Acne vulgaris (acne) is a prevalent skin disorder with substantial physical and psychological morbidity. Although dietary factors have long been considered unimportant, insulin resistance (IR) and dietary carbohydrates have recently been implicated in the etiology of acne [1,2]. The glycemic index (GI) of meals has been directly correlated to insulin response [3] and low GI diets have been shown to decrease IR [4]. High insulin concentrations in the fasting and/or postprandial state may exacerbate acne by increasing the proliferation of basal keratinocytes. Insulin also stimulates the synthesis of androgens leading to high sebum production, a recognized correlate of acne severity [5]. IR could also increase inflammatory responses within and adjacent to the comedo [6].

Postprandial insulin responses may be of particular relevance during puberty and adolescence when whole-body IR naturally increases [7]. Insulin sensitivity and sex-hormone-binding globulin (SHBG) are directly correlated, such that falls in insulin sensitivity increase the free androgen index (FAI) [8,9]. Similarly, compensatory hyperinsulinemia is associated with reductions in insulin-like growth factor binding protein-1 (IGFBP-1), corresponding to higher cellular concentrations of free insulin-like growth factor-1 (IGF-1) [9]. Indeed, the highest incidence of acne occurs when IGF-1 levels peak [10] and adult women with acne have high levels of IGF-1 [11]. Interventions that reduce fasting and postprandial insulinemia and IGF-1 concentrations would therefore be expected to decrease sebum production and keratinocyte proliferation [12].

In the first study of its kind, Smith et al. demonstrated that reductions in carbohydrate intake with particular emphasis on low glycemic index (GI) carbohydrate foods, improved cutaneous and hormonal markers of acne severity in young adult males [2]. Several confounders, however, complicated the interpretation of the study because the treatment group lost more weight, and had higher protein and fiber intake than the control group—factors which the authors discussed. Since weight loss as well as higher protein and fiber intake have all been linked to improved insulin sensitivity, it is not clear whether dietary GI played an important role in the improvement of acne.

In the present study, we hypothesized that in the context of weight maintenance and identical macronutrient and fiber intake, the replacement of high GI carbohydrates with low GI carbohydrates would improve acne severity by lowering postprandial insulin concentrations and/or decreasing IR. We studied adolescent males attending boarding school so that food intakes could be more easily controlled.

A total of 58 subjects met the inclusion criteria, of which 29 were assigned to the high GI and 29 to the low GI diet (per protocol results). Twelve subjects discontinued because they were unwilling to adhere to the diet (four on the low GI diet and eight on the high GI diet) and three subjects were omitted from data analysis due to missing acne gradings (dermatologist appointments were cancelled due to illness). Forty-three subjects completed the study (n = 23 low GI and n = 20 high GI). The mean ± SD age (16.6 ± 1.1 y) and BMI (23.6 ± 3.8 kg/m²) of the completers was not significantly different to the subjects who discontinued (age 16.4 ± 0.7 y and BMI 21.9 ± 2.3 kg/m², p = 0.72 and p = 0.12, respectively). On the whole, there were no significant differences in subject characteristics, baseline acne severity or blood biochemistry between the two diet groups (Table 1). However, at baseline, DHEA-S was higher in those allocated to the low GI diet (p = 0.04). As determined by the study protocol, diets differed significantly in GI (low GI diet 51 ± 1 vs. high GI diet 61 ± 2, p = 0.0002), but not energy, protein, total fat, carbohydrate, fiber, zinc or calcium content (Table 2). Diets also differed significantly in saturated fat (low GI diet 13 ± 0% vs. high GI diet 15 ± 1%, p = 0.01). Changes in weight were minor and not different within or between diet groups (low GI −0.3 ± 0.4 kg vs. high GI 1.1 ± 0.6 kg, p = 0.053).

Facial acne improved on both diets (low GI −26 ± 6%, p = 0.0004 and high GI −16 ± 7%, p = 0.01), but differences between diets did not reach significance (Figure 1, Table 3). Similarly, there were no significant differences within or between diets for most biochemical parameters. However, IGFBP-1 declined significantly on the low GI diet versus the high GI diet (low GI −6.2 ± 3.7 ng/mL vs. high GI 5.0 ± 2.6 ng/mL, p = 0.02) (Table 4). Testosterone increased significantly from 0 to 8 wk on the high GI diet and increased non-significantly on the low GI diet, but there was no effect of diet composition (low GI 3.4 ± 1.8 nmol/L vs. high GI 5.0 ± 1.7 nmol/L, p = 0.50).

Change in acne severity was not correlated with change in HOMA-IR or change in insulin (Pearson correlation r = −0.196, p = 0.244 and Pearson correlation r = −0.202, p = 0.225, respectively).

In the present study we found no effect of GI on facial acne severity over 8 wk and no correlation between changes in acne severity and insulin concentrations or sensitivity. However, our hypothesis that a low GI diet would improve facial acne to a greater extent than a high GI diet was consistent with the finding that acne severity decreased more on the low GI diet (~26%, p = 0.0004) compared with the high GI diet (~16%, p = 0.01). Because the trial duration was dictated by the school term, our study was shorter than the ideal and usual intervention period of 12 wk in acne treatment studies [2,17].

The mean change in acne severity on the low GI diet was a decrease of over half a point out of three, which represents a change that is likely to be highly significant to subjects [18]. More sensitive and objective dermatological assessment tools may have highlighted differences to a greater extent.

Improvements in acne on both the low and the high GI diet occurred despite the concurrent increase in testosterone levels, which reached significance on the high GI diet. Androgen levels increase during adolescence, hence higher testosterone may reflect a period of rapid sexual maturation [9,19]. In practice, androgen concentrations have been directly correlated to the severity of acne [20,21], thus the increases in testosterone might be expected to increase acne severity.

The findings of the present study contrast with those of Smith et al. [2]. Those investigators managed a 13 point difference in GI (43 vs. 56), and GI values were lower in both groups, whereas the current study produced a 10 point difference in GI (51 vs. 61). However, differences in GL were similar because we had slightly lower intakes of carbohydrate (Smith et al. [2] 101 vs. 174; current 102 vs. 157). Saturated fat was significantly different between interventions in both studies, which could have affected results because dietary fats can directly enter the pilosebaceous unit from the blood [22]. Recent evidence suggests that an increased proportion of saturated fatty acids in skin surface triglycerides may decrease abnormal skin cell replication and improve acne [23,24].

Overall, our findings imply that low GI carbohydrates were not responsible for the improvements in facial acne reported by Smith et al. [2]. The increased weight loss on the low GL arm of that study (nearly 3 kg compared to the weight gain of 0.5 kg in the control group) suggests negative energy balance and could explain the declines in fasting insulin and HOMA-IR. The higher protein content and higher fiber of their low GL diet may have contributed to increased satiety and reduced food intake [25,26]. Energy restriction and weight loss per se are recognized to improve fasting insulin concentration and HOMA-IR, independent of diet composition. Even in the absence of weight loss, higher protein or fiber intake may lead to improvements in glucose and insulin metabolism [27,28]. SHBG production also increases after weight loss, implying decreased androgenic activity [8]. Indeed, when Smith et al. adjusted for change in BMI, the differences in total inflammatory plus non-inflammatory acne lesion counts, HOMA-IR, SHBG and FAI between diets were no longer significant (although IGFBP-1 and inflammatory lesion counts remained significant) [2].

We found a significant effect of diet on IGFBP-1 but not IGFBP-3 or IGF-1. However, unlike Smith et al., IGFBP-1 decreased on the low GI diet [2].In an acute study in lean, young healthy subjects, Brand-Miller et al. observed that serum IGFBP-1 declined acutely after both high and low GI meals; the mean change at 4 h being significantly more prolonged after the low GI meal [29]. This is consistent with the decrease in IGFBP-1 in the present study. Considering the improvement in acne within both diets, we might have expected IGFBP-1 to increase on both diets and IGF-1 to decrease [10,21]. It is possible that blood levels of IGF-1 do not directly reflect levels in cells and tissues and that the complexity of IGF and IGFBP interactions and dietary effects on the system makes interpretation difficult.

Unlike Smith et al., we failed to see changes in SHBG or DHEA-S [2]. Improvements in acne severity might be expected to correspond to an increase in SHBG [30] and a fall in DHEA-S [21,31]. In postmenopausal women with hyperandrogenism, a low GI diet was able to increase SHBG and decrease testosterone concentration, although the study duration was longer [32]. Some studies have reported no association between free testosterone and acne severity [31], potentially because measurement of testosterone is prone to error. It is removed quickly from the circulation by the liver and target tissues, and may be produced locally in the pilosebaceous intracrine system [20].

The limitations of our study should be noted. Twelve subjects discontinued the study and three subjects were not available for the final acne gradings, representing a substantial drop out rate (26%). The 4-point acne grading method was novel and hence not internationally validated and capable of detecting only relatively large changes in the severity of inflammatory lesions (a change of one grade or more, e.g., from severe to moderate acne). Less obvious changes in inflammatory and non‑inflammatory lesions may have been overlooked. Assessing acne severity at four weeks would have increased the scope of the study, but this was not possible due to commitments of the dermatologist. Studying adolescents at boarding school may prevent generalization of results to wider non-boarding adolescent population. The non-randomized alternate allocation study design may have resulted in selection bias. A control group (who were not given any dietary advice) would have been useful to determine the true significance of within-diet changes. The subjects were not blinded to the purpose of the study and food intake may have been under- or over-reported. Changes in insulin sensitivity may have occurred but gone undetected because HOMA-IR is a comparatively “blunt” measure of insulin sensitivity, reflecting hepatic rather than peripheral glucose metabolism. Food intake was not measured at baseline, so differences between habitual diet and intervention diets could not be assessed. Food diaries were only completed on weekends because it was felt that asking the students to record their food intake in such detail for a longer period would be too onerous, which meant that any differences between week day and weekend food intakes would have been missed (although the vast majority of students remained at school on weekends and hence week day diet would be similar to weekend diet). Validation of the food diaries (e.g., via weighed food records) may have increased the accuracy of the intake data. Finally, the difference in GI between diets may not have been large enough to materially affect postprandial glycemic and insulin responses in adolescent males. Larger differences (ranging from 13 to 28 units) were employed to effect changes in glycated proteins in diabetic subjects [33]. The GI of our high GI intervention may have been similar to habitual intake.

A myriad of emotional and physiologic factors affect the pathogenesis of acne, making dietary studies in this area especially difficult, including stress [34,35] and nutrients other than carbohydrates (saturated fats [22,24], n-3 vs. n-6LCPUFA [36,37,38,39], vitamin D [40] and dairy [41]). n-3LCPUFA are increasingly recognized as anti-inflammatory agents [36], while n-6 fats are precursors of highly active mediators of inflammation [37]. n-3LCPUFA have been shown to increase IGFBP-3 in animals [38] and decrease IGF-1 in humans [39]. In this study, we found no difference in LCPUFA between diets, but did not differentiate between n-3 and n-6 isoforms.

This study suggests that a low GI diet does not significantly improve facial acne more than a macronutrient-matched high GI diet, although there was a trend for the low GI diet to improve acne to a greater extent than the high GI diet. However, the limitations of the study limit the conclusions that can be made from the results. The lack of a difference between diets may be due to short study duration and a relatively blunt acne severity assessment tool. It is possible that the observations of the previous study investigating dietary GI/GL may be explained by weight loss or differences in protein and/or fiber content. Future studies should employ randomized designs of at least 12 wk duration, subjects within a narrow age range, and using validated acne scoring systems [17]. Tight control of confounding factors, including weight, type and amount of fat, protein and carbohydrate, level of stress, and other limitations discussed above, is necessary.