Burden of Atopic Dermatitis in Adults and Adolescents: a Systematic Literature Review

Abstract

Introduction

Although previously regarded as a children’s disease, it is clear that atopic dermatitis (AD) is also highly prevalent in adults. Because AD is not associated with mortality, it is usually neglected compared with other, fatal diseases. However, several studies have highlighted that AD burden is significant due to its substantial humanistic burden and psychosocial effects. This study aims to summarize and quantify the clinical, economic, and humanistic burden of AD in adults and adolescents.

Methods

A systematic literature search was performed in PubMed, Scopus, Cochrane, Centre for Reviews and Dissemination (CRD), EconPapers, The Professional Society for Health Economics and Outcomes Research (ISPOR), The National Institute for Health and Care Excellence (NICE), and The Canadian Agency for Drugs and Technologies in Health (CADTH). Studies were included if they reported clinical, economic, or humanistic effects of AD on adults or adolescents, from January 2011 to December 2020. The Grading of Recommendations Assessment tool was used to assess risk of bias for the included studies. Regression models were used to explain the correlation between factors such as disease severity and quality of life (QoL).

Results

Among 3400 identified records, 233 studies were included. Itch, depression, sleep disturbance, and anxiety were the most frequently reported parameters related to the clinical and humanistic burden of AD. The average utility value in studies not stratifying patients by severity was 0.779. The average direct cost of AD was 4411 USD, while the average indirect cost was 9068 USD annually.

Conclusions

The burden of AD is significant. The hidden disease burden is reflected in its high indirect costs and the psychological effect on QoL. The magnitude of the burden is affected by the severity level. The main limitation of this study is the heterogeneity of different studies in terms of data reporting, which led to the exclusion of potentially relevant data points from the summary statistics.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13555-022-00819-6.

Plain Language Summary

Atopic dermatitis is a very common skin disease among children and adults. The disease is nonfatal but may lead to patients and families having a low quality of life and decreased productivity, especially in its severe state. Because atopic dermatitis is more common in children than adults, most published research is directed to studying the effect of the disease on children. Atopic dermatitis affects patients’ health, quality of life, financial state, and productivity. Therefore, our study aims to study and quantify the burden caused by the disease represented in the clinical burden, humanistic burden, and economic burden. We conducted a systematic literature review to determine all relevant studies providing specific values for the burden. The studies included are those providing information on the percentage of patients affected by specific symptoms, costs paid for treatment, number of days of productivity lost due to the disease, and quality-of-life questionnaire results for patients with atopic dermatitis or their caregivers. We analyzed the data from all relevant studies to calculate average values and quantify the burden. The results of our study should help healthcare sector decision-makers in understanding the real effect of the disease on adults and adolescents and rearrange their priorities for treating different diseases based on the specific burden of each disease.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13555-022-00819-6.

Key Summary Points

The burden of atopic dermatitis is significant, mainly owing to its high prevalence.

Itch, depression, sleep disturbance, and anxiety are the most common manifestations among atopic dermatitis patients.

Managing each atopic dermatitis patient costs about 4411 USD annually.

Indirect costs (productivity lost costs) of atopic dermatitis represent more than double its direct costs.

The quality of life of patients with atopic dermatitis is significantly affected by the disease, but the effect is largely dependent on the severity level.

Introduction

Atopic dermatitis (AD) is a nonfatal disease that significantly impairs patients’ quality of life (QoL). According to the global burden of disease study, AD has the highest disability-adjusted life-year (DALY) burden among all skin diseases. Its burden is ranked in the top 15 among all nonfatal diseases, and it is responsible for 0.36% of the total DALY burden of all 359 diseases and injuries analyzed in the study [1]. Compared with other dermatological diseases, AD poses a significantly higher burden. The age-standardized DALY rate of AD is 75% higher compared with psoriasis and 82% compared with urticaria, representing more than twice the burden of any other skin disease [1].

AD is also known as atopic eczema [2] and is a chronic disease that causes painful flares of inflamed, dry, and itchy skin periodically. Patients with AD usually have accompanying allergic disease, such as asthma or hay fever. To date, no cure has been found for AD, but treatments and self-care measures can relieve itching and prevent new outbreaks significantly [3].

Patients with moderate to severe AD often experience flares that negatively affect their productivity at work or school [4]. A cross-sectional study in Iran reported that 50% of dermatology patients suffered from psychiatric comorbidities as well [5]. An international study reported that 32% of participants believed that AD affected their school or work life, and 14% of participating adults believed that their career progression had been hindered by AD [4].

The prevalence of AD started to increase in the last decades of the twentieth century [6], with a prevalence up to 10–20% in children. Although AD had been regarded as a children’s disease, it has become clear that many adults also are affected, with an estimated prevalence of 3–5% in the general population [7].

Estimating the burden of AD on the basis of scientific evidence can help decision-makers make more informed treatment decisions. Understanding the burden of AD may also support public health policies, help to prioritize interventions, and allow for better resource allocation [8]. AD is a nonfatal disease and therefore usually neglected compared with more severe or fatal diseases. However, several studies have highlighted that the burden of AD is significant because of the substantial humanistic burden and psychosocial effects it can cause [9–11].

The aim of this systematic review is to summarize and quantify the clinical, economic, and humanistic burden of AD in adults and adolescents.

Methods

Databases and Literature Search Strategy

We conducted a systematic literature review (SLR) and reported its results according to Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines for reporting SLRs [12]. We searched PubMed, Scopus, the Cochrane library, Centre for Reviews and Dissemination (CRD), and EconPapers for relevant studies. Additionally, grey literature sources were searched, including The Professional Society for Health Economics and Outcomes Research (ISPOR) scientific presentations database, and websites of health technology assessment agencies [The National Institute for Health and Care Excellence (NICE) and the Canadian Agency for Drugs and Technologies in Health (CADTH)]. The search terms were constructed based on two domains: “Atopic dermatitis “and “Burden of disease.” To identify suitable keywords for the search term, Medical Subject Headings (MeSH) terms, Google search, and previous papers on the same topic were used as guidance. These helped to identify relevant search terms and their thesaurus.

We included studies that evaluated any type of burden related to AD. Because the burden of disease is dependent on factors such as prevalence and available treatment options, which vary significantly within 10 years, the literature search was limited to studies published since January 2011. The search was restricted to English-language papers. Although our review focused on adults and adolescents, no age restriction was applied during the literature search to avoid missing potentially relevant studies that were not labeled as containing data for a specific age group. Instead, studies not reporting any data on patients older than 10 years were excluded during the screening and full-text review phases. The detailed search strategy is described in Supplementary Table S1.

Owing to the overlap between databases, search results were first de-duplicated using the embedded feature of EndNote software version X9. Additional duplicates were manually identified and excluded during the screening phase. The snowballing technique was used to add relevant studies from the references cited in the papers found during the SLR. In case of eligibility, the pool of included papers was extended.

Title and Abstract Screening

Studies identified during the literature search were screened by two independent researchers through title and abstract screening. Disagreements were resolved by a third principal researcher. As a first step, the titles and abstracts of all studies were screened using the following predefined exclusion criteria: (1) duplicates, (2) no English abstract, (3) published before 1 January 2011, (4) letters, editorial, case reports, nonsystematic reviews, or animal studies, (5) not related to AD or eczema, (6) not reporting data for patients 10 years or older, and (7) not evaluating the clinical, economic, or humanistic burden of AD (e.g., those investigating treatment efficacy).

Full-Text Screening and Data Extraction

Studies that were eligible for inclusion from the title and abstract screening phase were downloaded, and their full texts were screened. The same previously mentioned exclusion criteria were used, in addition to excluding inaccessible studies and studies with experimental study designs (e.g., clinical trials) because they do not reflect the real-life burden. Other reasons for exclusion were studies in which AD was a comorbidity with other diseases [13] or if there was a confounding effect of a drug other than the usual treatment [14]. In these cases, the burden reported was not solely dependent on AD.

For the included studies, data were extracted in Microsoft Excel. Extracted data were validated by another independent researcher. The general information extracted included number of patients, average age, sex distribution, type of study, and most importantly, whether the study included information about any of the four domains: QoL scoring, humanistic burden other than QoL score, clinical burden, and economic burden. The included studies had data about at least one of the four domains. Risk-of-bias assessment of the studies was performed using the Grading of Recommendations Assessment (GRADE) tool [15]. Each study was assessed for risk of bias by one researcher and revised by another. In case of disagreement, the two researchers discussed to reach a valid decision. A summary of the quality assessment results is shown in Supplementary Fig. S1.

Because of the heterogeneity of the data collected, each domain was extracted in a separate Microsoft Excel sheet. In the clinical and humanistic burden sheets, data were extracted based on the conceptual model developed by Grant et al. [16] to illustrate the clinical and humanistic burden associated with AD in adults and adolescents. Data about signs and symptoms, as well as psychological impact and health-related QoL (HRQoL) impact, were extracted as “mentioned” or “not mentioned.” The number of unique studies reporting the specific impact as part of the results was calculated. In case a clinical questionnaire or assessment tool was used, details were extracted in a multirow format, including subgroup details. Similarly, QoL questionnaire results were extracted. The economic data reported were also extracted in a multirow format, including data about costs, healthcare resource utilization (HCRU), and productivity lost.

Grant et al. [16] categorized the impact of AD as signs, symptoms, mediating factors, proximal impact, and distal HRQoL impact. We adapted the model by recategorizing the same domains under clinical and humanistic burden. Based on the adapted model, clinical burden subgroups were considered to cover psychological impact, signs, and symptoms: (1) psychological impact (depression, anxiety, stress, suicidal ideation, other psychological manifestation), (2) signs (itch or pruritis, burning or heat or tingling sensation, skin sensitivity/sensitivity to sun, soreness/pain/tenderness, skin irritation, skin tightness), and (3) symptoms [redness (erythema), dryness (xerosis), bumps/blisters/papules/vesicles, hardening/flaking, cracking/fissuring, scaling/peeling, thickening/lichenification, bleeding, edema/swelling, other symptoms]. Psychological impact parameters were extracted in both clinical and humanistic burden because they were noted to affect both domains in the studies.

The humanistic burden subgroups included (1) mediating factors (scratching, skin picking), (2) proximal impact (sleep disturbance, lack of concentration, bodily/physical discomfort), (3) distal HRQoL impact (limitation in daily activity, psychological impact, physical limitation, limitation in social/leisure activities, limitation in role: work, limitation in role: school, problems with interpersonal relationships, problems with sexual functioning, suboptimal skin-related health perceptions/cognition, financial burden associated with buying special products), and (4) other humanistic burden manifestations.

Data Processing and Analysis

Simple statistics were obtained from the extracted data, including average number of patients, average study duration, type of data sources, and average age of patients. Frequency of articles by region and income groups was calculated based on the World Bank classification (June 2019 update) [17].

The frequency of mentions of the humanistic and clinical impact is reported, and the details of the clinical burden are narratively summarized. Further in-depth analysis was conducted for QoL and economic data. For this purpose, each type of data underwent processing as elaborated below.

Disease Severity

Reporting of disease severity in different studies was heterogeneous and used different terminologies that hindered the ability to assess severity as an independent variable, so severity ranks from different publications were transformed into an ordinal scale ranging from 1 to 5, where a higher value indicates higher severity. In case a study featured only two severity groups, the less severe group was labeled 2 and the more severe group was labeled 4, while if the study mentioned three subgroups, the subgroups were labeled 2, 3, and 4. In case of four severity groups the labels were 1, 2, 4, and 5, while in the case of five severity subgroups, the groups were labeled from 1 to 5. Studies reporting the whole population without specifying severity levels were excluded from the ordinal scale and labeled as “unstratified population.”

Economic Data

Economic data were converted to annual cost per patient values when possible. For studies reporting the time horizon as lifetime, the estimated life expectancy of patients was used (average age of death of AD patients − average age at onset) [18, 19]. Furthermore, for cost data, values were adjusted to inflation using the consumer price index (CPI) for 2020 from the World Bank database. If CPI values for the year 2020 were not available, the most recently reported values were used instead [20]. If more than one country was included explicitly in the study, the average CPI of all included countries was used. The CPI for Taiwan was not available, so it was obtained from an external source [21]. Next, values were converted to 2020 USD using the official exchange rate from the World Bank database [22].

QoL Data

Studies measured QoL using different questionnaires or scales. We unified QoL results into one unit to allow for aggregation of results and comparison. Utility values have reference points of 0 and 1, where 0 indicates death and 1 indicates perfect health. The European QoL Five Dimension (EQ-5D) index questionnaire is the QoL questionnaire that provides values on a utility scale, so the QoL values identified using other scales were transformed (i.e., mapped) to EQ-5D index values when possible.

Studies using the Dermatology Life Quality Index (DLQI) and Children’s Dermatology Life Quality Index (cDLQI) questionnaire results were transformed to the EQ-5D index using an online transformation tool [23] To transform EQ-5D Visual Analog Scale (VAS) values, there was no available tool, so we used a custom-made function based on linear regression in patients with AD.

To conduct the linear regression, we used all studies identified in our SLR that included both EQ-5D VAS and EQ-5D index values for the same AD patient subgroups. We identified five studies that included these values [24–28]. The data points in these studies were run through a linear regression model using the least-squares method.

The following linear regression equation was used to convert EQ-5D VAS QoL scores to EQ-5D index values on a scale from 0 to 1:

$$y=0.0136x-0.1534$$

y: EQ-5D VAS QoL score, x: EQ-5D index QoL value.

Productivity Lost

Similarly, productivity lost was reported either as the number of days or hours lost during a certain period, or as a percentage lost in some cases. All values were unified to number of days lost annually per patient by using the Organisation for Economic Co-operation and Development average working hours per year value of 1726 and assuming eight working hours per day [29].

Multiple Regression

Several multiple linear regression models were developed using IBM SPSS statistics software version 25 to determine the main drivers for economic costs and QoL of AD. Economic costs in USD were used as the outcome of one model, while QoL in utility score was used as the outcome of the other model. Different numeric and nominal variables were used as the main predictors (e.g., male percentage, age, severity score). Only clinically and statistically significant models are presented in the results.

Compliance with Ethics Guidelines

This study is based on previously conducted research and does not contain any new studies with human participants or animals performed by any of the authors.

Results

The systematic search yielded 3400 records after de-duplicating hits from different databases plus 48 studies identified via other methods. A total of 233 studies were included in the analysis. Further details are available in Supplementary Fig. S2.

General Results

The majority (66.1%) of the included studies reported data from Europe and Central Asia, yet the most frequent country considered in studies was the USA (46 studies), followed by Germany (35 studies). High-income countries represented more than 85% of the included studies, while only one study reported from a low-income country. More than 90% of the studies were observational, while only 9 studies used economic models and 36 were systematic literature reviews.

Clinical Burden

Itching (also known as pruritis in some studies), depression, and anxiety were the most frequently reported impact parameters in the clinical burden domain (51, 49, and 42 mentions, respectively). Figure 1 shows the frequency of the different clinical burden domains of impact. Itching was the most commonly mentioned clinical impact due to AD. Based on the aggregated data points, the itching or pruritis prevalence in patients with AD ranged from 21% up to 100% [30–35].

Fig. 1.

Frequency of mentioning different impacts related to clinical burden in the included studies

Eight studies reported the median severity of itch due to AD based on a 0–10 numerical rating scale. The median values describing the severity of itch ranged from 4 to 9, with an average of 6 (where 10 represents the highest level of itch) [28, 36–42]. A similar range exists with mean values ranging from 3 to 9, with an average of 6, for studies using a VAS (also 0–10) [9, 43–49].

Eleven studies reported diagnosis of depression prevalence values among patients with AD [26, 30, 50–58]. The average of all prevalence values was 18%. Prevalence estimates ranged from 3% to 57%. These results were slightly different from the self-reported depression values, which ranged from 10% to 37%, with an average of 26% [59–62].

The prevalence of anxiety among patients with AD ranged from 1.2% to 64%. These values were reported by 11 studies with an average anxiety of 24.12%. According to Mizara et al. [63], 41% of patients had a Hospital Anxiety and Depression Score of at least 11, which indicates a definitive case of anxiety.

Humanistic Burden

Concerning the humanistic burden, AD was shown to decrease QoL by impacting different aspects of patients’ lives. Among the included studies, the psychological impact was by far the most mentioned impact (78 times) causing loss in QoL, followed by sleep disturbance (55 times). The details of frequency of mentioning each aspect affecting patients’ QoL is shown in Fig. 2.

Fig. 2.

Frequency of mentioning humanistic burden impacts in the included studies

Sleep disturbance was very common among studies discussing AD burden and included nocturnal awakening due to itch and difficulty in sleep induction [40, 64, 65]. According to the included studies, sleep disturbance results in using sleeping pills or feeling sleepy, unproductive, or lacking concentration during the day [52, 66]. Several studies reported sleep disturbance in more than 70% of patients with AD [34, 45, 65, 67], while others showed lower prevalence, as low as 4.18% [51]. One study used subgroups for sleep disturbances and reported that 38.4% of patients had no difficulties, 23.9% had mild difficulties, 28.2% had moderate difficulties, and 9.6% had severe difficulties in sleeping due to AD [68]. One study also showed that controlling AD resulted in better outcomes related to sleep disturbance: only 8.5% of patients with adequately controlled AD experienced sleep disturbances compared with 23.8% in patients with inadequately controlled AD [69].

QoL Score Burden

The average utility value for the AD general population was about 0.779 based on 71 studies. Patients with the lowest severity had the highest HRQoL (utility), represented by an average utility value of 0.873. HRQoL decreased gradually with increasing severity, with an average utility value of 0.548 for the most severe patients (Table 1).

Table 1.

Utility values based on severity ranks

Severity rank	Number of studies reporting values	Average utility	Minimum utility	Maximum utility
Unstratified population	71	0.779	0.432	0.940
1	3	0.873	0.869	0.877
2	25	0.807	0.732	0.912
3	15	0.728	0.633	0.832
4	25	0.676	0.551	0.881
5	3	0.548	0.420	0.668

Among 597 data point estimates for the QoL questionnaires, several questionnaires were used, including VAS (77), the 36-Item Short Form Health Survey (66), Patient-Oriented Eczema Measure (POEM) (46), EQ-5D (40), AD Burden Scale (36), and Skindex (33). Yet, DLQI and cDLQI questionnaires were the most used to assess the QoL for patients with AD (299 data points).

Multivariate Regression Model for Utility

According to the multivariate regression model (Table 2), male patients with AD had significantly lower utility compared with female patients. Age was not a statistically significant explanatory variable for utility. Conforming with previous findings (Table 1), severity was inversely proportional to utility value.

Table 2.

Multivariate regression model for utility of patients with AD

Parameter	Beta coefficient (β)	Standard error	95% Wald confidence interval		Hypothesis test
			Lower	Upper	Wald chi-squared	Degrees of freedom (df)	Significance
(Intercept)	1.348	0.2433	0.871	1.825	30.675	1	0.000
Severity rank = 2	0.108	0.0256	0.058	0.158	17.746	1	0.000
Severity rank = 3	0.086	0.0504	−0.013	0.185	2.925	1	0.087
Severity rank = 4	0a
Age, years	−0.005	0.0031	−0.011	0.001	2.626	1	0.105
% of males	−0.863	0.2772	−1.406	−0.319	9.686	1	0.002
Scale	0.001b	0.0006	0.001	0.003

Dependent variable: quality of life

^aSet to zero because this parameter is redundant

^bMaximum-likelihood estimate

Economic Burden

Of the included studies, 70 provided data about costs and HCRU. Of those, 41 studies included (direct and indirect) cost data and 32 included HCRU data (e.g., number of outpatient visits). Twenty-eight studies included other economic data, of which the majority reported productivity loss.

Healthcare Resource Utilization

Data collected for AD comprise a wide range of severity and diversity in HCRU, including outpatient visits, emergency department visits, and hospitalizations. Studies usually reported separate data for different severity groups.

Dermatologist visits ranged from 2.8 to 16.3 per year for the unstratified population, with an average of 8.6 [28, 59, 70–72]. Primary care/general practitioner visits averaged 16.5 per year [70, 73, 74], with this number varying significantly by severity, where it reached 20.44 healthcare provider visits per year in patients with moderate to severe AD [50]. Two studies reported the visits of patients with AD to medical specialists other than dermatology, which were allergy and internal medicine, with a rate of 0.2–0.4 visits per year, respectively [70, 72].

As severity increased, the frequency of emergency visits increased. However, for all severity ranks, studies reported a low rate of emergency department admissions. Considering unstratified patients with AD, studies reported a minimum of 0.05 visits per year and up to 1.22 visits per patient per year, with an average of 0.80 visits [50, 68, 71, 74, 75].

For patients with rank 2 severity, the average number of annual emergency department visits per patient was 0.5 [50, 68, 71, 73, 74]. The average was 0.92 visits for patients with rank 3 severity [68, 74] and 1.41 for rank 4 severity [50, 68, 71, 73, 74]. The average annual number of hospitalizations (for the unstratified population) ranged from 0.03 to 1.2 admissions [50, 71, 73, 75]. Patients with severity rank 4 had an average annual hospitalization rate of 0.75 per year [50, 68, 71, 73, 75]. On the other hand, those with severity rank 2 had an average annual hospitalization rate of 0.45 per year [68, 71, 73, 75].

Costs

There was significant heterogeneity between individual studies since the studies came from different countries and several income levels. The total cost of AD per patient was mentioned in eight studies, in which the annual average cost was estimated to be 5246 USD (2020), with a minimum of 769 USD and a maximum of 23,638 USD [72, 74, 76–81]. The average total cost calculated from the studies was less than the sum of average total direct and total indirect cost due to the heterogeneity in sources and calculation methods. Nine studies reported total direct costs with an annual average cost of 4411 USD [48, 72, 76, 82–87]. The total indirect cost per patient was reported in three studies with an average cost of 9068 USD per year [72, 76, 88]. Cost details are presented in Table 3.

Table 3.

Average annual cost per patient with AD (unweighted)

Type of economic burden (direct/indirect)	Number of studies reporting the cost	Number of patients in the studies	Minimum reported cost (2020 USD)	Average cost (2020 USD)	Maximum reported cost (2020 USD)
Total direct cost	9	119,750	940	4411	11,536
Total indirect cost	3	218	1289	9068	15,650

Some studies reported economic data stratified by different factors, most commonly by severity (24 studies), followed by treatment groups (12 studies) and age (9 studies). The exact studies and strata are reported in Supplementary Table S2.

Productivity Lost

Several studies mentioned the economic burden incurred by AD due to productivity lost, which was usually quantified by the number of days of absenteeism and/or presenteeism. Among 28 studies reporting numbers or percentages of workdays lost due to AD as presenteeism or absenteeism, 20 reported absenteeism values separately, 13 reported presenteeism separately, and 14 reported both absenteeism and presenteeism values.

Productivity is significantly affected by AD, as seen by a total of 68.8 days lost annually due to absenteeism and presenteeism combined (for the unstratified population). The presenteeism (54 days lost) [42, 59–61, 72, 73, 78, 83, 89–91] effect was dominant, being more than three times the days lost due to absenteeism (14.8 days lost) [42, 59–61, 65, 72, 73, 78, 83, 88–93]. Productivity lost in days differed significantly among severity ranks, with patients with severity rank 5 losing on average 26.5 days due to absenteeism and 92.5 days due to presenteeism, compared with patients with rank 1, who lost an average of 2.5 days due to absenteeism and 13.6 days due to presenteeism [78, 94]. Table 4 presents the average number of days lost due to absenteeism and presenteeism based on the severity rank.

Table 4.

Average number of days lost per year due to absenteeism and presenteeism, by severity rank

Severity rank	Absenteeism only	Presenteeism only	Total
Unstratified population	14.8	54.0	68.8
1	2.5	13.6	16.1
2	14.0	58.5	72.5
3	23.3	78.5	101.8
4	24.0	95.5	119.4
5	26.5	92.5	119.0

Discussion

The highly prevalent chronic inflammatory skin disease AD affects adults and adolescents, with a significant DALY burden [1]. However, until recently, AD was generally considered to be merely a skin disorder [95]. Many efforts have been made to quantify different aspects of the burden of AD. We aimed to aggregate the findings from different studies to provide a holistic view of AD burden from the humanistic, economic, and clinical perspectives for adult and adolescent patients. Furthermore, due to the abundance of studies evaluating each burden element, we were able to stratify the impact based on additional factors, such as severity.

To date, there is no cure for AD [96]. However, based on these results that show a solid correlation between severity and HRQoL, as well as productivity lost, maintaining patients with mild disease severity could offset most of the burden. This study should be considered as a first step in mitigating the burden of AD by providing an overview of the scale and factors of AD burden. The next step to decrease AD burden should be to research further into specific policy actions that could improve the prognosis of patients with AD. This research should be validated from a local perspective to ensure its eligibility within the healthcare system structure and from the cultural perspective.

The burden of AD might be underestimated in low- and middle-income countries because, despite the abundance of literature on the topic, most of the literature came from higher-income countries; low- and middle-income countries were not equivalently represented in the literature. The global burden of disease study found a positive correlation between disease burden and gross domestic product [1]; however, this might be due to insufficient data and underreporting of AD in lower- and middle-income countries.

As expected, itching was the most commonly mentioned symptom in the literature for patients with AD, in some cases being reported to affect 100% of patients. This symptom was followed by depression and anxiety, which highlights the significance of the psychological illness impact on patients with AD, which was further confirmed by the humanistic burden data, where again, psychological illness ranked number one in terms of frequency of mentions in the literature. Sleep disturbance followed psychological illness in the ranking within the humanistic burden, which is not unexpected since it is linked to nocturnal awakening due to itch [64]. Although sleep disturbance might not be an issue if it is a one-night problem, the impact is amplified when the confounding factor is a chronic disease, and the majority of patients with AD do experience sleep disturbance. Sleep disturbance can lead to a cascade of implications, such as the use of sleeping pills, and usually causes a lack of concentration and lethargy [64, 66].

One of the main consequences of sleep disturbance is productivity loss due to lack of concentration, and lethargy, which might explain the significantly higher presenteeism compared with absenteeism. The productivity lost for the unstratified population by severity made up about one-third of the year, while for the most severe cases, the total productivity lost even exceeded half of the year.

Looking at the HRQoL, the variability of utility lost between different severity groups was significantly wide, which was further confirmed when we developed a multiple regression model that included severity, age, and sex as independent variables.

Our results concerning humanistic burden are concordant with a recent study in Europe assessing the AD burden of illness in adults [61]. It also states that anxiety, depression, sleep disorders, and overall and general impairment create a significant burden for patients with AD compared with controls. Another study by Reed et al. also confirms our findings of the significant losses in QoL and school or work absenteeism burden due to AD [97].

Drucker et al. estimated a similar total annual cost per patient in the USA in 2013 [75], ranging from 3302 to 4463 USD, compared with our estimate of 4411 USD. However, our estimate is not confined only to the USA. The similarity of these values is probably due to the underreporting of the burden in low- and middle-income countries, which might have decreased the average cost if their data were available, as these countries usually have lower unit costs owing to their relatively low gross domestic product.

Because of the diversity of the included studies, each had a different methodology and perspective; therefore, for some calculations, the values from two or more studies could not be used for summary statistics. However, we grouped similar methodological articles for each part of the burden and created summary statistics for specific subgroups. For the same reason, all summary statistics were calculated as nonweighted average values as it was not feasible to calculate the statistics based on the number of patients in each study due to the diversity of studies. Since severity was not measured in the same way in all included studies, we used the severity ranking approach. Although this approach may not provide the most accurate severity estimates, we assume that it is sufficient to provide useful insights about the burden. As costs from different studies were converted to USD and adjusted for inflation, the aggregated results should be interpreted with caution, as the purchasing power parity and treatment protocols, as well as the variance between drugs and medical services in different countries, might have significant effects. The regression was performed without considering the weights of patient numbers because of the difficulties in extracting the number of patients for each subgroup of patients as they usually overlapped.

Conclusions

The burden of AD is significant due to its high prevalence as well as the magnitude of its impact. While the disease is incurable, reducing the severity of the disease and modifying the prognosis of patients could significantly reduce the burden.

Supplementary Information

Below is the link to the electronic supplementary material.