To rule out potential false positive results from multiple tests of different modulators, all significant modulating variables were then included one omnibus regression model to evaluate the collective and individual residual effects of these multiple factors. We examined the effects of several potential modulating variables, to determine how they independently influenced P300 amplitude and whether they contributed to observed site differences. Each variable was added as a single additional factor to the original multi-factorial model. MMSE score [F=4.79. p<0.05] was positively associated with P300 amplitude independent of diagnosis, but had no impact on observed site differences. In separate Other measures of cognitive and functional status – GAF score and education – were not significant predictors of P300. Smoking, however, was found to be a robust modulator of P300 amplitude [F1,1195)=10.34, p<0.01]. Although the interaction between smoking and diagnosis was not significant [F=2.44, p=0.12], separate within-group analyses of smokers and nonsmokers revealed a significant patient control difference only among nonsmokers [F=29.69, p<0.00001]. Smoking differentially reduced P300 amplitude in healthy control subjects while having little effect in patients , which eliminated all diagnostic differences [F=1.09, p=0.30]. However, site differences remained robust even after controlling for smoking status. It should be noted, though,greenhouse rolling racks that only 70 control subjects were classified as smokers, compared to 50% of patients. The observed site differences appeared to primarily reflect racial stratification differences.

Inclusion of race as an additional predictor produced a significant race effect [F=16.29, p<0.000001], which eliminated the site effect while leaving the effects of both diagnosis [F=20.64, p<0.00001] and age [F=15.56, p<0.0001] intact. P300 amplitude was lower, overall, in the African American sample than in either the Caucasian or “Other/Mixed” racial groupings. There was a clear trend towards an interaction of race × diagnosis, but it did not reach statistical significance [F=2.85, p=.06]. In separate analyses, significant patient-control differences were observed within each racial subgroup, although the effect size was noticeably smaller within the African American sample . The differential impact of race on the association between schizophrenia and P300 was manifested primarily as an amplitude reduction among African American controls, rather than patients. Further consideration of potential modulating variables revealed that this apparent racial difference was due, in part, to the differential impact across the racial groupings of prior substance use disorders. When the sample was restricted to subjects with no history of substance use, the interaction of race × diagnosis was insignificant [F=1.56, p=0.21] and the magnitude of the patient-control difference was similar across racial categories . However, among those with a past history of substance abuse or dependence, there was a significant race × diagnosis interaction [F=6.77, p<0.001]. As illustrated , P300 responses of otherwise healthy African American controls were indistinguishable from those of African American schizophrenia patients. Comparable attenuating effects of past substance use were not observed for the Caucasian or Other/Mixed control samples. . The only significant interaction was diagnosis × substance use.

The 3-way interaction of diagnosis × African American race × substance use just missed significance threshold. The principal aim of this analysis was to determine the feasibility of acquiring comparable P300 data across multiple testing sites, both with and without specific ERP expertise, and to examine the clinical and socio-demographic factors that modulated the measurements across sites. Comparability across sites is a necessary predeterminant of the measure’s utility as an endophenotypic biomarker. To that end, the results are both very encouraging and somewhat cautionary. Across sites, 92% of subjects yielded technically acceptable EEG recordings with identifiable auditory evoked potential wave forms. Additional data loss, beyond this, resulted from 1)our failure to monitor subjects’ behavioral responses online in real time , and 2)our conservative strategy of rejecting any data lacking a reliable visibly identifiable P300 component. Many studies use an automated algorithm to measure P300 amplitude regardless of waveform appearance. Such an approach would have increased our final data yield from 74% to 85%. Given this overall yield, the fact that data quality did not differ between sites with or without prior electrophysiology experience, and the fact that the schizophrenia P300 deficit was replicated at each site, this study clearly demonstrates the feasibility of implementing large-scale ERP studies across diverse settings. The overall case-control effect size that we observed, 0.62, was somewhat lower than that reported in meta-analyses . Since the patient sample was older than the control sample and age significantly affected P300, the patient control difference was attenuated somewhat by inclusion of age as a covariate.

The effect size was almost certainly also lowered by our conservative data strategy, which likely excluded a number of subjects – primarily patients – with negligible but real P300 responses. This moderately large effect is, nevertheless, well within the expected distribution of published studies. Although we observed a significant difference across test sites, this did not reflect differences in data quality, methodology, or experimental rigor. Rather it reflected differences in the stratification of the samples across sites, as this relates to clinical and socio-demographic confounds or modifiers. In patients, site differences were entirely explained by differences in the level of positive symptomatology. Although the P300 deficit is traditionally thought of as being immune to changes in patients’ clinical status , it should probably be considered as more of a relatively stable deficit. It clearly does not normalize with treatment, even when symptoms dramatically improve. However, it still exhibits modulation over time in association with positive symptoms . Indeed, it is this ability to reflect increasing positive symptomatology that underlies the emerging utility of P300 as a predictive biomarker for imminent prodromal conversion to psychosis . Except for MMSE and UPSA-B, global indices of cognitive ability and real-world functional capacity, no other clinical measures were associated with P300, indicating that the association with positive symptoms is relatively specific. Since these patients were all clinically stable outpatients on stable medication regiments, differences in positive symptomatology presumably reflected relatively stable trait-like differences on this dimension of illness severity. P300 may therefore be an endophenotype that is especially informative regarding the genetic basis of positive symptoms. The associations with MMSE and UPSA-B highlight the utility of the P300 as a sensitive physiological index of differences in brain function, even within a relatively homogeneous clinical sample. The magnitude of the P300 response has long been considered a broad indicator of “cognitive fitness” and, more specifically, of the ability to appropriately process and respond to task-salient environmental inputs – i.e., to correctly detect a signal within noise. It is thought to require intact attentional and working memory capacities ,vertical grow and to reflect complex neural processes of temporal and spatial integration across multiple brain regions . It is not surprising, therefore, that the P300 would correlate with other measures of cognitive and functional capacity. A similar association between P300 amplitude and MMSE has been reported previously in chronic Alzheimer’s disease patients and, acutely, in uremic patients undergoing dialysis , where the two measures showed a correlated improvement, as well, following treatment. There have been no prior studies reporting a relationship between P300 and specific measures of functional capacity, including UPSA-B, either in schizophrenia patients or other clinical samples. However, this association is entirely consistent with the relationship between P300 and cognition.

Prior studies examining the relationship between neurocognitive and functional deficits have routinely found that cognitive ability, specifically working memory, is the strongest predictor of schizophrenia patients’ real-world functional capacity . Indeed, in our own data, we observed a similar robust correlation between MMSE and UPSA-B. These associations support the utility of P300 amplitude as a potential biomarker for predictive risk and treatment studies. However, they also emphasize the relatively non specific nature of the measure. This was evident, as well, in the control sample data. In these otherwise healthy subjects, P300 amplitude was affected by smoking, race and, as one mediator of the race effect, prior history of substance abuse or dependence. Previous studies have shown that nicotine reduces P300 , yet – despite the well-known propensity of schizophrenia patients to smoke – there has been virtually no consideration of the effect of smoking on the auditory P300 in patients. We observed no parallel effect of nicotine in patients, presumably because their ERPs were already suppressed. This is consistent with a recent small study of healthy subjects administered intravenous ketamine. Ketamine induced schizophrenia-like symptoms and attenuated the auditory target P300 response, but this was unaffected by co administration of nicotine vs. placebo . Similarly, reduced P300 has been associated with the use of stimulants , opioids and cannabis . Yet, again, we saw no effect of prior substance use on P300 in the schizophrenia patients. This mirrors what was recently reported in a study examining the effects of cannabis in prodromal subjects considered to be at ultra-high risk for developing psychosis. In this sample, those with a history of cannabis use were indistinguishable from those without. However, among the otherwise-healthy controls, those who used cannabis had reduced P300 responses that were indistinguishable from those of the prodromal sample . The impact of substance abuse on the African-American sample may reflect differences in the specific character and/or quantity of substance use within the different racial groupings, which are not captured by a simple dichotomous categorization. Similarly, the residual effects of race, independent of past substance use, could reflect the impact of other psychosocial stressors in the different racial communities. Unfortunately, we have no objective measures of either of stressful life events or physiological markers of stress to test this hypothesis. The fact that modulating factors such as nicotine and substance abuse can differential affect controls, but not patients, raises an important cautionary note about how to interpret study results, potential false negative findings, and what constitutes the best comparison sample for genetic or biomarker studies. A common recommended strategy is to recruit control subjects who are similar to the clinical sample on various modulating factors and co-morbid conditions. The results of this study would seem to temper that recommendation, at least for P300. It suggests that, in matching the samples, individual and group differences may be attenuated for reasons other than psychosis. Consequently, genetic associations with the endophenotype may be obscured and the ability of the measure to predict transition to psychosis may be weakened. This is an issue that clearly requires careful consideration in future analyses. However, the broad utility of P300 as a robust marker for large multi-site studies is confirmed, along with important associations with both positive symptoms and decreased cognitive and functional capacity. The prevalence of type 2 diabetes mellitus is increasing, and it is projected that in the USA alone, type 2 DM will increase to 48.3 million by 2050.In addition to defects in pancreatic b-cell function and insulin sensitivity, systemic inflammation is thought to be involved in its pathogenesis.Marijuana is the most commonly used illicit drug in the USA and is currently used by 14.4 million Americans.The major psychoactive CB is delta 9-tetrahydrocannabinol whose effect is mediated through the CB1 and the CB2 sub-types of CB receptors found in the brain and lymphoid tissues.The endocannabinoids, a group of neuromodulatory lipids also bind to these receptors.Cannabis, THC and other CBs have been shown to have both beneficial and detrimental effects.Marijuana users have higher caloric intake while eating less nutrient-rich foods,yet have similar or slightly lower body mass index than non users. We hypothesised that the prevalence of DM would be reduced in marijuana users due to the presence of one or more CBs because of their immunomodulatory and anti-inflammatory properties.We assessed the association between DM and marijuana use among adults aged 20e59 years in a national sample of the general population.Data on marijuana use were collected by self-report. Non-marijuana users included never users and those who reported ever having used marijuana, but who had not used marijuana in the past month . We classified participants who reported using marijuana in the past month by frequency of use as either light current users as previously described.The definition of marijuana for purposes of this survey includes ‘hash,’ ‘pot’ or ‘grass’ or any other references to the Cannabis plant.