases, these may have been highly inaccurate. Indeed, recent figures from the Health Survey for England 2011 suggest that women tend to underestimate their weights by 3.6 kg on average [21]. Women formed the majority of our study population and MDIS cohort. This may further explain why MDIS cases in our study had lower documented body weights than in previous reports [5].
We also found that weight as a factor was not a significant predictor for MDIS when adjusted for all other demographics, even though heavier patients tend to have multiple comorbidities and, as such, greater medication use [22, 23]. It may be that multiple comorbidities represent the most important risk factor. Deprivation however, was not found to be a significant risk factor for MDIS.
We showed that the likelihood of MDIS is increased in those with documented allergies to different specific drug classes. This may not be surprising given the definition of multiple drug intolerance, because patients with an allergy to a single drug are more likely to meet the MDIS criteria compared to those without allergies. Importantly, we showed that the propensity for different drug allergies to increase the likelihood of MDIS varies and that penicillin allergies do not share this relationship.
Limitations
The dataset used for this study was based on documented allergies within the PICS system, reported by patients on admission to hospital. This was dependent on complete and accurate documentation by physicians. Some data were excluded owing to the absence of information (e.g., weight), which reduced the size of the dataset.
We are unable to state whether the allergies reported were true type I hypersensitivity reactions. Given that allergy testing is only recommended in patients who experience anaphylactic reactions to a drug, patients who report drug “allergies” are unlikely to have undergone allergy testing unless this manifested as a suspected anaphylaxis [2]. If allergy testing was carried out, however, such tests may have been conducted in hospitals different to that of our study site. Because UK hospitals do not share medical information across sites, use of allergy test data in our study would only represent patients investigated at our study site, potentially producing misleading results.
We used data that was derived from a secondary care setting and assumed that patients would have similar drug allergy reporting habits in primary care. An additional limitation is that prescribing guidelines might prevent the investigation of potential drug relationships. For example, it is well documented that cephalosporins should be avoided in those allergic to penicillins owing to the risk of cross reactivity. As such, patients with a reported penicillin allergy are less likely to be exposed to a cephalosporin and are therefore less likely to report an allergy to this drug class. Most antiallergic agents would have been prescribed in a primary care setting and antihistamines may have been taken over the counter; as such, it is likely that the prevalence of their use was underestimated in this study.
Finally, we calculated the deprivation scores based on household income, which may not fully reflect patient background and education.