Role of glycemic index in the context of an overall heart-healthy diet.

Abstract

Randomized trials designed to assess possible effects of a dietary intervention on hard outcomes such as cardiovascular disease (CVD) events, cancer incidence, or all-cause mortality are nearly nonexistent. Clinical end point studies such as these require large sample sizes, long-term follow-up measured in years, and high levels of dietary adherence to attain a valid result. These studies are difficult and expensive to conduct and are consequently rare. An important exception was the PREDIMED Study wherein a Mediterranean diet supplemented with extravirgin olive oil or nuts reduced the incidence of major CVD events (mostly stroke) at 4.8 years in persons at high CVD risk.1 Based on many short-term studies and the few long-term studies such as PREDIMED, dietary recommendations currently favor an overall diet similar to the Mediterranean diet for CVD prevention.2,3 Unlike PREDIMED, most of the dietary literature consists of short-term interventions and surrogate (biomarker) outcome studies. The study by Sacks et al4 in this issue of JAMA represents an optimal experimental design of this type. The authors designed a randomized crossover feeding study to test alternative dietary patterns on important biomarkers of CVD risk among 163 overweight adults with prehypertension or mild hypertension. Four different diet patterns were compared, with food supplied to participants, each diet eaten for 5 weeks followed by at least a 2-week washout period. Two of the diets were high in calories from carbohydrates (58% energy), with either low– or high– glycemic index foods, and 2 of the diets were low in total carbohydrates (40% energy), also with either a predominance of low– or high–glycemic index foods. Low glycemic index was defined as less than 45 on the glucose scale, while high glycemic index was greater than 65 on the glucose scale. Each participant consumed 2 of the 4 diets in randomly assigned order, and overall, the 4 diets were compared. There has been a great deal of interest in diets with high–glycemic index foods for more than a decade. Therewere 5primaryoutcomes that included insulin sensitivity; levels of low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, and triglycerides; and systolic bloodpressure. Theparticipantswere a relatively high-risk group: 56% were obese, 26% had hypertension and the others prehypertension, 56% had an LDL cholesterol level greater than 130 mg/dL, and 30% exhibited impaired fasting glucose (>100 mg/dL). Many of the results were contrary to what had been expected. When glycemic index was lower in the highcarbohydrate diet, insulin sensitivity not only did not increase but decreased. With the same diet pattern, levels of LDL cholesterol and apolipoprotein B (a secondary end point) increased, with no changes in HDL cholesterol or triglyceride level or blood pressure. Insulin sensitivity was assessed from glucose and insulin concentrations after 75 g of oral glucose over 2 hours using the Matsuda index.5 Notably, fasting glucose was higher with the low–glycemic index, highcarbohydrate diet than with the high–glycemic index, highcarbohydrate diet: an effect the authors postulated was secondary to increases in morning insulin resistance, a compensation speculated to maintain fasting glucose. Alternatively, a lower–glycemic index diet over 5 weeks could have resulted in a modest and transient reduction in beta-cell function/insulin secretion.6 In this short-term, surrogate end point study, it remains unclear whether this small change in insulin sensitivity has any prognostic value, eg, a higher risk of type 2 diabetes, in this at-risk population, as suggested for diets with high glycemic load.7 Compared with the baseline LDL cholesterol level, all 4 study diets lowered LDL cholesterol; however, the significant increase in LDL cholesterol with the low–glycemic index, high-carbohydrate diet vs the other diets is of concern, although there is no apparentmechanism to explain this finding. HDL cholesterol was unchanged, but triglycerides were loweredwhendietswere restricted in total carbohydrate, glycemic index, or both. Yet only 17% of these participants had fasting triglycerides greater than 150 mg/dL. The glycemic index and glycemic load were first defined by Jenkins et al8 in 1981. Glycemic index describes a particular type of food and indicates the effect of that specific food on a person’s blood glucose after ingestion. The concept is based on observations that certain carbohydrate-containing foods increase blood glucose levels when ingested as single foodsmore so than other carbohydrate-containing foods containing similar amounts of carbohydrate and total calories. The glycemic load of a food serving is calculated by its carbohydrate content in grams multiplied by the food’s glycemic index divided by 100. Although these definitions perform consistently for an individual food eaten alone, concerns have arisen when considering the same food ingested as part of a meal in whichmany foods are consumed with or without fiber or other nutrients. Perhaps the most important clinical application of glycemic index or load is among patients with diabetes for whom postprandial glucose excursion and meaRelated article page 2531 Opinion