<i>JPEN</i> Journal Club 91. Choosing the population to study

Irritable bowel syndrome (IBS) is a chronic gastrointestinal disease that afflicts 4%–10% of the general population.1 Its symptoms consist of bowel dysfunction and, typically, constipation or diarrhea (or even both), along with abdominal pain. It usually begins in younger people, more commonly women, and is a member of a larger group of disorders known as functional bowel diseases. Individual foods have been associated with symptoms, but these foods vary from patient to patient. A diet that has been shown to be effective in relieving symptoms in many individuals is one that is low in fermentable oligosaccharides, disaccharides, monosaccharides, and polyols (FODMAPs).2 In 2010, a group of Brazilian scientists introduced a new classification of foods that was based on the use and extent of food processing.3 This system became known as the NOVA classification; the term NOVA is not an acronym but rather a word in the Portuguese phrase “uma nova classificação de alimentos.” The NOVA system initially used three categories of food: unprocessed or minimally processed (group 1), processed culinary ingredients (eg, salt, sugar, vegetable oils, butter, other fats) (group 2), and ultraprocessed foods (UPFs) (group 3). The last category was subsequently divided into groups 3.1 (processed foods) and 3.2 (UPF products), but ultimately, groups 3.1 and 3.2 were reassigned the numbers 3 and 4.4 Group 3 foods were those to which processed culinary ingredients were added; UPFs were food products that were derived from foods (or food parts) and to which nonfood substances not typically found in food are industrially combined to improve the storage capability, appearance, or palatability of the product. UPFs have been the target of numerous commentaries in both the medical and lay press and are widely regarded as unhealthy.5, 6 With regard to the issue of IBS, they have been associated with constipation (albeit not diarrhea) in the general public.7 Wu and five Chinese colleagues wondered if the consumption of UPFs was somehow related to the subsequent development of IBS.8 The report of their study8 is the focus for discussion in this issue of the JPEN Journal Club. Wu et al. used data collected in the United Kingdom Biobank. This latter entity is a large biomedical database that was created to store health information, including genetic, lifestyle, clinical, and follow-up data that could be used to address future questions. It is composed of 22 assessment centers in the United Kingdom that enrolled >500,000 volunteers between 2006 and 2010. Initially, these participants provided baseline questionnaires, anthropometric measurements, samples of body tissues and/or fluids (for DNA testing), and lists of their medical conditions. Between 2009 and 2012, these individuals were asked to complete five web-based 24-h diet-recall questionnaires; >210,000 of them responded at least once. The food intake was classified using the NOVA criteria. Eligible participants in this study had to have dietary-recall information available, to not have had an energy intake on the diet recall that was <500 or >5000 kcal, and to not have IBS diagnosed before the completion of the diet-recall survey nor any previous diagnoses of cancer, inflammatory bowel disease, or celiac disease. A total of 178,711 individuals met these criteria. Each participant's intake was assessed for the percentage of the total diet (in grams) that was composed of UPFs. They were then divided into quartiles of about 44,600 people each. These percentages of dietary UPFs for the quartiles were 0%–13.84%, 13.85%–19.57%, 19.58%–26.78%, and >26.78%. The primary outcome was the subsequent diagnosis of IBS; these diagnoses were established by linkages to primary care and hospital admission records and had to be validated with codes from the International Classification of Disease, 10th Revision (ICD-10). The follow-up for this study concluded in May 2022. A variety of other factors were recorded for each participant. These included age, sex, body mass index (BMI) (separated into underweight, normal, overweight, or obese), ethnicity (White or non-White), a socioeconomic measure (separated into quartiles), education level (university or not), smoking and alcohol status (current, past, or never for each agent), weekly alcohol intake, diabetes, physical activity (low, moderate, or high), total intake of various foodstuffs (energy, protein, fat, carbohydrate, fiber, and sugar), depression, and anxiety. These factors were used to create four models that were characterized by progressively more factors being adjusted in the analyses. These four were (1) raw data, (2) data adjusted for age and sex, (3) data adjusted for most of these factors, and (4) data adjusted for all of the factors. The average age of the 178,711 individuals was 55.8 years, and the average percentage of UPFs consumed was 21%. Those who consumed more UPFs were less likely to have attended university, to have participated in physical activity, or to have adequate socioeconomic status; they were more likely to have diabetes, higher BMI, depression, anxiety, and greater intakes of total energy, protein, fat, and carbohydrate. The average duration of follow-up was 11.3 years. During that time, 2700 individuals (1.5%) were diagnosed with IBS. The cumulative incidence of IBS was 1.7% in the highest quartile of UPF consumers and 1.4% in the lowest quartile; there was a linear increased risk of developing IBS in those consuming more UPFs. In the fully adjusted model, the hazard ratio for the highest quartile was 1.2 (95% confidence interval, 1.07–1.33) compared with the lowest quartile (P < 0.001). It was calculated that, for each 10% increment in UPF consumption, there was a relative 8% higher risk of developing IBS. A variety of subgroup analyses were performed. Associations between higher risks of IBS development and higher UPF consumptions were observed with regard to age, sex, BMI, and smoking or alcohol use; these all seemed to be dose-response effects. This same observation was made when the UPF consumption was divided by 10% increments instead of by quartiles. Wu et al. noted that this relationship between UPF consumption and IBS development was only an association and no causative relationship could be made because the study was an observational one. However, after declaring this limitation, they went on to state “these findings highlight the importance of considering food processing degree and limiting UPF consumption in diets to reduce the modifiable burden of IBS.” (This latter statement seemed to suggest that Wu et al. were proposing a causative effect.) Wu et al. believed that the study provided insight into the risk of developing IBS that was associated with UPF consumption. They based this belief on the facts that the UK Biobank population was a large and well-described prospective cohort, that the dietary-recall questionnaire was validated and provided more details than other such questionnaires, and that their sensitivity analyses addressed the issues of reverse causation, measurement bias, and selection bias. On the other hand, they did acknowledge the presence of other limitations to this study. They recognized that some of the participants only provided one dietary-recall response, so they could not confirm the occurrence of variability in those diets. (They did state that, when they removed those individuals, the findings were similar). They also admitted that for some of the foods, there was not much information available with which to be certain about the processing techniques. Residual confounding, especially related to lifestyle and nutrition factors, may have been present. Because only about half of the UK Biobank population completed the dietary-recall questionnaires, selection bias may have occurred. The diagnosis of IBS required that such a disorder be documented in the records with ICD-10 codes, so other patients with IBS may have been missed. Finally, individuals registered in the UK Biobank are largely White European, so it is not clear how generalizable the findings were. Case-control studies match individuals with a particular health outcome (eg, a specific disease), namely the cases, to others without that outcome, the controls. The investigators collect a variety of pieces of clinical and demographic data (referred to as exposures) from both groups. The trial methodology involves a comparison of the frequencies with which these exposures are present in each group. The objective of such studies is to identify potential associations between an exposure and an outcome. This type of observational study is typically a retrospective process, as the outcome has to either have occurred or not occurred at a particular time. One of the limitations of such studies is recall bias, as the methodology requires the participants to report whether the exposure occurred over a particular period of time, and this recall both may not be completely accurate and may be biased if recall is better in those with the outcome of interest. In contrast, a cohort study approaches this issue from the opposite direction. In such studies, participants have or do not have particular exposures at some time in the past and are assessed for a specific outcome (usually years later). Such studies may be prospective (when only the exposure is known) or retrospective (when both the exposure and outcome are known). Cohort studies are still observational studies, as there is some reason or reasons for the exposure to be present or absent, creating a confounding bias. The study by Wu et al. essentially analyzed participants with or without the outcome (IBS) and assessed whether or not the exposure (UPFs) was present. As such, it would appear to be a case-control one. Because they selected the UK Biobank participants, they were using individuals who had been prospectively followed. Hence, it might be tempting to attribute a causative effect from the UPFs. However, because of potential confounding (why UPFs were consumed or not consumed), such an interpretation would be inappropriate. Unfortunately, there is yet another problem with the study. As we noted, IBS is a disease that usually begins relatively early in life. However, the average age of the study participants was almost 56 years; it is unusual for IBS to develop in people of this age. Thus, when I read the paper, I was not convinced that the diagnoses of IBS were always correct, or at least that the IBS was not comparable to the disease that physicians commonly see in their practices. If this is the case, the described association between IBS and UPFs becomes irrelevant. The data from the UK Biobank may very well be useful for trying to identify relationships between UPFs and diseases that occur later in life. However, studying people who are middle-aged is not going to be helpful for assessing diseases that typically arise in much younger individuals. You may have other thoughts about this paper. If so, I would encourage you to discuss them with your colleagues or even send an email to me ([email protected]). I would also again remind you that the American Society for Parenteral and Enteral Nutrition (ASPEN) has created an information kit with guidelines for how chapters, hospitals, and universities can develop their own journal clubs (www.nutritioncare.org/journalclub). For the next installment, please read the following article: Lamontagne F, Masse M-H, Menard J, et al. Intravenous vitamin C in adults with sepsis in the intensive care unit. N Engl J Med. 2022;386(25):2387-2398. The author acknowledges the ongoing support of GIIssues Inc, a 501(c)(3) nonprofit organization dedicated to the practice and dissemination of evidence-based medicine. GIIssues Inc supports educational expenses for the author as well as any expenses related to the creation of papers and other educational products. GIIssues Inc does not provide any salary support to the author. There are no research materials that are related to this paper that can be accessed other than the stated references. None declared.