Seasonality of hair loss: a time series analysis of Google Trends data 2004–2016
Dear Editor, Hair loss is a frequently encountered dermatological complaint that often generates psychological distress. Prior studies have demonstrated a seasonal pattern to hair loss. Maximal proportions of telogen hairs have been found to occur in the summer,1, 2 whereas the lowest rates of telogen hairs occur in the winter.1 Additionally, a recent study demonstrated that maximal hair shedding occurs in August and September, and that the percentage of hairs in the anagen phase peaks at the beginning of spring.3 However, these prior analyses are constrained by small sample sizes or homogeneous patient populations in limited geographical locations, and focus on changes in the hair growth cycle stages. Additional studies are needed to describe further the relationship between hair loss and seasonality.
In this study, we explore the relationship between seasonality and hair loss at a population level using Google Trends data. As temperature and daylight levels have been hypothesized to be causative factors of observed seasonal hair loss patterns,3 we also sought to investigate whether temperature plays a role in affecting seasonal variation. We hypothesized that ‘hair loss’ search volume index (SVI), a proxy measurement for actual hair loss experienced in the population, would be highest in summer and lowest in spring.
Google Trends is an online, open-access database that aggregates Google search data from 2004 onwards. SVI is a normalized quantification of a search topic relative to all other Google searches in a given time frame and is indexed from 0 to 100.4 Monthly SVI data were evaluated worldwide and in eight English-speaking countries. These countries were chosen by selecting the top four countries by population in the top 15 countries by ‘hair loss’ SVI from January 2004 to October 2016 in each hemisphere. We chose to use the term ‘hair loss’ in this analysis, as the mean SVI for this term was more than 25 times higher than that of ‘hair shedding’. We assigned each month to a season based on meteorological definitions and corrected for hemisphere. We obtained monthly temperature data from the National Oceanic and Atmospheric Administration and 2013 nominal gross domestic product (GDP) per capita data (http://econfactbook.org) in order to account partially for the effect of countries’ access to technology. Multivariable Prais–Winsten time-series analyses were conducted to examine the association between hair loss SVI and seasonality, adjusted for temperature, hemisphere and per capita GDP.
Trends in monthly ‘hair loss’ SVI followed a cyclical distribution across all countries examined, such that patterns of peaks and troughs in SVI repeat annually. Across all eight countries analysed in aggregate, summer and autumn were associated with greater ‘hair loss’ SVI than spring (coefficients 5·74 and 5·05, respectively, both P < 0·001; Table 1), with the most pronounced increase in SVI occurring in summer. Winter also demonstrated a greater SVI than spring, albeit to a lesser extent than summer and autumn (coefficient 2·63, P < 0·001). Of the confounding variables, temperature was a minor contributor to SVI findings (coefficient 0·18, P = 0·020), while countries with higher per capita GDP were significantly associated with higher SVI (coefficient 0·62, P < 0·001) and countries in the southern hemisphere were associated with lower SVI (coefficient −20·23, P < 0·001).
| Characteristic | Regression coefficient (95% CI) | P-value |
|---|---|---|
| Season | ||
| Spring | Reference | Reference |
| Summer | 5·74 (3·91–7·58) | < 0·001 |
| Autumn | 5·05 (2·94–7·17) | < 0·001 |
| Winter | 2·63 (0·72–4·55) | 0·007 |
| Temperature | 0·18 (0·03–0·33) | 0·020 |
| Southern hemispherea
CI, confidence interval. Seasons for countries in the northern hemisphere were defined such that March, April and May constituted spring; June, July and August constituted summer; September, October and November constituted autumn; and December, January and February constituted winter. Seasons for countries in the southern hemisphere were defined as the reverse, such that northern-hemisphere spring, summer, autumn and winter constituted southern-hemisphere autumn, winter, spring and summer, respectively. aNorthern hemisphere used as the reference. bGross domestic product in thousands.
|
−20·23 (−23·71 to −16·76) | < 0·001 |
| Nominal GDP per capitab
CI, confidence interval. Seasons for countries in the northern hemisphere were defined such that March, April and May constituted spring; June, July and August constituted summer; September, October and November constituted autumn; and December, January and February constituted winter. Seasons for countries in the southern hemisphere were defined as the reverse, such that northern-hemisphere spring, summer, autumn and winter constituted southern-hemisphere autumn, winter, spring and summer, respectively. aNorthern hemisphere used as the reference. bGross domestic product in thousands.
|
0·62 (0·55–0·70) | < 0·001 |
- CI, confidence interval. Seasons for countries in the northern hemisphere were defined such that March, April and May constituted spring; June, July and August constituted summer; September, October and November constituted autumn; and December, January and February constituted winter. Seasons for countries in the southern hemisphere were defined as the reverse, such that northern-hemisphere spring, summer, autumn and winter constituted southern-hemisphere autumn, winter, spring and summer, respectively. aNorthern hemisphere used as the reference. bGross domestic product in thousands.
The results of this secular trend study suggest that hair loss in the population is significantly correlated with seasonality, and that hair loss occurs most frequently in the summer and autumn. These findings are consistent with prior studies that used trichograms and other hair samples to find that telogen hair loss occurs maximally in the summer1, 2 or the transition between summer and autumn.2 However, the physiology of hair loss as related to seasonal variation is unknown. Clinical implications of this pattern of hair loss seasonality include the potential for confounding diagnosis of hair loss conditions or efficacy of treatment started at different months of the year. While temperature was associated with hair loss seasonality in this study, it did not contribute significantly to hair loss in multivariate modelling. However, other seasonal trends were not evaluated. This suggests that other factors are contributive and that future studies exploring the effect of ultraviolet index variation on patterns of hair loss, for example, are warranted.
A limitation of this analysis is that Google Trends SVI data shift slightly depending on acquisition date, as relative percentages of total search volume are continuously altered.5 However, data for the same time range acquired on three separate days demonstrated close correlation (Spearman correlation 0·97). Nevertheless, this is an initial investigation into seasonal patterns of hair loss worldwide. Further studies evaluating the seasonality of hair loss, as well as exploring the effect of other potential mediating factors, are needed.
