Introduction

Acne Epidemiology

The Global Burden of Disease Study 2010 found that acne vulgaris (henceforth acne) is the eight most common skin disease, with an estimated global prevalence (for all ages) of 9.38%1. In different countries and among different age groups, the prevalence of acne varies, with estimates ranging from 35% to close to 100% of adolescents having acne at some point2.

Symptoms and Impacts of Acne

Acne patients typically present with comedones, papules and pustules3. Comedones can be subdivided into two types – open comedones (blackheads), which are clogged follicles with openings exposing its contents to the air, and closed comedones (white heads), which are clogged follicles without an opening4. Papules are raised lesions on the skin that are smaller than 1 cm in diameter while pustules are similar to papules but inflamed and filled with pus4. In patients with severe acne, nodules and cysts – inflamed, swollen lesions that are at least 5 mm large – may be present3,4. In addition, other symptoms such as the scars, erythema and hyperpigmentation may be observed in acne patients4.

On top of discomfort due to the clinical symptoms of acne, patients may experience other negative impacts. A study observed significantly higher unemployment rates among acne cases relative to controls, suggesting a correlation between acne and employment5. Further, acne has been found to adversely affect the social life6, self-esteem and body image of individuals and is often co-morbid with psychological disorders including depression and anxiety3. Additionally, acne is associated with substantial financial costs, with one study estimating that the cost of treating acne in Germany adds up to 400 million Euros annually7.

Aim

This review aims to analyze the epidemiology of acne around the world and investigate the factors that significantly modify the risk of presenting the condition.

Results and Discussion

Epidemiology of acne

The 35 articles reviewed differed in study design, acne definition and severity grading systems, variables studied and population characteristics. Population characteristics such as age and sex differed between studies depending on the type of acne and variables the researchers were interested in. For instance, Wei et al.8 studied adolescent acne, Kaminsky, Florez-White, Bagatin and Arias9 studied adult acne, and Park, Kwon, Min, Yoon and Suh10 studied childhood acne. Similarly, some studies only studied the risk factors for acne in females11 while others only studied acne in males12. Sample sizes used also varied, ranging from 88 in Ismail, Manaf & Azizan13 to 27,083 in Klaz, Kochba, Shohat, Zarka & Brenner14. Further, some studies investigated the co-morbidities of acne, however, since this paper is not intended to provide a review of acne co-morbidities, they will not be discussed here.

The acne and severity grading systems used by the different studies is described in Table 1. Depending on the acne definition and severity grading system used, the resulting prevalence estimates differ. Despite the presence of objective symptoms of acne (such as the presentation of comedones, papules and/or pustules), dermatologists disagree about the minimal criteria that should be used to diagnose the condition15. Similarly, efforts to create a standardized grading system for acne severity have been unsuccessful and over 25 different systems are currently in use16. As such, different studies use different definitions and grading criteria, making it difficult to compare their results and derived prevalence estimates3,16. The prevalence estimates obtained are also influenced by other factors such as the sample size and country studied. A larger sample will result in a more representative prevalence estimate. Prevalence estimates ranged from 26.8% in a study conducted in Germany17 to 96% in a study conducted in Brazil18. A summary and description of the articles reviewed can be found in Table 1.

Table 1 Summarised descriptions of journal articles on acne published between 1999 and 2019.
Full size table

Demographic factors that influence acne presentation

Many papers have demonstrated that acne presentation is influenced by demographic factors. The onset of acne typically correlates with the onset of puberty, when sebum production increases19. As such, the prevalence of acne increases with increasing age, showing highest incidence in teenagers and a relatively low incidence in pre-pubertal children19. After reaching the late teenage years or young adulthood, acne prevalence rates follow a decreasing trend with increasing age19,20. The results of the reviewed articles generally followed this trend, with higher odds of acne in teenagers compared to young adults and children (Tables 2, 3).

Table 2 List of risk factors for acne presentation analysed in the articles and results obtained for each risk factor.
Full size table
Table 3 Strength of association of risk factors with acne presentation.
Full size table

Previous reviews have reported that the prevalence of acne is higher in females than males20,21. Similarly, the Global Burden of Disease Study conducted in 2010 estimated that the prevalence of acne was 8.96% in males, lower than the estimated prevalence of 9.81% in females1. Lynn et al.21 also noted higher acne prevalence in females at younger ages, possibly due to the earlier onset of puberty in females relative to males. However, the papers reviewed in this study showed mixed results, with only two papers revealing a higher odds of acne in females while another three demonstrated a higher odds of acne in males (Table 3). These results may be due to differences in the characteristics of the sampled population or country studied. Notably, in a study that found lower odds of acne in females than males, the odds ratio was adjusted for stage of puberty, removing the potential confounders of age and onset of puberty22. When the pooled odds ratio was calculated, an OR of 1.07 (95% CI 0.42–2.71; males with reference to females) was obtained (Fig. 1) suggesting that male sex is only associated with a slight increase in acne risk.

Figure 1
figure 1

Individual and pooled odds ratio and 95% confidence intervals for acne presentation in association with male or female sex. Two studies were excluded from meta-analysis due to a lack of data, such as the odds ratio and/or 95% confidence interval.

Full size image

Relatively few studies considered other demographic factors. Factors including years of education, socioeconomic status and parent’s educational level did not significantly affect acne prevalence. Studies also found that low computer usage and marriage were protective for acne while working in an office was a risk factor for acne.

Demographic factors that influence acne severity

Demographic factors also influence the severity of acne. The articles reviewed revealed that the odds of severe acne are higher in older teenagers compared to younger teenagers or preteens (Table 4). As sebum production increases during puberty, older teenagers tend to have higher sebum production compared to younger teenagers. High sebum levels favor the growth of Propionibacterium acnes, a species of bacterium implicated in inflammatory processes in acne and the development of inflammatory acne lesions which are typically associated with more severe acne18,19.

Table 4 Strength of association of risk factors with acne severity (moderate/severe acne).
Full size table

Previous reviews have found that severe acne is more common in males compared to females21. Most of the articles reviewed in this study are in line with this trend, demonstrating an association between severe acne and being male (Tables 4, 5). One study reported higher severe acne prevalence in females relative to males, for the age group 11 to 13 but not the 14 to 15 or 16 to 17 age groups23, which may be, at least in part, due to the earlier onset of acne in females2.

Table 5 List of risk factors for acne severity analysed in the articles and results obtained for each risk factor.
Full size table

Other demographic factors were rarely investigated. In contrast to the results found for acne prevalence, parent’s education level and family education level were found to increase the risk of more severe acne. Further, a study reported a higher risk of severe acne in Whites at age 14 to 15 and those who were the only child at age 16 to 17 while household income was not significantly correlated with acne severity23.

Genetic and hormonal factors that influence acne presentation

Studies have also demonstrated the impact of genetic factors on acne presentation. Dreno and Poli24 reported that a positive family history of acne in parents was associated with increased acne risk in their offspring. While the articles reviewed in this study used different definitions of family history, with some considering only parents and others including siblings, first- or second-degree relatives, regardless of the definition used, the large majority of the articles consistently found that a positive family history was significantly associated with increased odds of acne in individuals. The pooled odds ratio of 2.91 (95% CI 2.58–3.28; family history in parents with reference to no family history in parents) suggests that family history in parents is associated with an increased risk of acne presentation (Fig. 2). Two case-control studies were excluded from the meta-analysis due to study design, but their results were still in line with the observed trend in the meta-analysis11,25. In addition, a study seems to suggest a possible additive effect of maternal and paternal family history on the prevalence of acne, as a larger odds ratio of 2.6 is observed if both parents have acne, compared to an odds ratio of 2.1 and 1.7 for maternal and paternal acne, respectively26. However, as few studies present data for maternal, paternal and both parents that allow for such comparison, further studies are required to determine if a true additive effect is present.

Figure 2
figure 2

Individual and pooled odds ratio and 95% confidence intervals for acne presentation in association with family history (parents) or no family history (parents). Four studies were excluded from meta-analysis due to a lack of data, such as the odds ratio and/or 95% confidence interval. Five additional studies were excluded from meta-analysis due to study design or because the comparison and reference groups used were different from the other studies. Two entries are used for Wolkenstein et al.49 as the odds ratios for maternal family history and paternal family history were presented separately.

Full size image

Further, an individual’s skin type (for example oily, neutral or dry skin) can be classified according to their skin sebum level. Twin studies suggest that skin sebum levels are controlled by genetic factors27. Oily skin shows a strong association while mixed skin shows an association with acne presentation relative to neutral or dry skin (Tables 2, 3). Since Propionibacterium acnes favor environments with high sebum levels19, having oily and mixed skin characterized by higher sebum levels increases the risk of acne presentation.

Other genetic factors were considered, with one study finding that acne was more prevalent in taller individuals and in those with a lighter skin tone. A few studies also considered factors related to hormones. Contraceptive use and the regularity of menstruation were not significantly linked to acne prevalence while pregnancy status and onset of puberty were associated with acne prevalence, with those who have never been pregnant and post-pubertal individuals at a higher risk of acne.

Genetic and hormonal factors that influence acne severity

Bhate and Williams19 observed that heritability estimates and twin studies suggest a genetic basis for acne and reported that individuals with a family history of acne tend to have more severe acne. In contrast to this observation, the articles reviewed suggest that a positive family history may not necessarily correlate with increased acne severity – approximately half of the articles reviewed found significantly higher odds of severe acne in those with a positive family history while the other half found that family history was not significantly associated with acne severity. A meta-analysis using loose criteria was conducted to study the association of acne severity with positive family history of acne with reference to no family history of acne (Supplementary Fig. S1), and the results suggest that family history may increase the risk of more severe acne. However, this result should be interpreted with caution due to the use of loose meta-analysis criteria. As mentioned earlier, some evidence seems to suggest a possible additive effect of family history on acne presentation. However, this presence or absence of this additive effect may depend on acne severity. A study observed an additive effect for those with mild acne but not for those with severe acne26. Further research is needed to determine if there is an interaction between additive effects of family history and acne severity.

Skin type can also influence acne severity. Oily (whether dermatologist- or self-evaluated) and seborrheic skin was observed to be associated severe acne (Tables 4, 5). Similarly, those with more severe acne were more likely to have higher sebum production and high usage of drugs to treat their acne. Most studies also noted that in general, menstrual characteristics (such as onset of puberty and regularity of menses) were not significantly related to acne severity. One study also found that skin color was not associated with acne severity. In addition, oral contraceptives may be protective for more severe acne.

Dietary factors that influence acne presentation

The importance of dietary factors in influencing acne presentation has been widely debated28. In particular, dairy and chocolate intake have received the most attention, possibly due to the hypothesis that Western diets are related to acne. The papers reviewed that studied the influence of the intake of various dairy products – including dairy in general, ice cream, yoghurt, cheese and different types of milk – on the presentation of acne elicited differing results (Tables 2, 3). Most studies found that cheese, yoghurt and ice cream intake did not significantly impact the risk of acne. The influence of milk intake on acne risk, however, was unclear, with inconsistent results between studies. For example, Ismail, Manaf and Azizan13 found that drinking milk on at least once per week was linked to increased odds of acne to 3.99 relative to those who drank milk less than once per week while Adebamowo et al.29 found that the intake of whole milk and low fat milk did not significantly influence the odds of acne. However, a meta-analysis review found that the intake of any amount of dairy in general or any kind of milk, regardless of the fat content (full-fat, whole, low-fat, skim) was linked to increased odds of acne30, suggesting that dairy and milk intake are likely to influence acne presentation. Possible explanations for this effect implicate steroid hormones or sugars present in milk21. Whey, a protein found in milk, may also be linked to acne presentation. One study reported that 5 healthy males developed acne after taking whey protein concentrates31. Similarly, Silverberg32 observed 5 patients who displayed acne soon after taking whey protein products. 2 patients who stopped taking whey protein showed good response to acne treatments that they did not respond to when they consumed whey protein. Simonart31 suggested a link between whey consumption and increased insulin levels, which in turn activates signaling pathways that eventually contribute to acne development. Notably, whey protein is found not only in milk, but also in some protein supplements used to support muscle building32.

The influence of chocolate intake on acne presentation is also the subject of debate19. Several studies, regardless of design, found that high or frequent chocolate intake was associated with increased odds of acne presentation (Tables 2, 3). A few hypotheses have been suggested to explain the possible effect of dairy and chocolate consumption on acne presentation. One hypothesis suggests that the sugars in dairy products and chocolate trigger insulin secretion, activating signaling pathways that eventually lead to increased keratinocyte proliferation, which can lead to the formation of acne lesions21,28. This may explain why some studies considered the combined effect of sweets and chocolates on acne presentation10,33. An alternative explanation suggests that chocolate consumption increases the secretion of inflammatory cytokines by cells and may influence the formation of acne lesions28. Meta-analyses using loose criteria were conducted to investigate the association of acne with chocolate intake (Supplementary Fig. S4), and the results suggest that high chocolate intake may increase the risk of acne. However, this result should be interpreted with caution due to the use of loose meta-analysis criteria.

High fat intake may also be a potential dietary risk factor for acne. Despite the inconsistency in the results from different studies, several studies found that high fat intake increases the odds of acne presentation. While potential mechanisms to explain this effect have been suggested, the evidence in support of them is insufficient34. More work is needed to establish the relationship between fat intake and acne and understand the possible mechanisms involved. Several dietary factors may be protective for acne. High intake of fish is suggested to reduce the risk of acne35, (Tables 2, 3). However, Wang et al.36 observed that seafood intake was associated with increased risk acne presentation, suggesting that the intake of other types of seafood may be associated with an increased rather than decreased risk of acne. Furthermore, some studies report that a high intake of fruits and vegetables may lower the risk of acne presentation. These observed protective effects may be due to the omega-3 fatty acids found in fish and the high fibre content in fruits and vegetables, which have been shown to lower the levels of Insulin-like growth factor 1 (IGF-1), thus reducing acne risk37.

In general, most studies reported that acne prevalence shows no relationship with the intake of carbonated drinks, fast food, meat, vitamins and minerals, nuts, and rice, pasta and semolina. Similarly, irregular mealtimes and a lack of nutritional information were not linked to acne prevalence. In addition, relatively few studies found that high glycemic load, spicy food and white bread was linked to increased risk of acne while a Mediterranean diet was linked to decreased risk of acne. The effect of diet in general and the intake of sugary foods, such as cakes and sweets, on acne is unclear, with several studies suggesting that they are risk factors while other studies found them insignificant. However, since these factors were only investigated in few studies, these results require further verification by future research.

Dietary factors that influence acne severity

The influence of chocolate intake on acne severity is also the subject of debate19. A few studies found that chocolate intake was associated with increased acne severity (Tables 4, 5). However, milk may confound the relationship between chocolate intake and acne severity, since many types of chocolate contain milk38. In a clinical study, 99% dark chocolate was used to control for this potential confounder. The study found that the daily intake of a small amount of 99% dark chocolate for four weeks resulted in a statistically significant increase in acne severity grades of participants38. However, it is unclear why chocolate intake is associated with more severe acne. One explanation implicates cocoa butter, a component of chocolate that contains high levels of oleic acid. Experiments conducted in animals demonstrated that oleic acid can affect the keratinization of skin and promote the development of comedones39. Oleic acid may have a similar effect in humans, promoting the formation of comedones and contributing to more severe acne.

Three studies also suggested that milk intake may increase the risk of more severe acne40,41,42, which is consistent with a meta-analysis that reported that high milk consumption was significantly associated with the presentation of moderate-severe acne43. Explanations for the association between milk intake and acne presentation may also explain the association observed for milk intake and acne severity. In addition, to investigate the association of acne severity with milk intake, a meta-analysis using loose criteria was performed (Supplementary Fig. S3), and the results suggest that high milk intake may increase the risk of more severe acne. However, this result should be interpreted with caution due to the use of loose meta-analysis criteria.

Some studies report that high dietary intake of fruits and vegetables may lower the risk of severe acne. However, processed fruit juices consumption does not confer this protective effect, but instead, is associated with increased acne severity44. A high dietary intake of fish may be associated with not only reduced acne risk, but also reduced risk of severe acne41 (Table 5).

In addition, a few studies found that acne severity was not significantly associated with diet in general, fasting, a lack of nutritional information, carbonated drink intake, bread/pasta intake, cheese/yoghurt intake, red meat intake, cured meat intake and spicy food intake. Intake of nuts, intake of eggs, intake of potato chips, high intake of fat and intake of oily food was linked to more severe acne while intake of whole grain bread, white rice and chicken and dieting to lose weight were linked to less severe acne. However, since these factors were only investigated in a small number of studies, these results need to be further verified by future research.

Personal factors that influence acne presentation

Personal factors including Body Mass Index (BMI), smoking status and alcohol intake have previously been linked to acne presentation. Most studies have noted increased prevalence of acne in overweight and obese individuals (typically defined as BMI ≥23 kg/m2 and BMI ≥25 kg/m2, respectively) relative to underweight individuals (BMI <18.5 kg/m2) or individuals of a normal weight (18.5 kg/m2 ≤ BMI < 23 kg/m2). A pooled odds ratio was calculated to establish the association between BMI and acne risk (Fig. 3). The obtained OR of 2.36 (95% CI 1.97–2.83; overweight/obese BMI with reference to normal/underweight BMI) suggests that BMI significantly influences acne presentation. While a case-control study by Lu et al.45 was excluded from the meta-analysis because of study design, their results were in line with the trend found in the cross-sectional studies. Obese and overweight individuals tend to have higher glycemic loads and androgen levels, which may increase sebum secretion, promoting the formation of acne lesions26. Dietary factors may confound the relationship between BMI and acne.

Figure 3
figure 3

Individual and pooled odds ratio and 95% confidence intervals for acne presentation in association with overweight/obese individuals or normal/underweight individuals. One study was removed from the meta-analysis due to study design. Two entries are used for Aksu et al.48 as the odds ratio for normal BMI and overweight BMI were presented separately.

Full size image

The effect of smoking status on acne prevalence is controversial, with inconsistent evidence on whether smoking is a protective or risk factor19. Rombouts, Nijsten and Lambert22 observed that in girls, smoking was linked to decreased odds of acne of 0.41 (95% CI 0.13–0.82) while Al Hussein et al.35 reported that smoking was a risk factor for acne, increasing the odds of acne to 2.859 (95% CI 1.467–5.576). Studies also considered smoking duration and number of cigarettes smoked. Smoking has been proposed to influence acne through multiple pathways that may affect processes such as inflammation, wound healing and immune responses22. For example, nicotine in cigarette smoke is suggested to activate receptors on cells such as keratinocytes, slowing the process wound healing and promoting acne in patients, however, there is a lack of evidence supporting this claim21. In contrast, smoking is also suggested to reduce inflammation, protecting against inflammatory acne22. The complex way in which smoking influences the pathogenesis of acne may contribute, in part, to the inconsistency in the findings from different studies. Study design may also affect the results obtained in different studies. A meta-analysis found a no significant association between smoking and acne prevalence when all studies were included, however, when studies with a quality score ≤6 were excluded, a significant protective effect of smoking on acne prevalence was observed46.

A small number of studies have also investigated the relation between substance use and acne prevalence. Alcohol intake was consistently found to have no significant relation to acne prevalence. Tobacco use may be protective for acne while cannabis use may be linked to increased acne risk. In addition, a few studies found that lack of sleep or insomnia, sun exposure, high mental stress, study pressure and cosmetic usage may be associated with increased acne presentation. In contrast, frequent face washing may be linked to lowered acne risk. One study also found no significant association between acne and both physical exercise sunbed usage. However, some factors show inconsistent results. For example, Hogewoning et al.47 found that living in an urban environment was linked to increased acne risk, while the findings from Aksu et al.48 suggest the opposite. For country of residence, living in certain countries, like Poland and Belgium, was linked to reduced acne risk while living in Czech or Slovak Republic was linked to increased acne risk. Further research is necessary to establish the association link and understand possible mechanisms underlying these associations.

Personal factors that influence acne severity

Previous studies have suggested a link between personal factors and acne severity. Most studies suggest a strong association between overweight or obese BMI and acne severity, with Al Hussein et al.35 reporting an odds ratio of 5.02 (95% CI 1.284–19.682; BMI ≥25 relative to BMI <25) and Lu et al.45 reporting a similar odds ratio 5.027 (95% CI 2.758–9.162; overweight/obese BMI relative to underweight/normal BMI). A meta-analysis using loose criteria was conducted to determine the association of acne severity with overweight/obese BMI with reference to normal or underweight BMI (Supplementary Fig. S2), and the results suggest that overweight/obese BMI may increase the risk of more severe acne. However, this result should be interpreted with caution due to the use of loose meta-analysis criteria. Similar mechanisms may explain the association between BMI and both acne presentation and acne severity.

The effect of smoking status on acne severity is controversial. Rombouts, Nijsten and Lambert22 observed that in girls, smoking was protective of severe acne while Karadağ et al.44 found that severe acne occurred more frequently among smokers. The effects of smoking duration and number of cigarettes smoked on acne severity were also investigated in some studies. While one study observed a significant dose-dependent relationship between acne severity and the number of cigarettes smoked per day17, other studies did not find a significant dose-dependent effect22. Explanations for the association between smoking and acne presentation may also explain the association between smoking and acne severity.

A small number of studies have considered the influence of other factors on acne severity. Makeup use, high mental stress and exposure to chemical substances were shown to be associated with more severe forms of acne while cannabis use, tobacco use, lack of sleep, sun exposure, seasons of the year, climate, travel to humid regions, birthplace, duration of residence in the US and frequent face washing were not significantly associated with acne severity. Living in a semi-rural environment or certain regions in Turkey and physical exercise may be protective for severe acne. Further research is necessary to establish the association link and understand possible mechanisms underlying these associations.

Limitations and conclusions

When selecting articles for review, no exclusion criteria based on geographical location was used. However, because the selected articles were unable to fully represent all regions and continents in the world (for example, no studies from the Australian continent were included based on the selection criteria), the result may not apply to the areas that were not included in the study. In addition, most of the articles reviewed are cross-sectional or case-control studies where the variables have not been experimentally manipulated, thus we cannot exclude the possibility of confounders that may cloud the true relationship between a factor and acne presentation or severity. Furthermore, despite the use of objective criteria to evaluate potential factors influencing acne presentation and severity, the analysis may be subject to researcher bias. Inadvertent errors may also have occurred when summarizing the data from primary literature.

We conducted meta-analysis on risk factors with replicated results reported in a minimum of three independent acne publications. However, we faced some difficulties as some studies reported incomplete information and failed to report the odds ratio and/or 95% odds ratio confidence interval. In addition, we noted that for each particular factor, the comparison and reference groups used in different studies were not compatible, making it difficult to do a meta-analysis, as the results were not comparable. Further, the different study designs used made it difficult to compare the results of different studies via meta-analysis. Thus, meta-analysis was not conducted on all factors reviewed.

Based on the analysis, the potential factors that are the most likely to influence acne presentation and severity are family history and BMI. However, more work needs to be done on dietary factors, smoking, mental stress and sleep duration to understand their effects on acne presentation and severity.

This study summarizes the potential factors that may affect the risk of acne presentation or severe acne and can help researchers and clinicians to understand the epidemiology of acne and severe acne. In addition, the findings can guide future research on risk factors with the hope of better understanding the pathophysiology of acne and developing effective therapeutics.

Methods

Search strategy and selection criteria

A search was conducted on the Web of Science database in September 2019 and only papers of the document type ‘article’ published between 1990 and 2019 containing the search term ‘acne’ in the title and the terms ‘epidemiology’ and ‘risk factor’ in the topic were included. ‘Acne’ was used instead of ‘acne vulgaris’ as it is more general and commonly used. 274 articles were identified after this initial search. Since this article is only interested in modifiable risk factors of acne vulgaris, articles that discussed other forms of acne (such as acne rosacea), polymorphisms associated with acne, acne as a risk factor for other psychological or medical conditions, the efficacy of acne treatments and any other irrelevant articles were excluded. One article with an underspecified study design, unclear acne definition and small sample size was also excluded. Unlike other studies that determined acne status via self-report or clinical evaluation, no indication was made as to how acne status was determined in this particular study, leading to exclusion. Based on these criteria, 247 articles were excluded from analysis. A total of 35 articles – 27 articles from the described search and 8 additional articles for cross-referencing – were analyzed carefully for study design, acne prevalence, acne definition, acne severity grading system used and risk factors evaluated. The process followed to select studies for review is shown in Fig. 4.

Figure 4
figure 4

Flowchart of the process used to select studies for meta-analysis.

Full size image

Meta-analysis

Stata/MP 14.0 was used to conduct meta-analysis using the random effects model to investigate the effect of risk factors with replicated results reported in a minimum of three independent acne publications. Studies with incomplete information such as those that failed to report the 95% odds ratio confidence interval or those that did not clearly define the reference and comparison groups used to calculate the odds ratio were excluded from the meta-analysis. Further, only the results from studies that used a similar design (in this case, cross-sectional design) were included in the meta-analysis.

For meta-analysis using loose criteria, results from studies with different study designs (cross-sectional, case-control and longitudinal) and results that used similar (but not identical) comparison and reference groups were included.