martes, 28 de febrero de 2017

Gastrointestinal disorders associated with migraine: A comprehensive review

Resultado de imagen de migrañas con aura
Core tip: Migraine is a recurrent and disabling primary headache disorder that commonly affects a significant proportion of the global population. Recent reports demonstrate an increased frequency of gastrointestinal (GI) disorders in patients with migraine compared with the general population. We review the available literature linking migraine with GI complications and comorbidities.

INTRODUCTION

Migraine is a disabling primary headache disorder, defined by the International Classification of Headache Disorders as recurrent, moderate to severe headache attacks lasting 4-72 h with associated features including nausea and/or vomiting[] that affects over 17% of women and 5%-8% of men[,]. Recent publications have suggested that its worldwide prevalence may surpass 20%, and that it consistently rates as one of the most disabling conditions[]. The subtype of chronic migraine affects up to 2%-5% of the population worldwide[,].
Susceptibility to migraine is thought to be multi-factorial, with genetic hormonal and environmental factors all playing an important role. Unbiased genome-wide association studies have identified 13 migraine-associated variants pointing at genes that cluster in pathways for glutamatergic neurotransmission, synaptic function, pain sensing, metalloproteinases and vasculature[]. The physiopathology of migraine is complex and the precise mechanisms and pathways involved remain to be fully elucidated. Many different neuropeptides, neurotransmitters and brain pathways have been implicated, but whether pain generates in central or peripheral structures is a matter of debate. A review of all of these mechanisms is beyond the scope of this article, and excellent recent papers by Noseda et al[] and Burstein et al[] thoroughly summarize the current concepts of migraine physiopathology.
Briefly, sequential steps of neurogenic inflammation, peripheral trigeminovascular input, and central cortico-trigeminal nuclei activation are thought to mediate migraine pathogenesis. Headache arises initially from an inflammatory response in the dura, mediated by vasoactive peptides, including calcitonin gene-related peptide (CGRP), substance P and neurokinin A. These are all released by trigeminal fibers and lead to activation of nociceptive perivascular sensory nerve terminals located in the meningeal vasculature[]. Perpetuation of headache is thought to be secondary to increased cortical responsiveness, while cortical spreading depression (CSD) has been hypothesized to represent the pathophysiological correlate not only for crisis onset but also for migraine aura.
The many stages of migraine headache from prodromal to postictal symptoms involve alterations in multiple cortical and subcortical structures. CSD itself can also activate the trigeminovascular system. Nociceptive input from trigeminal fibers converge onto the spinal trigeminal nucleus, which later modulates the activity of second-order structures such as the ventrolateral area of the upper cervical and medullary dorsal horn, the periaqueductal gray matter, rostral trigeminal spinal nuclei, brainstem reticular areas, superior salivatory nuclei and the cuneiform nuclei[,]. Activity in these brain regions modulates the myriad of symptoms that follow migraine headache. The frequency of accompanying autonomic symptoms, such as nausea, vasoconstriction, vasodilation and diaphoresis as well as the participation of the hypothalamus, has led some investigators to propose that autonomic dysfunction may also play an important part in migraine pathophysiology.
It is widely known that migraine is associated with a variety of comorbidities, mainly cardiovascular, psychiatric and other neurological conditions. Hypertension, hyperlipidemia, sinusitis, asthma, pulmonary emphysema, insomnia, affective disorders and fibromyalgia have all been associated with migraine. Recent studies have found that gastrointestinal (GI) disorders also appear to be more frequent in patients with migraine (PWM) than in the general population[]. In this review, we survey the available literature linking migraine with GI complications and comorbidities.

HELICOBACTER PYLORI INFECTION

Interest in a possible association between Helicobacter pylori (H. pylori) infection (the most relevant cause of gastritis and peptic ulcer) and migraine first arose after this microorganism was recognized as the cause of myriad extra digestive manifestations, including neurological diseases[,]. Its association with both cardiovascular (including typical functional vascular disorders such as primary Raynaud phenomenon) and autoimmune diseases has been established, possibly due to a chronic inflammatory response with local secretion to the circulatory system of numerous inflammatory and vasoactive mediators[,]. Both vascular and inflammatory hypotheses have been proposed as mechanisms mediating migraine physiopathology, making a link between H. pylori and migraine at least plausible enough to warrant investigation.
The evidence for a possible association between H. pylori infection and migraine is surrounded by some controversy. In one study[], investigators performed endoscopic procedures in 31 children with migraine and abdominal complaints and found a remarkably high prevalence of esophagitis (41.9%), corpus gastritis (51.6%), antral gastritis (38.7%) and duodenitis (87.1%). However, only 7 had H. pylori, thus failing to support an association between H. pylori and migraine.
A study assessing 200 subjects affected by primary headache found a H. pylori infection prevalence of 40%. It also found a higher prevalence of migraine without aura in infected patients[]. A similar study performed on 225 PWM found that 40% were colonized by H. pylori, and that the intensity, duration and frequency of migraine attacks were significantly reduced in all patients who underwent H. pylori eradication[]. These results were questioned by a later case-control study of 103 PWM and 103 control subjects, in whom the proportion of H. pylori infection was almost identical; in addition, there were no clinical or demographic differences between colonized and non-colonized migraine patients[]. In a follow-up of the former study[], 175 PWM were compared with 152 matched controls and investigators found no difference in prevalence of infection (40% vs 39%, respectively). However, among infected subjects, they found a significantly higher prevalence of CagA-positive strains in patients affected by migraine with aura compared with those affected by migraine without aura (41% vs 19%) and with controls (41% vs 17%), suggesting a pathogenic role for that specific strain of H. pylori. In contrast, a group of Turkish investigators compared the prevalence of H. pylori infection in 70 PWM and found a greater prevalence when compared to 60 matched controls, as well as a slight clinical benefit after eradication therapy[].
Besides different H. pylori strains, differences between migraine subtypes could also explain some of these inconsistent results. In a study on 49 PWM without aura, H. pylori infection was more prevalent compared to controls, and interestingly investigators also found much higher prevalence in PWM without family history or hormonal fluctuations, suggesting that H. pylori infection could be particularly relevant in patients with fewer known risk factors for migraine[]. However, an earlier study had reported that the association with H. pylori was exclusive for migraine with aura[]. In an attempt to resolve the controversy over the epidemiological association between migraine and H. pylori, a recent meta-analysis that included five studies and 903 patients found an overall H. pylori infection rate in PWM of 45%, compared to 33% in controls, with subgroup analysis finding greater infection rate of H. pylori in Asian patients compared to Europeans, but no difference among migraine subtypes[]. Recently, a case-control study demonstrated higher IgM antibody titers against H. pylori in PWM compared to controls; also, they found a positive correlation between IgG antibody titers and severity of migraine[]. This suggests the importance of H. pylori active infection in migraine.
The previous epidemiological studies suggested that eradication therapy could be beneficial for migraine control, but in a non-controlled setting. In the only available double-blind, randomized, controlled clinical trial, 64 patients diagnosed with migraine-type headache were randomized to receive migraine treatment and H. pylori eradication treatment or migraine treatment and placebo[]. Using the Migraine Disability Assessment (MIDAS) questionnaire, on enrollment patients in the treatment group had more severe symptoms, but these differences disappeared after completing the study. Analysis of the change in MIDAS scores between baseline and completion of the study revealed a slight benefit for the treatment group.
As for possible pathophysiological mechanisms, few studies have made relevant findings. In small studies evaluating the redox state of PWM, H. pylori infection did not influence plasma accumulation of peroxidative substances, values of nitrite/nitrate concentrations or expression of systemic nitric oxide[,]. Others have speculated that elevated levels of interleukin-10 might be implicated, considering that studies have shown elevations in this cytokine both in PWM as well as in patients with infection with H. pylori, particularly with CagA-positive strains[,]. This has a therapeutic implication since the administration of sumatriptan (5-HT1D receptor agonist) reduces the levels of interleukin-10 during a migraine attack[]. The serum levels of CGRP, which has been suggested as a biomarker for chronic migraine[], are also higher in those with H. pylori-associated duodenal ulcers compared with healthy controls[].
In conclusion, whereas epidemiological evidence appears to support an association between migraine and H. pylori infection, the data is limited and investigations should focus on subgroups of patients and of different ethnicities, and should consider the regional variations in H. pylori infection[]. Furthermore, intervention studies suggest a small benefit for eradication therapy, but the long-term benefit has not been established, and a possible intrinsic role for antibiotic or antacid treatment cannot be ruled out completely[]. Available studies are heterogeneous in both populations and treatments used, both in migraine control and H. pylori eradication.

IRRITABLE BOWEL SYNDROME

Irritable bowel syndrome (IBS), considered a neuro-gastroenterologic functional disorder, shares some environmental risk factors with migraine (predominately affecting the female sex and younger individuals), and could be associated with conditions of smooth muscle dysfunction. IBS is associated with a number of extra-intestinal manifestations, and both diseases are widely prevalent and share many somatic and psychiatric comorbidities[]. In systematic reviews of the existing literature on IBS and its comorbidities, patients were found to present a 2-fold increase in other somatic disorders, suggesting a common physiopathological mechanism[]. Indeed, chronic headache was reported in 34%-50% of all IBS patients[,].
Large population-based studies of the prevalence of IBS and accompanying symptoms show migraine, as well as heartburn, dyspepsia, flushing, palpitations and urinary symptoms, to be more common in patients with IBS[]. Specifically, out of 1620 subjects, 350 fulfilled criteria for IBS, and of these, 32% complained of migraine-type headache, compared with 18% in controls. Of course, within a population of PWM, comorbid constellations vary, differing in headache and psychosocial profiles, highlighting the heterogeneity of environmental and genetic factors[]. Nevertheless, an epidemiologic association has been since confirmed by various studies. In a cohort of 208 patients with IBS, 17% had migraine, compared to 8% in 1240 controls[]. A more recent cohort study investigating 97593 IBS patients exhibited a migraine prevalence of 6%, compared to 2.2% in healthy controls[]. A migraine cohort of 14117 newly-diagnosed patients presented similar results, with an adjusted incidence of IBS 1.95-fold higher than in controls (73.87 vs 30.14 per 10000 person-years), particularly in those under 30 years of age[].
There is some indication that comorbid headache disorders in IBS patients could negatively alter their clinical course. In a study of IBS patients comparing first-time attendees with chronic attendees to an outpatient clinic, 40% of first-time attendees complained of mild headache and 1% of severe headache, compared to 59% and 23% of the chronic attendees, respectively[].
In a prospectively identified, hospital-wide population of migraine patients, another study found that 70% of patients met the Rome III criteria for concurrent functional gastrointestinal disorders (FGID), with 40.4% meeting criteria for IBS[]. The authors also demonstrated a clear link between coexistent FGID symptoms, psychological comorbidity and worse scores in anxiety/depression scales.
In the search for a common pathogenetic mechanism for IBS and migraine, neuroendocrine factors, immunological factors, the brain-gut axis and even the intestinal microbiota have been postulated[]. A role for serotonin (5-HT) has been postulated in both diseases, as well as for other factors such as biopsychosocial dysfunction, heredity, genetic polymorphism, central/visceral hypersensitivity, somatic/cutaneous allodynia and the neurolimbic pain network[]. Another possible physiopathologic link could be derived from treatment studies. A recent small, clinical study, showed that an IgG-based food elimination diet could reduce symptomatology and attack prevalence of both disorders in patients with comorbid IBS and migraine[].

GASTROPARESIS

Gastroparesis can be defined as delayed emptying of the stomach in the absence of mechanical obstruction, and its clinical manifestations include nausea, vomiting, bloating and weight loss, among other symptoms[]. An association between migraine and alterations in gastric motility has been noted, describing this motility disorder as a stomach with a functional “vagotomy”[]. A delay in gastric emptying time and increased pyloric tone characterized these alterations. The recognition of GI stasis in PWM led to initial concerns on the absorption of analgesics and the effect on therapeutic efficacy, but some authors emphasized the accentuation of overall distress and inconvenience caused by GI symptoms[]. Studies confirmed an effect of gastric emptying delay time over the absorption of paracetamol and acetylsalicylic acid, among other common analgesics[]. One study established the close association of gastroparesis with migraine attacks. PWM during a pain-free period showed normal gastric emptying times measured with an epigastric impedance method, but these times were significantly delayed during severe or moderate attacks, and delay times were significantly correlated with the intensity of headache, nausea and photophobia[].
The physiology of emesis during migraine attacks could, in some manner, mirror those of gastroparesis, and these considerations have had therapeutic implications, as both dopaminergic and 5-HT4 agonists have prokinetic properties[]. Early findings of an apparently normal gastric motility outside of migraine attacks[] suggested that the physiopathology of gastroparesis in migraine was mediated by pain mechanisms, such as adrenergic and endogenous opiates, or to factors shared with the pathophysiology of a migraine attack itself. An excessive sympathetic response coupled with a decreased parasympathetic tone has also been postulated[]. However, a recent study with gastric scintigraphy showed that gastric emptying time is delayed in both ictal and interictal periods, suggesting an alteration in enteric autonomic function[,]. Another study showed that PWM have meal-induced hypersensitivity of the stomach, due to a low postprandial discomfort threshold, irrespective of the presence of dyspepsia[].
More recent studies have further characterized migraine-associated gastroparesis, but these have also questioned the existence of an alteration outside the ictal period. Using gastric scintigraphy in 3 PWM, investigators found that gastric emptying delay occurs not only in ictal and interictal periods, but also in both drug-induced and spontaneous migraine attacks[]. However, other groups using similar methods were not able to find differences in emptying time between patients without migraine and PWM in the interictal period[]. Of note, both studies included a very small sample, and only the latter included an age- and sex-matched control group. In a larger gastric scintigraphy study, involving 27 PWM and 12 healthy controls, again there was no gastric emptying delay interictally[]. However, in this study, PWM who experienced other GI symptoms were excluded. The controversy over this issue also stems from methodological issues, such as the criteria used to define gastric emptying delay, which should be adapted to the specific population studied, as well as small sample sizes and inadequate control group selection[].
Although there is enough evidence to link gastroparesis with migraine, the nature, causes, correlates and consequences of gastric stasis in migraine are just beginning to be elucidated[,]. There is controversy over whether gastroparesis occurs both ICTALLY and interictally, but it is clear that it is associated with increased discomfort and affects the effectiveness and absorption of orally administered drugs. This suggests that non-oral formulations of commonly used migraine medications as well as the addition of prokinetic drugs could theoretically offer an advantage[]. At least one small trial showed an additive effect of a prokinetic drug (trimebutine) over the efficacy of rizatriptan, a 5-HT1B/1D agonist[].

OTHER FUNCTIONAL GI DISORDERS

Other functional GI disorders have been linked to primary headaches, but there is doubt as to a specific link with migraine. In population-wide registries, both diarrhea and constipation are significantly more frequent in headache sufferers compared to the general population, with no apparent difference between PWM and non-PWM[]. In a cross-sectional study of 326 children, nearly 20% of those who complained of headaches had constipation, a significantly higher number than in controls[]. However, no such association was found among PWM. Another retrospective study of 96 patients with primary headache also found comorbid constipation in 25%, but this was mostly associated with tension-type headaches[]. All studies reported a close correlation of constipation with headache severity, suggesting that it is this factor, together with the related affective and emotional distress, that more adequately explains the association. In a well-designed web-based survey aiming to screen for symptoms of reflux, among 1832 PWM, 22% reported having the diagnosis of GERD, 11.6% reported experiencing heartburn, and 15.8% reported undiagnosed reflux symptoms[]. The most common used medications were triptans but a significant number used NSAIDS. Whether reflux symptoms are a side effect of these medications or are independently associated with migraine is unknown.

HEPATO-BILIARY DISORDERS

There is little evidence linking migraine with hepato-biliary disorders. In a community study on the prevalence of chronic complaints, migraine and some types of biliary colic or right upper abdominal quadrant discomfort occurred together with some regularity, but a statistical association was not established[]. In another study, 488 healthy patients and 99 migraine patients reported upper abdominal symptoms, including unexplained right upper quadrant discomfort. These symptoms were more than twice as frequent in PWM, after adjusting for age, sex, smoking and consumption of analgesics and alcohol[]. A study of twin samples including a cohort of 1200 patients, found an association between migraine and biliary tract disorders, with higher ORs in monozygotic twins (OR = 3.5) than in dizygotic twins (OR = 1.7-2.7)[]. Waist circumference and female sex were also associated with migraine, but the association with biliary tract disorders remained even after controlling for these factors. The stronger association in monozygotic twins suggests a genetic influence. A weakness of the study was that the migraine diagnosis was based on iterated questionnaires and personal interviews and not guideline-based criteria.
As with other functional GI disorders, such as IBS, functional biliary tract disorders would be expected to be more robustly represented in PWM. In a recent study involving 972 patients with biliary dyskinesia (over 80% women), 14.6% were PWM[], a proportion similar to worldwide prevalence[,]. In this study, 30% of the cohort presented comorbid anxiety or depression as well. Interestingly, migraine was an independent predictor (OR = 2.13) for continued medical resource utilization (for recurrent attacks of biliary symptoms), which could suggest a more severe course in PWM.
In experimental studies, cholecystokinin (CCK) coexists with CGRP in the trigeminal ganglion; and trigeminal ganglion stimulation was able to induce local increases of CCK[]. CGRP is also able to influence biliary motility in vitro, and impaired CGRP release is associated to biliary tract disease in humans[,]. This evidence is proposed as a possible common physiopathologic mechanism linking biliary tract disorders and migraine, in addition to a possible role of obesity and female hormones and a vasodilatory effect of CCK[,].
An association between liver disease and migraine and other headache types is even rarer and of dubious physiopathological standing[]. A recent cross-sectional study showed that migraine patients with non-alcoholic fatty liver disease (NAFLD) had significantly more attacks and a higher frequency of auras, but also higher waist circumferences and metabolic disturbances[]. The study was not designed to establish an independent association of NAFLD and migraine, but did suggest a more aggressive disease with a simultaneous diagnosis. Obesity and metabolic disturbances, which are important determinants of NAFLD, are also associated with an increased risk of migraine[,].
Based on the available data, CCK, which seems to have a role in the physiopathology mechanism in migraine[,,], is released in response to fatty acids in the proximal intestine[,]. In this respect, a cross-over study including 83 PWM randomly assigned to a low lipid diet or a normal lipid diet found that the number (2.9 ± 3.7 vs 6.8 ± 7.5, P < 0.001) and severity [1 pt indicated mild, 2 pts moderate and 3 pts very severe headache (1.2 ± 0.9 vs 1.7 ± 0.9, P < 0.01)] of migraine attacks significantly decreased during the low lipid diet intervention periods[]. Likewise, a cross-sectional study showed that after a weight loss program, those who achieved a significant weight loss and metabolic control presented improvement of migraine[].

CELIAC DISEASE

Celiac disease (CD) is an autoimmune condition that occurs in those genetically predisposed to dietary gluten hypersensitivity, affecting about 1% of the general population[]. Besides GI symptoms, it is now known to have systemic involvement. Neurological complications are well-known manifestations of CD, including epilepsy, occipital calcification and migraine-like headaches; also, celiac antibodies have been described in patients with a wide range of neurological disorders, including encephalopathy, ataxia, neuropathy and myopathy, among others[]. Nonetheless, it remains unclear whether gluten sensitivity contributes to the pathogenesis of these disorders or whether it represents an epiphenomenon. Moreover, case reports of patients with concomitant CD and migraine describe the total disappearance of headaches after treatment of CD, and others describe particular combinations of signs and symptoms, such as CD, cerebral calcifications and migraine[,], suggesting the existence of a pathophysiological association. Other single-case reports have described migraine as the presenting symptom in patients later confirmed to have CD[].
The search for antibodies associated with CD in migraine patients aims at establishing an epidemiological and possibly a physiopathological association. In an early study on the association between celiac antibodies and children with neurological disorders, including migraine, epilepsy and hypotonia, among others, investigators found only 13% of cases of anti-gliadin IgG antibodies and no cases of endomysial antibodies[]. Cases were not followed up with biopsies, but none seemed to have clinically evident CD. Another study involving 25 PWM (ages: 14-37; 22 females) did not find a single case of anti-gliadin antibodies in peripheral blood samples[]. However, the limitation of anti-gliadin antibodies is its low diagnostic accuracy; these antibodies have a high prevalence in normal population[]. Therefore, the reported association between CD and migraine using this test may reflect a sampling error. In a more recent study of 100 children with migraine, only 2% had serologic tests positive for CD antibodies, a finding not different from a control group of 1500 healthy subjects[]. The only positive study, which included 73 children with migraine, found that 5.5% of patients have positive celiac antibodies compared with 0.6% in the control group[]. However, most of the seropositive children had normal duodenal biopsies, putting in doubt the diagnosis of CD. On the other hand, only 1 patient out of 147 controls was seropositive, a prevalence lower than what would be expected in the general population. Together, this data does not support screening for CD antibodies in PWM, but does not exclude a causal link in those affected with both disorders simultaneously.
The occurrence of few cases of CD in PWM and the high frequency of headaches in CD provide some support for a possible asymmetrical association. In a study of 72 adult patients with biopsy-proven CD who were screened for neurological disorders, 28% had migraine, among other neurological symptoms[]. A recent study of 188 patients with proven CD found that they had a significantly higher prevalence of migraine headaches compared with controls (OR = 3.79), particularly in women and those under 65 years of age[]. Additionally, there was a trend of more severe headaches in CD patients compared with controls. In a retrospective study of 354 children with CD, 24.8% had experienced headaches, compared to 8% in an age and sex-matched control group[]. The screening for migraine in patients with CD would, therefore, seem to be justified.
The occurrence of more severe headaches in patients with CD and migraine raises the question of a possible therapeutic effect of a gluten-free diet. In a study of 4 PWM and serologically and endoscopically confirmed CD, a gluten-free diet resulted in complete remission of migraine symptoms in one patient and improvement in frequency, duration, and intensity of migraine in the other 3[]. These improvements were associated with a reduction in single photon emission CT regional baseline brain tracer uptake in all 4 patients. Of note, these 4 patients were screened out of a population of 90 PWM (less than 5%). In another study of a cohort of Italian patients with CD, migraine-type headaches were more common than in controls (32%) and subsided partially with gluten-free diet[]. Other large cohorts of CD patients have similarly described improvement in migraine-type headaches with dietary intervention[,].
Although the link between CD and migraine is not fully elucidated, they do share many psychosocial and pathophysiological characteristics. A physiopathological link is suggested by imaging studies. In a case series of 10 patients with CD and episodic headaches, brain MRI showed diffuse and heterogeneous hyperintensities involving white matter, similar to lesions described in PWM[]. The majority of those patients had symptom improvement with a gluten-free diet. Authors have speculated that a state of hypervigilance, associated with an exaggerated response to future threats and episodic attacks, could transform a genetically sensitive nervous system into one susceptible to the alterations underlying these chronic and disabling diseases[]. Another hypothesis is that neurological complications in CD may be caused by a general inflammatory response, rather than be directly antibody-mediated. This idea is supported by a study that showed no correlation between neural antigens and neurologic symptoms in patients with CD[]. Hypothetically, elevated levels of interferon-gamma and tumor necrosis factor-alpha, both independently implicated in migraine and CD, and known to modulate the neuropeptide CGRP, could explain the apparition/progression of migraine symptoms in patients with CD[].
A summary of GI disorders associated with migraine, proposed physiopathological mechanisms, and clinical implications is presented (Table (Table11).
Table 1
Summary of gastrointestinal disorders associated with migraine

MIGRAINE AND THE GUT MICROBIOME

There is evidence that suggests gut microbiota can modulate the brain-gut axis through many pathways, with a potential to influence brain function and nociceptive behavior[,]. The intestinal surface contains 100 trillion microorganisms, separated from the host by a layer of columnar intestinal epithelial cells. Indirect links have been made between gut microbiota and the function of the major pathophysiological mechanisms associated with migraine: Serotoninergic transmission, CGRP activity and cortical reactivity[]. Dysbiosis has an impact on immune function, epithelial barrier permeability, absorption and metabolism of nutrients affecting, in consequence, GI and central nervous system (Figure (Figure11).
Figure 1
Role of the gut microbiota in migraine. Immunological, endocrine, metabolic and neural signals are important pathways by which the gut microbiota influences brain functions. Altered gut microbiota (dysbiosis) affects the normal assimilation of nutrients ...
The serotoninergic system has been shown to be differentially affected by the gut microbiota in experimental studies. Whole-brain tryptophan, tyrosine and glutamine levels are lower in germ-free mice compared to those re-colonized by normal microbiota[], while concentrations of 5-HT and 5-hydroxyindoleacetic acid in hippocampal slices are elevated in germ-free mice[]. Furthermore, circulating levels of 5-HT and tyrosine are elevated in germ-free animals, compared to those with normal gut microbiota[,].
CGRP functions not only as a transmitter but also as a gut hormone, and its signaling could be influenced by microbiota through multiple pathways[]. Although studies demonstrate changes in the expression of sensory-related peptides in the gut by modulating the whole microbiome, no direct effect has been found for CGRP[]. However, at least one study did show that after treatment with the probiotic Pediococcus acidilactici, the number of CGRP-immunoreactive neurons increased in the submucosal plexus ganglia of the small intestine[], although no such effect was observed with Saccharomyces boulardii[].
Whether the gut microbiome has any effects on large-scale cortical function is a matter of theoretical debate. In an interesting study, a group of healthy volunteers who were given fermented milk product with probiotics (Bifidobacterium animalis subsp Lactis, Streptococcus thermophiles, Lactobacillus bulgaricus, and Lactococcus lactis subsp Lactis) showed a reduced task-related response of a distributed functional network in affective and viscerosensory cortices on functional MRI[]. The results of a small, non-randomized trial, including 29 migraine patients, showed a significant reduction in migraine severity after 12 wk of probiotic supplementation compared to baseline[]. Moreover, an open-label study using a combination of two nutritional formulations (combination A: Enzymatically rendered fish protein high in bioactive peptides and amino acids plus probiotics and chlorophyll; combination B: Composed of 21 different ingredients designed to improve the nutritional status of the kidneys and liver) demonstrated a significant and sustained improvement in quality of life (determined by the Migraine Specific Quality of Life Questionnaire), supporting the idea that dysbiosis and altered assimilation of nutrients could have an important role in the physiopathology of migraine[].
There is no direct evidence to conclusively support that the gut microbiome can affect migraine. However, the prospects of a therapeutic strategy based on probiotic dietary interventions or modifications of the gut microbiome, considering that these would intuitively have a high safety profile and cost-effectiveness, make this issue an interesting topic for further research.

FUTURE AREAS OF RESEARCH

Several unanswered questions arise related to this topic. Therefore, further research in GI disorder associated to migraine is warranted in order to evaluate the real impact of some screening and therapeutic measures as well as to clearly define the common inflammatory and neurotransmitter pathways in GI disorders and migraine (Table (Table22).
Table 2
Future areas of interest on gastrointestinal disorders associated with migraine

CONCLUSION

Currently, sufficient evidence exists linking the increased frequency of several GI disorders with migraine compared to the general population. The gut-brain axis plays an important role in the association between GI disorders and migraine. Multiple inflammatory and vasoactive mediators are significantly implicated in the physiopathology of migraine, mainly through the GI microbiota modulation of the GI immunological and autonomic system.
Based on the several implicated mechanisms between different GI disorders and migraine, several pharmacologic and non-pharmacologic therapeutic options for specific GI disorders have shown to improve frequency and severity of migraine attacks. Also, based on the implicated mechanisms, some screening measures (e.g., H. pylori infection) seem to be justified in PWM. Treatment of GI comorbidities in migraine might not only lead to a better quality of life but could also open roads for novel therapeutic strategies for this prevalent and disabling disease.

Brain-gut axis in the pathogenesis of Helicobacter pylori infection.

Resultado de imagen de Helicobacter pylori

Abstract

Helicobacter pylori (H. pylori) infection is the main pathogenic factor for upper digestive tract organic diseases. In addition to direct cytotoxic and proinflammatory effects, H. pylori infection may also induce abnormalities indirectly by affecting the brain-gut axis, similar to other microorganisms present in the alimentary tract. The brain-gut axis integrates the central, peripheral, enteric and autonomic nervous systems, as well as the endocrine and immunological systems, with gastrointestinal functions and environmental stimuli, including gastric and intestinal microbiota. The bidirectional relationship between H. pylori infection and the brain-gut axis influences both the contagion process and the host's neuroendocrine-immunological reaction to it, resulting in alterations in cognitive functions, food intake and appetite, immunological response, and modification of symptom sensitivity thresholds. Furthermore, disturbances in the upper and lower digestive tract permeability, motility and secretion can occur, mainly as a form of irritable bowel syndrome. Many of these abnormalities disappear following H. pylori eradication. H. pylori may have direct neurotoxic effects that lead to alteration of the brain-gut axis through the activation of neurogenic inflammatory processes, or by microelement deficiency secondary to functional and morphological changes in the digestive tract. In digestive tissue, H. pylori can alter signaling in the brain-gut axis by mast cells, the main brain-gut axis effector, as H. pylori infection is associated with decreased mast cell infiltration in the digestive tract. Nevertheless, unequivocal data concerning the direct and immediate effect of H. pylori infection on the brain-gut axis are still lacking. Therefore, further studies evaluating the clinical importance of these host-bacteria interactions will improve our understanding of H. pylori infection pathophysiology and suggest new therapeutic approaches.

lunes, 27 de febrero de 2017

Diet-dependent acid load and type 2 diabetes: pooled results from three prospective cohort studies

Resultado de imagen de diabetes tipo 2

Abstract

Aims/hypothesis

Studies suggest a potential link between low-grade metabolic acidosis and type 2 diabetes. A western dietary pattern increases daily acid load but the association between diet-dependent acid load and type 2 diabetes is still unclear. This study aimed to assess whether diet-dependent acid load is associated with the risk of type 2 diabetes.

Methods

We examined the association between energy-adjusted net endogenous acid production (NEAP), potential renal acid load (PRAL) and animal protein-to-potassium ratio (A:P) on incident type 2 diabetes in 67,433 women from the Nurses’ Health Study, 84,310 women from the Nurses’ Health Study II and 35,743 men from the Health Professionals’ Follow-up Study who were free from type 2 diabetes, cardiovascular disease and cancer at baseline. Study-specific HRs were estimated using Cox proportional hazards models with time-varying covariates and were pooled using a random effects meta-analysis.

Results

We documented 15,305 cases of type 2 diabetes during 4,025,131 person-years of follow-up. After adjustment for diabetes risk factors, dietary NEAP, PRAL and A:P were positively associated with type 2 diabetes (pooled HR [95% CI] for highest (Q5) vs lowest quintile (Q1): 1.29 [1.22, 1.37], ptrend <0.0001; 1.29 [1.22, 1.36], ptrend <0.0001 and 1.32 [1.24, 1.40], ptrend<0.0001 for NEAP, PRAL and A:P, respectively). These results were not fully explained by other dietary factors including glycaemic load and dietary quality (HR [95% CI] for Q5 vs Q1: 1.21 [1.09, 1.33], ptrend <0.0001; 1.19 [1.08, 1.30] and 1.26 [1.17, 1.36], ptrend <0.0001 for NEAP, PRAL and A:P, respectively).

Conclusions/interpretation

This study suggests that higher diet-dependent acid load is associated with an increased risk of type 2 diabetes. This association is not fully explained by diabetes risk factors and overall diet quality.

Keywords

Acid–base balanceDietary acid loadGlucose intoleranceInsulin resistance

Abbreviations

AHEI
Alternative Healthy Eating Index
A:P
Animal protein-to-potassium ratio
FFQ
Food frequency questionnaire
HPFS
Health Professionals’ Follow-up Study
MET
Metabolic equivalent task
NEAP
Net endogenous acid production
NHS
Nurses’ Health Study
NHS2
Nurses’ Health Study II
PRAL
Potential renal acid load

Introduction

Type 2 diabetes is an important cause of mortality and morbidity globally and its increasing prevalence is driven by the obesity epidemic and ageing populations [1]. A western dietary pattern, characterised by high intake of foods containing animal products and beverages containing sugar and low intake of fruit, vegetables and whole grains, is associated with an increased risk of type 2 diabetes [2]. This association can be partially explained by low intake of whole grains and poor dietary fat quality, which have been related to glucose homeostasis in both experimental [34] and observational studies [12]. However, potential mechanisms behind other components of a western dietary pattern, such as foods rich in animal protein, are not yet fully elucidated.
It has been hypothesised that a western-style diet may cause low-grade metabolic acidosis, possibly leading to metabolic disturbances [56]. Sulphur-containing amino acids (e.g. methionine and cysteine) are found in animal proteins, particularly in meat and fish, and are important determinants of dietary acid load due to sulfate generation after their oxidation [5]. Indeed, several studies have shown that the positive association between protein intake and type 2 diabetes is mainly driven by animal protein intake [78].
Physiological effects of high dietary acid load include increased urinary excretion of sulfate, phosphorus and chloride, increased elimination of calcium, intrarenal vasodilatation and increased glomerular filtration rate [59]. Hence, dietary acid load has been studied extensively in relation to kidney disease [1011], blood pressure [1213] and bone health [14]. Dietary acid load may also play a role in glucose homeostasis. Associations between markers of metabolic acidosis (i.e. low serum bicarbonate, higher anion gap and low urine pH) and insulin resistance have been reported in individuals with insulin resistance [15] and in a nested case–control study in type 2 diabetes [16]. In addition, a recent study found that dietary acid load was associated with insulin resistance in healthy Japanese workers [17]. The E3N-EPIC cohort study found diet-dependent acid load to be associated with an increased risk of type 2 diabetes [18]. However, this was not confirmed in a cohort of community-dwelling men in Sweden [19] and a Japanese cohort study only found an association between dietary acid load and type 2 diabetes risk in younger men [20]. These contrasting findings may be explained by differences in other dietary habits, sex, age and other population characteristics. Therefore, further research on this topic is needed to replicate and validate previous findings and facilitate dietary recommendations related to diet-dependent acid load for public health purposes. We aimed to evaluate the prospective association between dietary acid load and the incidence of type 2 diabetes in three cohorts of adults in the USA.

Methods

Study population

Data was used from three prospective cohort studies of health professionals in the USA: the Nurses’ Health Study (NHS), Nurses’ Health Study II (NHS2) and the Health Professionals’ Follow-up Study (HPFS).
The NHS was established in 1976 and included 121,700 registered nurses aged 30–55 years living in the 11 most populous states of the USA (California, Connecticut, Florida, Maryland, Massachusetts, Michigan, New Jersey, New York, Ohio, Pennsylvania and Texas). NHS2 was established in 1989 and included a younger population of nurses. The NHS2 population included 116,430 women between the ages of 25 and 42 years. The HPFS began in 1986 and included 51,529 men in health professions aged 40–75 years (29,683 dentists, 4185 pharmacists, 3745 optometrists, 2220 osteopath physicians, 1600 podiatrists and 10,098 veterinarians).
At baseline and every 2 years, participants in the NHS, NHS2 and HPFS completed questionnaires about diseases and health-related topics such as smoking, physical activity and medication use. In addition, food frequency questionnaires (FFQs) to assess dietary intake were administered at 4 year intervals. The follow-up rate was greater than 90% in all cohorts.
For the present analyses we excluded participants with prior diagnosis at baseline of cardiovascular disease (including coronary heart disease and stroke), cancer, type 1 or type 2 diabetes or history of gestational diabetes (in NHS cohorts only). Additionally, we excluded participants with implausible daily intakes of total energy (<2510 kJ or >14,644 kJ for women and <3347 kJ or >17,573 kJ for men) and participants with missing FFQ data. The final population of analyses consisted of 67,433 women in the NHS, 84,310 women in the NHS2 and 35,743 men in the HPFS.
The study protocols were approved by the institutional review board of Brigham and Women’s Hospital and Harvard T. H. Chan School of Public Health and all participants provided informed consent.

Dietary assessment and dietary acid load

Habitual dietary intake over the preceding year was assessed by validated FFQs described in detail previously [2122]. Dietary data were collected in 1984 for the NHS, 1986 for the HPFS and 1991 for the NHS2 and were updated every 4 years with similar FFQs.
For each food item, participants were asked how often on average they consumed a serving of that food. The frequency of a food item was recorded as number of times per day, week or month.
To calculate nutrient intake, the frequency of consumption of each food item was multiplied by the nutrient content of one serving and then summed across all food items. Nutrient values of foods were obtained from the US Department of Agriculture [23].
We calculated dietary acid load by using three previously defined algorithms: net endogenous acid production (NEAP) [24], potential renal acid load (PRAL) [25] and animal protein-to-potassium ratio (A:P) [26]:
NEAP(mEq/day)=(54.5×protein[g/day]/potassium[mEq/day])10.2
PRAL(mEq/day)=0.4888×protein [g/day]+0.0366×phosphorus [mg/day]--0.0205×potassium [mg/day]--0.0125×calcium[mg/day]--0.0263×magnesium [mg/day]
A:P=animal protein(g/day)/potassium (g/day)
The NEAP algorithm by Frassetto et al [24] estimates dietary acid load from an acid precursor (dietary protein) and an index of base precursors from organic anions (potassium) and has been previously validated in healthy men and women aged 17–73 years [24]. NEAP only includes protein and potassium as pH-altering nutrients from diet whereas the PRAL algorithm by Remer and Manz [25] estimates dietary acid load taking into account average intestinal absorption rates of ingested protein and additional minerals and has been validated against urine pH in healthy adults [25]. Since it is considered that animal protein constitutes a major determinant of dietary acid load [27], and NEAP and PRAL do not make a distinction between animal and vegetable protein, we also assessed the ratio of animal protein to potassium, as described previously [26].
To account for measurement error and to remove extraneous variation arising from total energy intake, all nutrients (including those for NEAP, PRAL and A:P) were adjusted for total energy intake by using the residual method [28].
The FFQs were validated against the diet records of 173 participants in the NHS in 1980 and 127 participants in the HPFS in 1986 [2122]. Correlation coefficients adjusted for energy and within-person variation varied from 0.52 (NHS) to 0.44 (HPFS) for total protein, from 0.53 (NHS) to 0.73 (HPFS) for potassium, from 0.51 (NHS) to 0.54 (HPFS) for calcium, from 0.53 (NHS) to 0.57 (HPFS) for phosphorus and 0.72 for magnesium when compared with 1 week food records [2122].

Assessment of type 2 diabetes

Every 2 years, participants in all three cohorts were asked whether they had any physician diagnosis of diabetes. Participants who reported physician-diagnosed diabetes were sent a supplementary questionnaire to obtain information about symptoms, diagnostic tests and diabetes drug use.
Participants were originally defined as having type 2 diabetes if they experienced one or more symptoms of polydipsia, polyuria, weight loss and hunger and met at least one of the following criteria [29]: increased blood glucose levels (fasting levels ≥7.8 mmol/l, random blood levels ≥11.1 mmol/l and/or 2 h blood glucose levels ≥11.1 mmol/l during OGTT); raised blood glucose levels on two different occasions in the absence of symptoms; or treatment with glucose-lowering drugs.
In June 1998 the diagnostic criteria of type 2 diabetes were changed and a fasting blood glucose level of 7 mmol/l, rather than 7.8 mmol/l, was used as the threshold for diagnosis [30]. The questionnaires for the diagnosis of type 2 diabetes have been previously validated in subsamples of the cohorts [3132]. These validation studies showed that 98% and 97% of the cases of self-reported type 2 diabetes were confirmed by medical records in the NHS [32] and the HPFS, respectively [3132].

Covariates

Updated information on anthropometric and lifestyle factors for type 2 diabetes, including body weight (height was ascertained at baseline), cigarette smoking, physical activity, family history of diabetes and history of kidney stones was collected in the 2-yearly questionnaires. Among participants in the NHS and NHS2, menopausal status and postmenopausal hormone use was ascertained by questionnaires; oral contraceptive use was ascertained by questionnaire in the NHS2 only. An approximation of moderate/vigorous physical activity levels was determined by multiplying the metabolic equivalent tasks (METs) measured in hours per week of each activity by the number of hours spent on the activity and taking the sum of these values (six METs or greater was defined as moderate/vigorous activity).
Based on the FFQ, an adherence score for the Alternative Healthy Eating Index (AHEI), an indicator of adherence to healthy eating behaviour, was derived as described in detail elsewhere [33], as well as indices of glycaemic load. Briefly, the AHEI included intake of the following foods and nutrients: vegetables, fruits, whole grains, sugar-sweetened beverages and fruit juice, nuts and legumes, red and processed meat, trans fat, long-chain n-3 fat, polyunsaturated fat and sodium. Alcohol intake was excluded from the AHEI and treated as an individual covariate. Each component was scored on a scale of 0–10 and the overall score ranged from 0–100, with a higher score representing a healthier diet. Principal component analysis was used for a posteriori-derived western-like dietary pattern scores as described in detail previously [34]. The overall glycaemic load was calculated by taking the sum of the following product: carbohydrates per food item × glycaemic index of food item × mean servings of food item per day [35].

Statistical analyses

To assess long-term dietary acid load exposure and to reduce within-person variation, we calculated the cumulative average of dietary acid load from baseline until type 2 diabetes diagnosis, death or end of follow-up, whichever came first. Missing dietary data on follow-up visits were replaced by the cumulative average of prior dietary assessments.
Participant characteristics by dietary acid load were standardised to the age distribution of the study population.
We used multivariable Cox proportional hazard models with time-dependent covariates to assess the relationship between quintiles of NEAP, PRAL and A:P and type 2 diabetes. Linear trends across the quintiles were examined by using the median values of each quintile of dietary acid load as continuous variable in the analyses.
To account for differences in age distribution, time-dependent Cox regression analyses were stratified by age in months (i.e. an age-adjusted model allowing for different baseline hazard functions for different age groups).
Results were further adjusted for total energy intake (quintiles), BMI (continuously), family history of type 2 diabetes (yes/no), menopausal status (premenopausal, postmenopausal without hormone use, postmenopausal with past hormone use, postmenopausal with current hormone use), history of hypertension and hypercholesterolaemia (yes/no), smoking status (never smoking, former smoker, current smoker [1–14 cigarettes, 15–24 cigarettes and ≥25 cigarettes/day]), alcohol intake (0, 0.1–4.9, 5.0–14.9, 15.0–19.9, 20.0–29.9 and ≥30 g/day) and moderate/vigorous physical activity (0, 0.01–1.0, 1.0–3.5, 3.5–6.0 and ≥6.0 h/week) (multivariate model 1). Additionally we adjusted the analyses for other dietary exposures associated with type 2 diabetes [12] such as glycaemic load (quintiles), AHEI (quintiles) and the western dietary pattern (quintiles) (multivariate model 2).
Sensitivity analyses were performed by stratification according to BMI status (<25 kg/m2, 25–29 kg/m2 and ≥30 kg/m2), age (<60 years and ≥60 years), smoking status (ever vs never smoker), presence of hypertension and history of kidney stones [34]. Effect modification by BMI, smoking status, hypertension and kidney stones was evaluated by using the likelihood ratio test. Furthermore, since it has recently been found that animal protein but not vegetable protein was associated with an increased risk of type 2 diabetes [8], we additionally adjusted the association between NEAP and type 2 diabetes for animal protein intake (quintiles). Last, to assess the potential influence of dietary changes due to diabetes-related symptoms before diagnosis, we used a 4 year lag between dietary acid load and type 2 diabetes in sensitivity analysis. Results from the three cohorts were pooled using random effects meta-analysis and are reported as HR and 95% CI. All analyses were performed with SAS 9.3 Unix (SAS Institute, Cary, NC, USA). A p value <0.05 was considered as statistically significant.

Results

Characteristics of the study population

Mean (SD) dietary NEAP, PRAL and A:P was, respectively, 43.8 (8.7) mEq/day, −3.1 (10.6) mEq/day and 18.4 (4.1) for the NHS, 49.9 (10.5) mEq/day, 6.4 (12.2) mEq/day and 20.3 (5.1) for the NHS2 and 47.7 (10.5) mEq/day, 5.7 (13.2) mEq/day and 19.5 (5.1) for the HPFS.
Participants with high dietary NEAP tended to consume more red meat and fish and to consume fruit, whole grains and sugar-containing beverages less often. In addition, participants with high dietary NEAP were likely to consume less alcohol and had an overall lower AHEI score and a higher western dietary pattern score (Table 1).
Table 1
Age-adjusted characteristics of the NHS and HPFS at median follow-up time
Characteristic
NHS (1994)a
NHS2 (1999)a
HPFS (1996)a
Quintile 1
(n = 11,449)
Quintile 5
(n = 14,974)
Quintile 1
(n = 18,030)
Quintile 5
(n = 13,878)
Quintile 1
(n = 6428)
Quintile 5
(n = 7472)
Age, yearsb
62.4 (6.8)
57.6 (6.8)
45.1 (4.5)
43.4 (4.7)
64.6 (9.4)
59.5 (8.4)
Premenopausal, %
10.2
11.0
74.1
70.5
Family history of diabetes, %
17.3
19.1
14.9
17.4
19.1
20.2
BMI, kg/m2
24.9 (4.4)
27.3 (5.6)
24.8 (5.0)
28.0 (6.9)
25.4 (3.5)
26.5 (3.9)
Hypertension, %
21.3
24.4
9.3
15.2
35.6
36.6
Hypercholesterolaemia, %
27.7
30.1
18.6
24.6
41.0
43.4
Moderate/vigorous intensity activity, h/week
2.6 (4.1)
1.5 (3.0)
3.0 (4.0)
1.7 (2.8)
5.0 (6.4)
3.1 (5.0)
Current smoker, %
14.2
13.0
10.1
8.5
4.4
7.1
Alcohol intake, g/day
5.8 (10.1)
4.5 (9.1)
5.0 (8.3)
2.7 (6.4)
11.8 (15.8)
9.7 (14.0)
AHEI score
50.6 (10.1)
45.0 (9.8)
50.1 (9.5)
41.3 (8.5)
53.0 (9.0)
44.0 (9.3)
Western dietary pattern score
0.4 (0.9)
0.6 (0.8)
−0.3 (0.7)
0.1 (0.8)
−0.4 (0.7)
0.2 (1.0)
Glycaemic load, units/day
113.5 (16.8)
95.6 (15.7)
134.5 (18.6)
114.2 (19.5)
145.6 (23.9)
117.5 (22.3)
Red meat intake, no. of servings/day
0.8 (0.5)
1.1 (0.6)
0.6 (0.4)
1.2 (0.6)
0.7 (0.5)
1.4 (0.9)
Sugar-containing beverage intake, no. of servings/day
1.5 (0.9)
0.9 (0.6)
1.4 (1.2)
0.9 (1.0)
1.5 (1.2)
1.0 (0.8)
Dairy intake, no. of servings/day
1.6 (0.9)
1.4 (0.8)
2.3 (1.3)
1.8 (1.1)
1.9 (1.3)
1.7 (1.2)
Vegetable intake, no. of servings/day
4.3 (1.7)
3.4 (1.2)
3.9 (2.0)
2.3 (1.1)
3.9 (2.0)
2.4 (1.2)
Nut intake, no. of servings/day
1.2 (0.8)
0.8 (0.6)
0.4 (0.4)
0.2 (0.2)
0.6 (0.5)
0.4 (0.4)
Fruit intake, no. of servings/day
1.7 (1.0)
0.8 (0.5)
1.8 (1.1)
0.7 (0.5)
2.6 (1.6)
1.0 (0.7)
Whole grain intake, g/day
15.4 (10.4)
11.1 (7.9)
28.3 (15.9)
18.2 (11.8)
32.3 (20.1)
20.3 (15.5)
Fish intake, no. of servings/day
0.2 (0.2)
0.3 (0.2)
0.2 (0.2)
0.2 (0.2)
0.3 (0.3)
0.4 (0.3)
Energy intake, kJ/day
7196 (2192)
6987 (2205)
7510 (2025)
7234 (2058)
8284 (2251)
2029 (1017)
Fat intake, % of energy
29.0 (4.9)
34.7 (4.6)
27.5 (5.0)
34.0 (5.1)
26.7 (5.5)
34.2 (5.3)
Saturated fat intake, % of energy
10.1 (2.2)
12.2 (2.1)
9.3 (2.1)
11.9 (2.2)
8.6 (2.3)
11.7 (2.4)
Polyunsaturated fat intake, % of energy
4.5 (1.3)
4.9 (1.2)
5.1 (1.2)
5.7 (1.2)
5.3 (1.4)
5.9 (1.2)
Carbohydrate intake, % of energy
54.5 (6.6)
44.8 (6.1)
56.4 (6.3)
45.8 (6.4)
55.6 (7.6)
43.5 (6.8)
Protein intake,% of energy
16.6 (2.3)
19.8 (2.9)
17.0 (2.5)
20.4 (3.1)
16.3 (2.4)
20.0 (3.1)
Animal protein intake, % of energy
11.1 (2.5)
15.1 (2.9)
11.0 (2.8)
15.7 (3.0)
10.4 (2.6)
15.3 (3.1)
Potassium intake, mg/day
3583 (930)
2673 (688)
3449 (941)
2449 (708)
3875 (1060)
2805 (802)
Dietary NEAP, mEq/day
32.1 (3.5)
55.7 (6.0)
36.8 (4.5)
66.7 (7.3)
33.8 (4.4)
62.5 (7.4)
Dietary PRAL, mEq/day
−17.4 (8.0)
8.7 (6.4)
−8.8 (9.1)
22.0 (7.3)
−12.9 (10.3)
21.3 (7.2)
Dietary A:P
13.2 (2.2)
23.5 (3.0)
14.3 (3.0)
27.8 (3.8)
13.2 (2.8)
26.1 (3.9)
Values are expressed as means (SD) or %, standardised to the age distribution of the study population and grouped according to quintiles of NEAP (quintile 1, low; quintile 5, high)
aYear represents median follow-up
bValue is not age-adjusted
During 1,709,638 person-years of follow-up, 7655 new cases with type 2 diabetes were ascertained in the NHS; during 1,513,932 person-years of follow-up, 4109 cases were ascertained in the NHS2 and during 801,561 person-years of follow-up 3541 cases were ascertained in HPFS.

Dietary acid load and type 2 diabetes

The association between indices of dietary acid load and type 2 diabetes is shown in Table 2. In the NHS and HPFS cohorts, dietary NEAP was associated with an increased risk of type 2 diabetes after adjustment for diabetes risk factors (HR [95% CI] for highest vs lowest quintile of NEAP was 1.28 [1.18, 1.38], ptrend <0.0001 for NHS, 1.30 [1.17, 1.44], ptrend <0.0001 for NHS2 and 1.32 [1.18, 1.47], ptrend <0.0001 for HPFS).
Table 2
Association of type 2 diabetes mellitus with NEAP, PRAL and A:P
Modela
Quintile 1
Quintile 2
Quintile 3
Quintile 4
Quintile 5
ptrend
NHS
  Dietary NEAP
    No. of cases/person-years
1051/342,428
1373/341,989
1598/341,556
1699/341,826
1934/341,838
 
    Age-adjusted
1.00
1.27 (1.18, 1.38)*
1.50 (1.39, 1.62)*
1.65 (1.53, 1.79)*
2.07 (1.91, 2.23)*
<0.0001
    Multivariate model 1
1.00
1.13 (1.05, 1.23)*
1.23 (1.13, 1.33)*
1.22 (1.12, 1.32)*
1.28 (1.18, 1.38)*
<0.0001
    Multivariate model 2
1.00
1.14 (1.05, 1.23)*
1.23 (1.13, 1.34)*
1.23 (1.13, 1.33)*
1.29 (1.19, 1.41)*
<0.0001
  Dietary PRAL
    No. of cases/person-years
1083/342,058
1382/341,855
1516/341,733
1708/341,945
1966/342,047
 
    Age-adjusted
1.00
1.27 (1.17, 1.37)*
1.42 (1.31, 1.54)*
1.67 (1.31, 1.54)*
2.15 (1.99, 2.32)*
<0.0001
    Multivariate model 1
1.00
1.13 (1.04, 1.22)*
1.17 (1.08, 1.27)*
1.21 (1.12, 1.31)*
1.26 (1.16, 1.36)*
<0.0001
    Multivariate model 2
1.00
1.10 (1.02, 1.20)*
1.14 (1.05, 1.24)*
1.18 (1.09, 1.28)*
1.23 (1.13, 1.33)*
<0.0001
  Dietary A:P
    No. of cases/person-years
1062/342,228
1332/341,912
1611/341,662
1738/341,775
1912/342,026
 
    Age-adjusted
1.00
1.23 (1.14, 1.34)*
1.51 (1.40, 1.63)*
1.70 (1.57, 1.83)*
2.08 (1.93, 2.25)*
<0.0001
    Multivariate model 1
1.00
1.10 (1.01, 1.19)*
1.21 (1.12, 1.31)*
1.22 (1.13, 1.32)*
1.26 (1.16, 1.36)*
<0.0001
    Multivariate model 2
1.00
1.11 (1.02, 1.20)*
1.23 (1.13, 1.34)*
1.26 (1.15, 1.37)*
1.31 (1.19, 1.43)*
<0.0001
NHS2
  Dietary NEAP
    No. of cases/person-years
595/303,044
752/302,999
819/302,865
904/302,564
1039//302,459
 
    Age-adjusted
1.00
1.32 (1.19, 1.47)*
1.57 (1.41, 1.75)*
1.98 (1.78, 2.19)*
2.82 (2.54, 3.12)*
<0.0001
    Multivariate model 1
1.00
1.03 (0.93, 1.15)
1.07 (0.96, 1.19)
1.14 (1.03, 1.27)*
1.30 (1.17, 1.44)*
<0.0001
    Multivariate model 2
1.00
1.01 (0.91, 1.13)
1.04 (0.93, 1.16)
1.10 (0.98, 1.23)
1.22 (1.09, 1.37)*
<0.0001
  Dietary PRAL
    No. of cases/person-years
684/303,268
806/303,105
820/302,908
861/302,517
952/302,395
 
    Age-adjusted
1.00
1.34 (1.21, 1.48)*
1.57 (1.42, 1.74)*
1.95 (1.76, 2.16)*
2.83 (2.55, 3.13)*
<0.0001
    Multivariate model 1
1.00
1.13 (1.02, 1.25)*
1.11 (1.00, 1.23)*
1.17 (1.05, 1.30)*
1.33 (1.20, 1.48)*
<0.0001
    Multivariate model 2
1.00
1.10 (0.99, 1.22)
1.07 (0.96, 1.19)
1.30 (1.00, 1.24)*
1.25 (1.12, 1.40)*
0.0065
  Dietary A:P
    No. of cases/person-years
563/302,924
749/303,056
839/302,969
896/302,618
1026/302,366
 
    Age-adjusted
1.00
1.37 (1.22, 1.52)*
1.69 (1.51, 1.88)*
2.06 (1.85, 2.29)*
3.07 (2.77, 3.41)*
<0.0001
    Multivariate model 1
1.00
1.08 (0.97, 1.21)
1.09 (0.97, 1.21)
1.18 (1.06, 1.32)*
1.35 (1.21, 1.50)*
<0.0001
    Multivariate model 2
1.00
1.07 (0.95, 1.20)
1.06 (0.95, 1.19)
1.15 (1.02, 1.29)*
1.30 (1.15, 1.47)*
<0.0001
HPFS
  Dietary NEAP
    No. of cases/person-years
585/160,351
636/160,660
701/160,617
770/160,289
849/159,654
 
    Age-adjusted
1.00
1.07 (0.95, 1.20)
1.21 (1.08, 1.35)*
1.41 (1.27, 1.57)*
1.74 (1.56, 1.94)*
<0.0001
    Multivariate model 1
1.00
1.00 (0.89, 1.12)
1.08 (0.96, 1.20)
1.16 (1.04, 1.29)*
1.32 (1.18, 1.47)*
<0.0001
    Multivariate model 2
1.00
0.92 (0.82, 1.03)
0.95 (0.84, 1.06)
0.99 (0.88, 1.11)
1.09 (0.96, 1.23)
0. 0370
  Dietary PRAL
    No. of cases/person-years
585/160,199
688/160,651
694/160,536
752/160,402
825/159,773
 
    Age-adjusted
1.00
1.17 (1.04, 1.30)*
1.21 (1.08, 1.35)*
1.39 (1.25, 1.55)*
1.71 (1.54, 1.91)*
<0.0001
    Multivariate model 1
1.00
1.10 (0.98, 1.23)
1.09 (0.97, 1.22)
1.17 (1.04, 1.30)*
1.29 (1.16, 1.44)*
<0.0001
    Multivariate model 2
1.00
1.01 (0.90, 1.13)
0.95 (0.85, 1.07)
0.99 (0.88, 1.12)
1.07 (0.94, 1.20)
0.3583
  Dietary A:P
    No. of cases/person-years
553/160,474
633/160,629
662/160,617
816/160,334
877/159,507
 
    Age-adjusted
1.00
1.1 (1.02, 1.28)*
1.22 (1.09, 1.37)*
1.60 (1.44, 1.79)*
1.94 (1.74, 2.16)*
<0.0001
    Multivariate model 1
1.00
1.04 (0.92, 1.16)
1.03 (0.92, 1.15)
1.26 (1.12, 1.40)*
1.39 (1.25, 1.55)*
<0.0001
    Multivariate model 2
1.00
0.94 (0.84, 1.06)
0.90 (0.80, 1.01)
1.07 (0.94, 1.20)
1.15 (1.01, 1.31)*
0.0016
Pooled cohorts
  Dietary NEAP
    Age-adjusted
1.00
1.22 (1.09, 1.37)*
1.42 (1.23, 1.64)*
1.67 (1.40, 1.98)*
2.16 (1.68, 2.79)*
<0.0001
    Multivariate model 1
1.00
1.06 (0.98, 1.15)
1.13 (1.03, 1.24)*
1.18 (1.11, 1.25)*
1.29 (1.22, 1.37)*
<0.0001
    Multivariate model 2
1.00
1.03 (0.91, 1.16)
1.07 (0.91, 1.26)
1.10 (0.97, 1.26)
1.21 (1.09, 1.33)*
<0.0001
  Dietary PRAL
    Age-adjusted
1.00
1.26 (1.17, 1.35)*
1.39 (1.22, 1.60)*
1.66 (1.39, 1.97)*
2.19 (1.64, 2.94)*
<0.0001
    Multivariate model 1
1.00
1.12 (1.06, 1.18)*
1.13 (1.07, 1.20)*
1.19 (1.13, 1.26)*
1.29 (1.22, 1.36)*
<0.0001
    Multivariate model 2
1.00
1.08 (1.02, 1.14)*
1.06 (0.96, 1.18)
1.10 (1.06, 1.22)*
1.19 (1.08, 1.30)*
0.0093
  Dietary A:P
    Age-adjusted
1.00
1.24 (1.13, 1.36)*
1.46 (1.24, 1.72)*
1.77 (1.55, 2.04)*
2.31 (1.78, 3.02)*
<0.0001
    Multivariate model 1
1.00
1.08 (1.02, 1.14)*
1.11 (1.00, 1.23)*
1.22 (1.15, 1.29)*
1.32 (1.24, 1.40)*
<0.0001
    Multivariate model 2
1.00
1.04 (0.95, 1.15)
1.06 (0.88, 1.27)
1.16 (1.06, 1.28) a
1.26 (1.17, 1.36)*
<0.0001
Data are shown as HR (95% CI). The results were pooled using random effect meta-analysis
aMultivariate model 1: additionally adjusted for total energy intake (quintiles), BMI (continuously), family history of diabetes (yes/no), menopausal status (premenopausal or postmenopausal, never, past or current menopausal hormone use), presence of hypertension and hypercholesterolaemia (yes/no), smoking status (never smoker, former smoker, current smoker: 1–14, 15–24 or ≥25 cigarettes/day), alcohol intake (0, 0.1–4.9, 5.0–14.9, 15.0–19.9, 20.0–29.9 and ≥30 g/day), moderate/vigorous intensity activity (0, 0.01–1.0, 1.0–3.5, 3.5–6.0 and ≥6 h/week). Multivariate model 2: additionally adjusted for glycaemic load (quintiles), AHEI index (quintiles; including the following components: sugar-containing beverages, fruit, vegetables, nuts, red meat, whole grains, EPA, DHA, other PUFAs, trans fat, and sodium) and the western dietary pattern (quintiles)
*p < 0.05
Similar results were found for PRAL. Higher PRAL was associated with an increased risk of type 2 diabetes after adjustment for diabetes risk factors (HR [95% CI] for highest vs lowest quintile was 1.26 [1.16, 1.36], ptrend <0.0001 for NHS, 1.33 [1.20, 1.48], ptrend <0.0001 for NHS2 and 1.29 [1.16, 1.44], ptrend <0.0001 for HPFS).
In addition, a higher A:P was associated with an increased risk of type 2 diabetes after adjustment for diabetes risk factors (HR [95% CI] for highest vs lowest quintile was 1.26 [1.16, 1.36], ptrend <0.0001 for NHS, 1.35 [1.21, 1.50], ptrend <0.0001 for NHS2 and 1.39 [1.25, 1.55], ptrend <0.0001 for HPFS). Additional adjustment for dietary risk factors for type 2 diabetes attenuated some of the associations but they remained statistically significant except for PRAL in HPFS (HR [95% CI] for highest vs lowest quintile of NEAP was 1.29 [1.19, 1.41], ptrend<0.0001 for NHS, 1.22 [1.09, 1.37], ptrend <0.0001 for NHS2 and 1.09 [0.96, 1.23], ptrend = 0.0370 for HPFS and HR [95% CI] for PRAL was 1.23 [1.13, 1.33], ptrend <0.0001 for NHS, 1.25 [1.12, 1.40], ptrend <0.0001 for NHS2 and 1.07 [0.94, 1.20], ptrend = 0.358 for HPFS; Table 2). The pooled estimate of HR (95% CI) for type 2 diabetes risk according to dietary NEAP, PRAL and A:P was 1.21 (1.09, 1.33), ptrend <0.0001, 1.19 (1.08, 1.30) ptrend = 0.009 and 1.26 (1.17, 1.36), ptrend <0.0001, respectively, for highest vs lowest quintile after adjustment for diabetes and dietary risk factors.

Additional analyses

After stratification by BMI, the associations between type 2 diabetes and NEAP, PRAL and A:P were slightly stronger in participants with BMI <25 kg/m2 than in those who were overweight or obese but the difference did not reach statistical significance across strata (HR [95% CI] for highest vs lowest quintile of NEAP was 1.33 [1.13, 1.57] for participants with BMI <25 kg/m2, 1.24 [1.12, 1.38] for overweight participants and 1.14 [0.96, 1.35] for obese participants, pinteraction = 0.4237; electronic supplementary material [ESM] Table 1). Also, stratification by smoking status and hypertension yielded comparable results (HR [95% CI] for highest vs lowest quintile of NEAP was 1.23 [1.10, 1.39] for never smokers and 1.20 [1.08, 1.33] for ever smokers, pinteraction = 0.3272, and 1.25 [1.15, 1.36] for participants without hypertension and 1.17 [0.99, 1.38] for participants with hypertension, pinteraction = 0.8875; ESM Table 1). In addition, no differential associations were found after stratification by history of kidney stones and age (HR [95% CI] for highest vs lowest quintile of NEAP was 0.99 [0.70, 1.25] for those with kidney stones and 1.11 [0.99, 1.29] for those without kidney stones, pinteraction = 0.3089, and 1.19 [1.02, 1.38] for participants below 60 years of age and 1.13 [0.98, 1.31] for those aged 60 years and above; ESM Table 1).
Results did not differ in sensitivity analyses using a 4 year lag of exposure to dietary acid load (HR [95% CI] for highest vs lowest quintile of NEAP was 1.30 [1.18, 1.41], ptrend <0.0001 for NHS, 1.24 [1.10, 1.39], ptrend <0.0001 for NHS2 and 1.13 [0.99, 1.28], ptrend = 0.0026 for HPFS). Additional adjustment for animal protein weakened the association between dietary NEAP and type 2 diabetes but results remained significant among women in NHS and NHS2 (HR [95% CI] for highest vs lowest quintile was 1.23 [1.12, 1.34], ptrend <0.0001 for NHS, 1.14 [1.01, 1.29], ptrend = 0.0103 for NHS2 and 0.97 [0.85, 1.11], ptrend = 0.8328 for HPFS).

Discussion

In three prospective cohorts of men and women in the USA, we found that dietary acid load was associated with an increased risk of type 2 diabetes. In addition, differences in BMI and other dietary risk factors for type 2 diabetes, such as glycaemic load, the AHEI index and an a priori-defined western dietary pattern, did not fully explain the observed associations.

Comparison with other studies

The results of this study are in line with results from the E3N-EPIC cohort study wherein dietary acid load was found to be associated with the risk of type 2 diabetes in a cohort of French women aged ∼50 years [18]. Another study revealed that dietary NEAP and PRAL were associated with insulin resistance and beta cell function in Japanese men and women aged 19–69 years, with the latter association being mainly present in normal-weight individuals [17]. Recent findings from The Japan Public Health Center-based Prospective Study showed that PRAL but not NEAP was associated with the risk of type 2 diabetes only in Japanese men [20]. Moreover, the E3N-EPIC study found that the association between dietary acid load and type 2 diabetes was stronger in women with BMI <25 kg/m2, which is in line with our findings. In contrast, another study in community-dwelling older men (aged 70–71 years) did not confirm the aforementioned associations [19]. It has been proposed that these contrasting results may be due to differences in age distribution since the results from Akter et al showed that the association between dietary acid load and type 2 diabetes was mainly observed in younger individuals [20]. However, our stratified analyses did not reveal any significant effect modification by age on the association between dietary acid load and type 2 diabetes. We found that the association between diet acid load and diabetes risk was slightly stronger in women from the NHS than in men from the HPFS, suggesting that diet-dependent acid load may affect the development of type 2 diabetes in a sex-specific manner (e.g. due to differences in sex hormones, such as oestrogen, which may affect acid–base balance) [36].

Potential mechanisms

Dietary acid load was determined by validated indices based on protein and potassium intake. We have previously shown that animal protein but not vegetable protein is associated with increased type 2 diabetes [8]. In this analysis, we found that the association between dietary NEAP and type 2 diabetes was explained by animal protein intake to some extent. Sulphur-containing amino acids, such as methionine and cysteine, are found in animal protein and are main determinants of acid load as sulfate is generated after their oxidation [37]. Some studies of animal protein in relation to type 2 diabetes have shown conflicting results [38]. Nevertheless, red meat intake, a major source of sulphur-containing amino acids, has been consistently associated with an increased risk of insulin resistance, the metabolic syndrome and type 2 diabetes [39].
The main food sources of potassium are fruit and vegetables, which also provide other base cations (e.g. magnesium) [23]. A previous review showed that green leafy vegetable intake was associated with a reduced risk of type 2 diabetes but inconsistent results were found for intake of other vegetables and fruit [40]. Potassium is involved in acid–base balance by assisting electro-neutrality through exchange across cellular membranes for hydrogen ions [4142]. As a result, potassium-rich foods have been shown to be more alkalising than animal product-based foods [43].
The hypothesis that low-grade-metabolic acidosis may play a role in the aetiology of type 2 diabetes has been suggested by others. Some experimental studies have shown that reduction of the extracellular pH decreases beta cell response [44], reduces insulin secretion [44] and increases cortisol production [45], which in turn may affect the development of type 2 diabetes [46]. In addition, several studies have confirmed that markers of low-grade metabolic acidosis, such as lower plasma bicarbonate [61516], higher anion gap [15], lower urine pH [47] and high levels of plasma lactate [48], are linked to insulin resistance, suggesting that low-grade metabolic acidosis could be involved in the aetiology of type 2 diabetes.

Methodological considerations

Strengths of this study include the prospective design with long follow-up, the large sample size, high rates of follow-up and repeated dietary measurements during follow-up. In addition, the wide range of data on confounding variables and the homogeneity of the study population help to reduce residual confounding. However, several limitations need to be considered. First, dietary data was based on self-reported FFQs. While these are susceptible to measurement error, we used energy-adjusted nutrient intake for the calculation of dietary acid load as well as repeated measurements of diet, which can reduce the magnitude of measurement error [49]. Nonetheless, phosphorus intake for the calculation of PRAL did not include the assessment of phosphorus-containing food additives. Although the results for PRAL were similar to those for NEAP and A:P, they should be interpreted with caution. Second, we did not have data on kidney function available in the cohorts. Since kidney function is an important determinant of acid–base balance [27], it may be speculated that the relationship between dietary acid load and type 2 diabetes is more pronounced in participants with impaired kidney function due to altered haemodynamic adaptation to high acid load [10]. Hence, future studies on dietary acid load that include detailed urinary markers of acid–base balance (e.g. ammonium) and kidney function are needed to clarify the observed associations. Third, this study is of an observational nature. Although we adjusted for many confounders, residual confounding cannot be fully excluded. Last, the study population consisted primarily of white health professionals. This may limit the generalisability of our study results but could strengthen internal validity since confounding by ethnicity and socioeconomic status is greatly reduced.

Public health implications

It has recently been demonstrated in healthy individuals that both the capillary and urine pH can be modified by a diet high or low in protein combined with a high or low intake of fruit and vegetables [43]. Therefore, our findings may have important public health implications for the future. If future studies confirm that food-induced improvement in acid–base balance is accompanied by improved insulin sensitivity, this may facilitate development of dietary guidelines, such as more specific recommendations on the ratio of potassium to protein in diets.

Conclusions

Findings from this prospective study in men and women suggest that a higher dietary acid load is associated with an increased risk of type 2 diabetes. These results are not fully explained by BMI or diet-related risk factors associated with type 2 diabetes, such as glycaemic index and overall healthy or western dietary patterns. Randomised controlled trials are needed to clarify whether specific dietary interventions to reduce dietary acid load (e.g. a diet high in fruit and vegetables and low in animal protein) could improve glucose homeostasis and reduce the risk of type 2 diabetes.