Studies considered in this chapter

In this chapter validation studies for diagnostic criteria are considered. Validation studies for the UK Working Party’s diagnostic criteria were identified. Although other diagnostic criteria for atopic eczema have been described, such as the Lillehammer criteria and questionnaires used for epidemiological studies, no validation studies were identified for these criteria.

No evidence comparing outcomes for children diagnosed with atopic eczema using different criteria was identified. Studies comparing epidemiological data obtained by using different diagnostic criteria are not relevant to this section.

Overview of available evidence

The UK Working Party criteria were developed by comparing observations made by two observers (dermatology registrars or senior registrars) using 31 of the Hanifin and Rajka criteria, with the definitive diagnosis of atopic eczema being made by a physician with an interest in dermatology.²⁷ The observers were unaware of the true purpose of the study. Sixteen physicians were involved in the study, 13 of whom had a special interest in atopic eczema, including six paediatric dermatologists. The study population consisted of consecutive new cases of ‘typical mild to moderate atopic eczema’ (patients aged 6 months to 50 years) and two control groups (patients with an inflammatory skin disorder other than atopic eczema attending the clinic, and patients from the community with no overt skin disease; total n = 224, 120 cases and 104 controls). Overall, 53% of the cases were aged under 10 years; 35% of the total study population were aged under 10 years and 46% were aged under 16 years. Cases were significantly younger than controls (P < 0.01). The study population was predominantly white (82%), and the ethnic origin of the remaining individuals was the Indian subcontinent (5%), Afro-Caribbean (9%), oriental (3%), and ‘other’ (1%). Non-white people were significantly under-represented in the control group (P = 0.01).²⁷ [EL = 2+/DS II]

The sensitivity and specificity of each criterion was calculated using the physician’s diagnosis as the gold standard and the observer’s diagnosis as the ‘test’. Regression techniques were used to derive the minimum set of criteria that best discriminated between cases of atopic eczema and controls; these techniques included the chi-squared test, consideration of the intra-observer reliability, and the sensitivity and specificity values. Six criteria were found to provide good separation of atopic eczema cases from controls, namely:²⁷

• history of flexural dermatitis
• history of dry skin
• onset under the age of 2 years
• history of a pruritic skin condition (‘presence of an itchy rash’)
• personal history of asthma
• visible flexural dermatitis.

The investigators also explored whether the six criteria were influenced by ethnic group. They reported that there was no evidence of a difference, but no data were presented.

The proposed composite criteria (itchy skin as a major criterion, with three or more of the other five criteria) were validated in studies undertaken in outpatient settings.²⁸ [EL = DS Ib] The populations considered were dermatology outpatients (27% of whom were children aged 10 years or under) and paediatric outpatients. While the dermatology outpatients study included some data for children within the age group of interest to this guideline, no demographic data were provided and therefore that part of the study is not considered further.

Some criteria were modified after the dermatology outpatients validation study. In younger children the criteria age of onset under 2 years and personal history of hay fever may not be applicable. Therefore, for children under 4 years, the criterion onset under 2 years was not used, and history of asthma/hay fever was replaced with history of atopic disease in a first-degree relative. In addition, because distribution of atopic eczema may be different in young children, visible dermatitis on the cheeks and/or the outer aspects of the limbs were included as part of ‘visible flexural dermatitis’ in children under 4 years, and ‘history of flexural dermatitis’ included dermatitis on the cheeks in children under 10 years.²⁸

The paediatric outpatient study, conducted in the London area, included 114 children aged up to 16 years (39 children with atopic eczema and 75 controls). The median ages of cases and controls (interquartile range) were 5 years (2–10 years) and 6 years (3–9 years), respectively. Overall, 51% were female, 51% were white, 27% were Afro-Caribbean, 11% were from the Indian subcontinent, and 11% were Chinese, Middle Eastern or of mixed race. Control groups had conditions such as other inflammatory dermatoses or infections.²⁸

The conclusion was that optimal discrimination was given by itch plus three or more other criteria. The sensitivity of these diagnostic criteria was 85% (95% confidence interval (CI) 60% to 94%) and the specificity was 96% (95% CI 89% to 99%).²⁸ This indicates that 85% of children diagnosed with atopic eczema by a dermatologist were also diagnosed with atopic eczema using the composite criteria. The specificity value indicates that 96% of those who were not diagnosed with atopic eczema by a dermatologist were also not diagnosed with the condition using the composite criteria. When the specificity is very high, the rate of false positives is conversely low. Therefore, a positive test result implies a correct diagnosis. The sensitivity and specificity of the composite criteria were considered to be similar in the Afro-Caribbean subgroup to those in the total population.²⁸

Validation studies of the UK Working Party’s diagnostic criteria for atopic dermatitis have also been undertaken in community populations (schoolchildren in London,³⁰ Romania,³¹ and South Africa,³² and in Scottish infants aged 1 year³³). There was one study in a clinical setting in India.³⁴ Other validation studies identified have included both children and adults, but do not report data separately for children and therefore are not considered further.^35,36

The validation studies tended to focus on the predictive value of individual criteria and of composite criteria (itch plus a number of other criteria). In the South African study,³² questionnaires including all six questions were administered by fieldworkers. [EL = DS III] In the other studies, parents, children or schoolteachers completed questionnaires that included five of the six UK Working Party criteria. A nurse independently assessed whether the sixth criterion (visible flexural dermatitis) was present. The diagnostic accuracy of each criterion was then compared with the diagnosis made by a dermatologist (regarded as the gold standard diagnosis).^30,31 [EL = DS II]

The studies in schoolchildren in London (n = 695) and Romania (n = 1114) were identical in design. The London children were aged 3–11 years and included a range of ethnic groups (43% white, 8% Indian subcontinent, 32% black, 15% mixed, 2% other).³⁰ The Romanian children were aged 6–12 years and were predominantly white Romanian (98%), the remainder being Gypsy (1%), mixed race (1%), or ‘other’ (0.1%).³¹ The prevalence of atopic eczema in the London and Romanian school children was 8.5% and 2.4%, respectively.

From these studies, the composite criterion of itch plus three or more other criteria was regarded as providing the best diagnostic information (that is, providing the best separation of cases from non-cases). Compared with a dermatologist’s diagnosis, the composite criterion provided the following diagnostic accuracy statistics:

• sensitivity 70%, specificity 93%, PPV 47%, NPV 97% in London schoolchildren³⁰
• sensitivity 74%, specificity 99%, PPV 63%, NPV 99% in Romanian schoolchildren.³¹

The results show that the level of agreement for a negative diagnosis is high. The relatively low PPVs reflect the low prevalence of atopic eczema in the study populations. It is expected that in clinical situations where the diagnostic criteria are to be used that the prevalence would be much higher and therefore the PPV would also increase.

The validity of the criteria in certain subgroups (including groups based on age and ethnicity) was also explored, although results were given only for those aged under 4 years and according to severity. The study in London schoolchildren also considered the retest reliability of the questionnaire in 73 cases. Kappa scores were above 0.85, indicating a good level of agreement between first and second questionnaires.³⁰

The South African study comprised Xhosa-speaking schoolchildren (n = 3067, age 3–11 years) from urban, peri-urban and rural areas.³² The original questionnaire was translated into Xhosa, validated in a pilot study and administered by a bilingual interviewer. For the UK diagnostic criteria, specificity was high (97.9%, 95% CI 97.3% to 98.4%). Sensitivity of 43.7% (95% CI 26.3% to 62.3%) means that over half of the children diagnosed with atopic eczema by a dermatologist were misclassified by the diagnostic criteria. The single criterion of visible flexural eczema had sensitivity of 81.2% (95% CI 63.5% to 92.7%) and specificity of 99.0% (95% CI 98.6% to 99.3%), implying that this criterion alone has the ability to distinguish between cases and non-cases in this population. The prevalence of atopic eczema in this group was 1.0% (95% CI 0.6% to 1.4%).

The validation study of infants in Scotland considered level of agreement (percentage and kappa scores) between a parent’s and a nurse’s diagnosis of atopic eczema in cases and controls using the UK Working Party’s criteria (n = 108).³³ [EL = 2+] Parents completed a postal questionnaire listing the criteria. The percentage agreement for five of the six criteria ranged from 88% to 97% (kappa scores 0.75 to 0.94). (The criterion ‘onset in age under 2 years’ is irrelevant in this study because the entire study population was aged under 2 years). The levels of agreement between mothers and nurses for composite criteria were 96% for itch plus three or more other criteria, and 94% for itch plus all UK criteria.³³

The study in India³⁴ (n = 149, age 2 months to 14 years) compared the Hanifin and Rajka criteria and the UK Working Party diagnostic criteria with each other and with clinical diagnosis by a dermatologist. A questionnaire was designed that included all of the features of both sets of criteria. This questionnaire was administered to 101 children with atopic dermatitis and 48 children with other skin conditions. It was not stated whether the clinical diagnosis was known by the interviewers. The UK Working Party diagnostic criteria were found to have high sensitivity (86%) and specificity (96%). [EL = DS III]

Evidence statement for diagnosis

A range of diagnostic criteria for atopic eczema in children have been described in the literature, but only the UK Working Party criteria have been assessed adequately for validity and repeatability. The use of composite criteria of itch plus another three or more of the five criteria is considered to provide optimal separation of children with or without the condition. In validation studies in European children aged 1–12 years, the UK Working Party criteria provided a valid tool for diagnosing atopic eczema in community settings. [EL = 2+/DS II] In the South African study, the composite criteria did not distinguish cases from non-cases adequately, although the single criterion of visible flexural eczema did. [EL = DS III] The high specificity in all of the validation studies means that the false positive rate is low and therefore a diagnosis of atopic eczema according to the UK working party criteria should be believed.

Cost-effectiveness

Published evidence relating to the cost-effectiveness of diagnostic criteria was not sought because the use of healthcare resources was not the focus of the clinical question.

From evidence to recommendations

In the absence of outcome data for any diagnostic method, the GDG consensus view was that the UK Working Party’s diagnostic criteria would help clinicians with little knowledge or experience of dermatology to diagnose atopic eczema in children. Using the diagnostic criteria may also optimise the use of consultation time.

It is the GDG’s view that the proposed diagnostic criteria apply to all ethnic groups, although it is recognised that there are differences in the pattern of atopic eczema among different ethnic groups. For example, in children of black African, black Caribbean or Asian origin atopic eczema may present on extensor surfaces as well as on flexures, and is more likely to produce lichenification (thickening of the skin), lumpy or papular skin (papular or follicular eczema) and a change in pigmentation. [EL = 4]

The potential impact of using the proposed criteria on consultation time for diagnosis was considered by the GDG. The likelihood is that using diagnostic criteria such as these would focus history taking and physical examination compared with not using formal criteria, and therefore would not increase consultation time or cost.

The GDG also believes that taking a thorough history that includes questions about potential trigger factors and the presence of other atopic diseases is an important step in the management of atopic eczema in children.