Distinguishing Three Classes of Androgenetic Alopecia based on Concerns in Youth: Latent Class Analysis of a Cross-Sectional Survey


 Background: While androgenetic alopecia (AGA) is a common condition, the concerns of youth about this condition, and how these concerns influence quality of life, personal well-being and health behaviour have been not been thoroughly investigated.Methods: Thai youth were asked to complete an AGA-related questionnaire including the dermatology life quality index, to assess their concerns and the clinical relevance. Latent class analysis was used for data analysis.Results: Data from 278 youth were collected. The mean participant age was 22 years (SD=0.79), and 57% of the participants were men; 94.6% of the participants reported any hair loss, mostly in the AGA severity of grade 1-2. A three-class model showed the best fit. Latent class 1 (LC1) participants (58%) had the lowest probabilities of concerns regarding AGA. LC2 participants (22%) had a generally low probabilities of AGA concerns, except when they were flirting, when they were with their lover and regarding how their hair looked when they were photographed. LC3 participants (20%) were the most concerned about AGA in all aspects. Compared with LC1 participants, LC3 participants were more often to use more than 3 hair styling products (relative risk, 6.7;1.5-29.6;P=0.013) and their AGA severity was more severe in both males (relative risk, 3.2;1.1-8.9;P=0.023) and females (relative risk,10.2; 3.3-31.0;P=<0.001).Conclusions: Youths who are more worried about AGA experience higher AGA severity and poorer quality of life, and are more likely to use multiple hair styling products.

3. Are you concerned that if you have any children, they will have a hair problem? 4. Are you concerned about your relative's hair loss problem? 5. Are you concerned about your best friend's hair loss problem? 6. Are you concerned that other people will make fun of your hair? 7. Are you concerned about how your hair looks when you are swimming? 8. Are you concerned about how your hair looks when you are photographed? 9. Are you concerned about how your hair looks when you are irting with someone? 10. Are you concerned about how your hair looks when you are with your boyfriend or girlfriend? 11. Have you ever considered getting a hair transplant?
Participants with missing data in any of these questions were excluded. The optimal number of classes was determined by comparing the convergence of iterative routines for parameter estimation and goodness-of-t statistics (Akaike's information criteria (AIC), the corrected AIC, the Bayesian information criteria (BIC) and the adjusted BIC. LCA produces predicted probabilities of each class for each subject, with the nal class membership of each subject is determined by the highest predicted probability ensuring each subject is assigned to only one class. After determining the number of classes, we compared classes regarding the distribution of sociodemographic factors (gender, number of hair products used, number of people you interact with per day), disease severity and duration of disease using Pearson chi-square for categorical covariates, and Kruskal-Wallis tests for continuous covariates. To identify the associations between the above-listed variables and membership in each class, we use multinomial logistic regression to quantitate relative risks and 95% con dence intervals (95%CI). The dependent variable was the latent class. Summary data for AGA-related quality of life, personal beliefs about how whether food can affect AGA, and their treatment-seeking behaviours of each class were analyzed and presented as proportions. All comparisons were two-sided, and P-value < 0.05 was considered signi cant. All analyses were performed using STATA version 15.1 (StataCorp, College Station, TX).

Results
A total of 305 questionnaires were answered. As the LCA's classi cation method can be applied only to the subjects with complete data, we con ned our analysis to 278 (91%) respondents without any missing data in the questions mentioned above. The mean age of respondents was 22 (SD 0.79) years. The majority (57%) of the participants were men. Of 278 participants, 59% had a monthly income of 200-400 USD, 66% did not use any hairstyling product and 75% interacted with 11-100 people per day. 147 (92.5%) men and 116 (97.5%) women reported any hair loss; the majority of both men and women who reported hair loss reported AGA severity of grade 1-2, with 61% reporting a duration of AGA of > 6 months but less than 5 years. Almost 80% of the subjects had a same gender 1st degree relative with a hair loss problem.

Latent class analysis and characteristics of each class
In total, 3 latent classes were tted to the data. (Table 1) reports the conditional probabilities of the 11 previously selected indicators among the three LCs, interpreted as the probability of each indicator being present in the members of each class. Of 278 subjects, 162 (58%) belonged to LC1, 62 (22%) belonged to LC2 and 55 (20%) belonged to LC3. LC1 subjects had the lowest probabilities of concerns regarding AGA. In LC2 subjects, probabilities of AGA concerns were low in general, except when they were irting, when they were with their lovers and regarding how their hair looked in photos. LC3 subjects were most concerned about AGA. Consequently, we chose the following terms to designate the three classes: LC1, "Concern-free"; LC2, "Lover"; LC3 "Worrisome".  Table 2) compares the subjects' socio-demographic data and disease characteristics across the three LCs. The Worrisome class (LC3) had distinct characteristics of higher income, they used more hair styling products and interacted with a greater number of people per day. The severities of AGA in both men and women were also signi cantly higher in this group. The Concern-free and the Lover classes had quite similar characteristics in terms of income per month, use of hair styling products, number of people they interact with per day, and the severities of AGA in males and females. The proportion of subjects with same gender 1st degree relative with hair loss problem were relatively high in all classes. The duration of AGA was longest in the Worrisome class (LC3) followed by the Lover class (LC2), then the Concern-free class (LC1).  How AGA affects the youths' quality of life (Table 3) demonstrates the burden of AGA-related quality of life was highest in the Worrisome class (LC3) with 20% having more than 5 points in DLQI, compared to only 2.5% and 3.23% in the Concern-free class (LC1) and in the Lover class (LC2), respectively. More than 67% of the subjects in the Worrisome class reported at least one aspect of their daily life that was affected by their AGA condition, versus less than 25% in the other two classes. Most affected aspects of quality of life in all three classes were similar, their social characteristics and their self-con dence (data not shown).
Personal perspective about food and its effect on AGA in each class Personal beliefs about food and lifestyle among each class are shown in (Table 3). The Worrisome class (LC3) were most likely to believe that all the listed foods or ingredients, namely monosodium glutamate, high fat diets, chocolate and vitamin supplements, had some effects on AGA, versus the other classes. The Concern-free class (LC1) were least likely to believe that any of the listed foods could help or worsen their AGA condition. Of all the listed foods, the food additive monosodium glutamate was considered by all three LCs consider as something that could possibly worsen their AGA condition.

Treatment-seeking behaviours in each class
For treatment-beliefs and behaviours, a higher percentage of subjects in the Worrisome class (LC3) thought it was important to treat AGA urgently, compared to the other two classes. They were willing to spend more money on AGA treatment with more than 5% of the subjects willing to spend more than their entire monthly income on treatment. Half of the LC3 group were also willing to spend more than 10% of their one-month income on AGA treatment, versus only approximately 20% of the other two classes.
Interestingly more than 30% of the Worrisome class (LC3) and the Lover class (LC2) were willing to live a healthier life if it would improve their AGA condition, while only 14.5% of the Concern-free class (LC1) considered to do so.

Discussion
This is the rst study to use latent class analysis to describe the effect of AGA-concerns in youths. Using the data from a cross-sectional survey, we were able to categorize youths into three distinct classes which are explicable and relevant to clinical practice. The rst class, the "Concern-free" class (LC1), accounted for 58% of all subjects and was characterized by a low-level of concern about AGA. Even though the prevalence of Grade 1-2 AGA was high amongst our entire student cohort, this class showed low levels of interest in all AGArelated matters. The second, "Lover" class (LC2, 22%) was characterized by a high level of AGA concern only when they are irting or with their lovers and when considering how their hair appeared in photos. They had these concerns despite the fact that their AGA severities were not different from the Concern-free class (LC1), and more than one-third of the subjects in this group were willing to live a healthier life if it would help with their conditions. In contrast, the third "Worrisome" class (LC3), had the highest commitment and determination to improve the condition and also a higher proportion of both males and females with grade 3-5 AGA severity. This Worrisome class (LC3, 20%) represented the class where participants were most concerned about AGA in every aspect. The increased AGA severity and number of hair products they used were statistically signi cantly higher than the Concern-free class (LC1). Their AGArelated quality of life was poorest compared to the other two classes, and they harbored beliefs that food and lifestyle impacted on AGA and were willing to spend more money for the treatments.
AGA has been stigmatized as a dermatological condition that signi cantly affects self-image satisfaction, with a negative impact on patients' well-being (5,8,15). Although few studies have examined how concerning thoughts about AGA impact on quality of life, higher disease severity and higher impacts on quality of life have been observed in groups that are more concerned about AGA. For example, a previous prospective, multicenter study, reported that the quality of life of AGA patients decreased as severity of AGA increased (5). This emphasizes the importance of addressing AGA concerns in clinical practice. Our study showed that youths who are more concerned about AGA also think it is important to treat their AGA urgently. Indeed, treatment which is initiated early leads to a better response rates and clinical outcomes (16).
Our study also found that the group who was more concerned about AGA used more of hair products. However, we de nitely cannot conclude whether their condition led these participants to seek products they believed might help improve their looks, or whether those products might be a factor contributing to their conditions. Nevertheless, this is an important discussion point, since one main obstacle to optimal AGA treatment outcomes is believed to be poor compliance and lack of adherence to the treatment regimens (17). Our result could imply that the patients, particularly youths, with more AGA-related concerns are more likely to adhere to therapies because of their usual complex hair care regimens. Incorporating topical drugs such as minoxidil, an FDA-approved treatment, could provide an e cient and user-friendly treatment for AGA in this group (18).
Other than negative social impact, AGA has been suggested as a potential risk factor for asymptomatic atherosclerosis, myocardial infarction, and metabolic syndromes (19,20). Microvascular insu ciency theory has been proposed as one of the mechanisms (21,22). An excessive of circulating insulin causing vasoconstriction results in increased peripheral resistance, which decreases the blood supply to the scalp and enhances the effect of dihydrotestosterone on follicular miniaturization is one proposed mechanism (23). The link between AGA and metabolic disease is even stronger in the group of patients with early onset AGA (24)(25)(26). Even though an association between healthy lifestyle and AGA is still not clear, evidence about the associations between healthy living and its bene t in cardiovascular disease and metabolic disease prevention are well recognized (27)(28)(29). We therefore propose that it is important to encourage any youths with concerns about AGA, to live a healthier life not because it can help their AGA condition, but because it can help minimizing the increased cardiovascular disease and metabolic disease in this group of patients.
Our study is the very rst to focus on androgenetic alopecia concerns in youths. Most about AGA concerns and impact have been conducted in middle-aged subjects (5,(30)(31)(32)(33). A previous study found AGA affected over 70% of adult males and 50% of adult women (34). However, in this study, we found that 92.5% of men and 97.5% of women in their twenties reported hair loss considered as AGA. We cannot de nitively conclude presence and severity of AGA in our study participants, since the data was collected by self-report. Although this might cause controversy regarding diagnostic accuracy, it nevertheless re ects the perception of the students towards their conditions, and the impacts on their lives. This is important from an ethical point of view, since it is essential to consider whether people who have concerns about AGA should be treated as healthy, if their conditions are relatively mild, given their concerns have proven effects on their well-being and health-seeking behaviours.
The concept of sickness should refer to the social and societal dimensions of an ailment; It is about the perceptions and subjective feelings of those with the condition, not only the severity of disease. This research further emphasizes the signi cance of patient perspectives on AGA. Furthermore, our ndings are based on a classi cation method involving no pre-established de nitions and suggest three distinct group of concern regarding AGA in youths. These three subtypes, "Concern-free", "Lover" and "Worrisome", are consistent with those observed in clinical experience and provide a useful classi cation for clinicians to integrate into the clinical practice for better understanding and treatment outcomes in youths with AGA. For a pragmatic use of classi cations in real-life practice, we suggest that the 11-point questionnaires could be shortened into 4 most distinct topics (from parameter estimation, Table 1.): concern about hair loss in the future or concern about a relative's hair loss to include them in the "Concern-free" class (LC1) concern about hair when irting or when you are with your lovers to include them in the "Lover" class (LC2). If all the answers to these topics are no, include them in the "Concern-free" class (LC1). Further prospective studies should be conducted to evaluate the bene t of incorporating these classi cations which could be determined by administration of a short questionnaire, into real-world treatments for AGA, to assess treatment adherence and clinical outcomes after treatments, in participants from each class.

Conclusions
More than 90% of youths reported the hair loss considered as AGA. Youths who are more worried about AGA have experienced higher severity of AGA, worse AGA-related quality of life, and are more likely to use multiple hair products. LCA which was used to classify 3 groups of AGA in the youths based on their concerns might be useful for integrating into the clinical practice to assess adherence and outcomes after treatment. Further research should be focused on prospective evaluating the advantage of these classi cations.

Declarations
Ethics approval and consent to participate: The study was approved by the institutional Review Board, Faculty of Medicine, Chulalongkorn University (IRB 441/59). Informed consent to participate in the study was obtained from all participants.

Consent for publication:
Not applicable.
Availability of data and materials: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.