Brain-derived neurotrophic factor (BDNF) regulates the survival and growth of neurons, and influences synaptic efficiency and plasticity. Several studies report reduced peripheral (blood) levels of BDNF in schizophrenia, but findings are inconsistent. We undertook the first systematic review with meta-analysis of studies examining blood BDNF levels in schizophrenia compared with healthy controls, and examined potential effects of age, gender and medication. Included are individual studies of BDNF blood (serum or plasma) levels in schizophrenia (including schizoaffective disorder, or first episode psychosis), compared with age-matched healthy controls, obtained by electronic Medline and Embase searches, and hand searching. The decision to include or exclude studies, data extraction and quality assessment were completed by two independent reviewers. The initial search revealed 378 records, of which 342 were excluded on reading the Abstract, because they did not examine BDNF blood levels in schizophrenia compared with healthy controls. Of 36 papers screened in full, 17 were eligible for inclusion, but one was subsequently removed as an outlier. The remaining 16 studies provided moderate quality evidence of reduced blood BDNF levels in schizophrenia (Hedges g=−0.458, 95% confidence interval=−0.770 to −0.146, P<0.004, random effects model). Subgroup analyses reveal reduced BDNF in both drug-naïve and medicated patients, and in males and females with schizophrenia. Meta-regressions showed an association between reduced BDNF in schizophrenia and increasing age, but no effects of medication dosage. Overall, blood levels of BDNF are reduced in medicated and drug-naïve patients with schizophrenia; this evidence is of moderate quality, that is, precise but with considerable, unexplained heterogeneity across study results.
Brain-derived neurotrophic factor (BDNF) is a neurotrophin that regulates neuronal survival, differentiation and growth during brain development, with important effects on neurogenesis and neuroplasticity.1 Activity-dependent effects of BDNF on neuronal transmission in the hippocampus, cortex, cerebellum and basal forebrain are important for learning and memory processes in the mature brain.2, 3
Neurodevelopmental models of schizophrenia implicate reduced BDNF in the central nervous system,4 and specifically suggest that reduced BDNF may affect synaptic efficiency and connectivity in schizophrenia that is believed to underlie core behavioural signs and symptoms of the disease.5, 6 This is consistent with findings from animal studies showing that BDNF controls the development and activity of neurotransmitter systems implicated in psychotic disorders.7, 8
Several studies report altered BDNF mRNA and protein in prefrontal cortical regions of post-mortem brain tissue9, 10, 11, 12, 13, 14 in people with schizophrenia, with variation in levels reported across different brain regions. It is likely that peripheral BDNF blood levels are derived from central nervous system sources, and thus provide a less invasive source for sampling BDNF levels to index the brain's neurotrophic potential in living people.15 However, there is no widespread agreement on the degree of peripheral BDNF reduction in schizophrenia, as measured in blood serum or plasma. The majority of studies report reduced peripheral (blood) BDNF levels in schizophrenia,15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28 whereas two studies report higher levels of BDNF concentration in serum29 or serum protein30 in schizophrenia compared with healthy controls. There are also studies that report no significant difference in BDNF plasma levels in non-medicated schizophrenia patients compared with controls, with increased BDNF levels observed in these same subjects following antipsychotic treatment.21 Some studies suggest that peripheral BNDF levels increase in association with antipsychotic medication exposure or responsivity to specific agents,17, 21 but conflicting evidence of no change in BDNF following antipsychotic treatment has also been reported.23 Notably, the majority of existing studies in schizophrenia have been conducted in chronic, medicated samples with only recent studies reporting specific investigation of first episode (drug-naïve) patients. Noting these inconsistencies, we aimed to systematically review studies examining peripheral BDNF in schizophrenia, and quantitatively assess by meta-analysis the robustness of previously reported findings of reduced blood BDNF levels in schizophrenia compared with age-matched healthy controls.
This study provides the first meta-analysis of blood BDNF levels in schizophrenia. This analysis has implications for the role of reduced BDNF as a potential biomarker for psychotic disorders, and may also provide clues about the relationship between brain and blood BDNF levels in humans, given robust findings of reduced brain BDNF in schizophrenia.9, 10, 11, 12, 13 In addition, it is of interest as to whether blood BDNF levels decrease with age at a greater rate in schizophrenia, in line with findings of age-related decline in brain BDNF levels in schizophrenia10, 12 and noted decreases in blood BDNF levels in adulthood in the general population.31 Finally, it remains unclear as to whether the observed medication dosage effects on brain BDNF levels4 apply also to blood BDNF levels in schizophrenia; as noted above, individual study findings on the relationship between antipsychotic medication and blood BDNF levels are currently mixed. We therefore used meta-regression to examine the influence of age and medication dosage on BDNF reduction in schizophrenia.
We hypothesized that (1) BDNF levels would be reduced in patients compared with age-matched controls when all study data are pooled; (1a) that the observed reduction in blood BDNF levels in schizophrenia patients would increase significantly with age, relative to normal rates of age-related decline; and, (1b) that antipsychotic dosage may be associated with the observed reduction in blood BDNF levels in schizophrenia. Using subgroup analyses we grouped patients by gender and medication exposure (drug-naïve and medicated patients) to examine further hypotheses that (2) the reduction in blood BDNF levels would be present early in the disease course, in both first episode, drug-naive patients as well as more chronic medicated patients and (3) that males with schizophrenia would show a greater reduction in peripheral BDNF levels than do females with schizophrenia, when compared with controls, due to the clinical observation of male cases displaying more severe symptoms in schizophrenia.
Included in the meta-analysis are studies that report a gold standard (enzyme-linked immunosorbent assay) measurement of BDNF levels in serum or plasma in patients with a diagnosis of schizophrenia, schizoaffective disorder or first episode schizophrenia/psychosis and in age-matched, healthy controls. We excluded studies reporting data for patients with comorbid substance abuse or neurological disorders. The decision to include or exclude studies was taken independently by two of the authors (MG and SM) with any disagreements settled by discussion.
Studies were identified in January 2010 by searching Medline and Embase (the Embase search revealed duplicate records to the Medline search and were therefore removed at the outset). The search terms were: exp schizophrenia/, schizophreni$.tw, exp psychotic disorders/, schizo$.tw, brain-derived neurotrophic factor.mp, brain-derived neurotrophic factor.tw, BDNF.tw, serum brain-derived neurotrophic factor.tw, plasma brain-derived neurotrophic factor.tw, neurotrophic factor.tw and neurotroph$.tw. Hand searching reference lists of included reviews were also conducted. Further, authors of included studies were contacted to enquire if they knew of any relevant unpublished studies and to request raw data that may not have been included in their publication (that is, subsequently collected data, gender breakdown of BDNF levels, antipsychotic type and dosage, antidepressant use and any clarification required of their published data).
Study quality was assessed using the STROBE checklist, which outlines a preferred way to report observational studies (http://www.strobe-statement.org). Studies were assigned a low, medium or high possibility of reporting bias depending on how many items were checked. For instance, a low possibility of reporting bias would be assigned to studies checking over 66% of items, a medium possibility between 33 and 66% and a high possibility would be assigned to studies checking less than 33%.
Data quality was assessed using the GRADE approach,32, 33 in which evidence such as that gained from randomized controlled trials is assumed to be of high quality and may be downgraded to moderate, low or very low (where further research is needed), if review or study quality are limited, if samples are small, if there is inconsistency in results (significant heterogeneity), if there are indirect comparisons or populations or if data are imprecise (wide confidence intervals). Conversely, evidence such as that gained from observational studies is assumed to be of low quality, and may be upgraded if sample sizes are large or if results are reasonably consistent, precise or direct. Quality assessments and data extraction have been completed independently by three of the authors (MG, SM and AS) who were not masked to study authors. The following variables were extracted: (1) mean and standard deviation of the BDNF levels for each participant group (and for patient-type and gender subgroups, where provided); (2) demographic, clinical and treatment characteristics (for example, number of patients, age, gender, previous exposure to antipsychotic medications, type of pharmacological treatment), (3) characteristics of measurement (pg ml−1 or ng ml−1 of plasma or serum; enzyme-linked immunosorbent assay kit used).
We used Comprehensive Meta Analysis (CMA V2) software34 to conduct the meta-analysis. We report standardized weighted mean differences (SMDs), represented as Hedges g, as this provides a more conservative measure for meta-analyses of many studies with a small sample size. SMD has the benefit of allowing effect sizes to be directly compared across different measurement scales (that is, ng ml−1 vs pg ml−1), because study mean differences are first standardized (that is, divided by the pooled standard deviation) before weighting is applied. A SMD of 0.2 represents a small treatment effect, 0.5 a moderate effect, and 0.8 and over represents a large effect.35
A random effects model was used as heterogeneity across study results was expected. The random effects model assumes that the magnitude of the effect may differ across populations, rather than assuming the existence of a single ‘true’ effect size in all sampled populations (as assumed by a fixed effects model). With consideration of the between-subject heterogeneity in schizophrenia, and expected between-study error in biological measurements, a random effects model was most appropriate here. Comprehensive Meta Analysis software provides heterogeneity measures of Q statistics, their related P-values and the I2 statistic, which is the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error. A guideline for interpreting I2 is that 0–40% might not be important, 50–90% ‘may’ be substantial and over 75% should be regarded as ‘considerable’.35 Data were considered imprecise if the confidence interval (CI) surrounding the point estimate was greater than 0.5 in either direction.32
Subgroup analyses and meta-regressions were performed to investigate possible sources of heterogeneity and to test a priori hypotheses (see above); these include possible effects of disease progression and medication exposure (drug-naïve patients in the early stages of illness versus established, medicated schizophrenia patients), medication dosage in medicated patients, age and gender effects. Publication bias was investigated by Classic fail-safe N: the number of missing (unpublished) studies that would bring the observed P-value to >0.05. In addition, Egger's test for publication bias was used to test the degree of funnel plot asymmetry.36
The Medline search resulted in 752 (less duplicates=376) records and two additional research reports were identified by hand search. Of these 378 records, 342 were excluded from full review because they did not report age-matched controlled studies of BDNF levels in human subjects with schizophrenia. Of the 36 papers screened in full, a further 19 were excluded because they did not meet inclusion criteria (see ‘Excluded studies’ section and Figure 1). Data required for meta-analysis were extracted from the remaining 17 studies.
Of the 19 studies excluded after full review, 6 reported data that duplicated cases reported in a more recent study by Xiu et al,27 which met criteria and provided the most comprehensive data set for inclusion in the meta-analyses.37, 38, 39, 40, 41, 42 A further six studies were excluded because they did not report BDNF blood levels in comparison with a healthy control group,43, 44, 45, 46, 47, 48 five reported BDNF levels in post-mortem brain tissue,9, 10, 11, 12 one reported serum BDNF taken from birth in a sample later diagnosed with schizophrenia, and therefore did not fit either of our patient group categories of current BDNF blood levels in medicated or drug-naïve first-episode patients.49 The last study to be excluded reported BDNF levels in pg per μg protein, rather than serum or plasma concentration.30
Characteristics of the 17 studies included in this main analysis are summarized in Table 1, showing that most of the studies used moderate sized samples of schizophrenia subjects (median 89 patients), except for the large study by Xiu et al.27 that included 364 schizophrenia and 324 control patients. All studies had a low probability of reporting bias as determined by the STROBE checklist. Four papers reported BDNF concentration levels in plasma16, 21, 28, 50 and 13 studies reported BDNF concentration levels in blood serum.17, 18, 19, 20, 22, 23, 24, 25, 26, 27, 29, 51, 52
Where possible, raw data were obtained from corresponding authors of individual studies to enable calculation and analyses of BDNF levels according to subgroups defined by gender and patient type (drug naïve or medicated), and/or to enable planned regressions with age and medication dosage on blood BDNF levels (10 studies in total,18, 21, 24, 26, 27, 28, 50, 51 although one study was excluded30 for reporting in pg per μg protein rather than serum or plasma, and one study16 reported a very large effect size and was subsequently removed from the analyses; see Results).
When a study measured blood BDNF levels at two different time points, such as the two studies that reported BDNF levels in schizophrenia patients before and after antipsychotic treatments,21, 23 we included only data from the post-treatment antipsychotic-medicated patient groups, because pretreatment data from these studies did not fit our criteria for drug-naïve patients (that is, at baseline, these patients were only temporarily medication-free). An exception was one study,50 which provided data for both the drug-naïve patient subgroup at baseline (Visit 1) and medicated patients (comprised of mean data calculated across post-treatment visits at 3, 6 and 12 months after baseline). Assessment of publication bias using the Fail-safe N (file drawer statistic) revealed that 476 unincluded or unpublished studies would be required to make P>0.05 in all the patients analysis, supporting the view that the results of our meta-analysis are not likely to be a result of publication bias. Egger's test for publication bias confirmed the symmetry of the funnel plot (intercept=0.003, 95% CI −3.891 to 3.898, t15=0.002, one-tailed P=0.499).
Meta-analyses for all patients
Meta-analysis of 17 studies was conducted on a total of 1114 patients with schizophrenia compared with 970 age-matched healthy controls. The random effects estimate showed a moderate reduction in blood BDNF levels in schizophrenia patients when compared with controls (g=−0.594, s.e.=0.172, 95% CI −0.930 to −0.257, P=0.001; see Figure 2a). Data were precise (CI <0.5 in either direction) but with considerable heterogeneity (I2=91.5%, P<0.001). One study16 showed a greater reduction in blood BDNF levels compared with controls with an effect size at a magnitude of 25 s.e. from the overall mean (g=−4.850, s.e.=0.732, 95% CI −6.285 to −3.415, P<0.001; see Figure 2a, and Table 2). We therefore tested the robustness of the meta-analysis by excluding this study as a potential source of bias (that is, as an extreme score in the predicted direction; and removed from all subsequent analyses). The subsequent random effects model (for 16 studies of 1109 patients with schizophrenia and 956 age-matched controls; see Figure 2b) provided a similar, moderate effect of reduced BDNF in patents with schizophrenia compared with controls (g=−0.458, s.e.=0.159, 95% CI=−0.770 to −0.146, P=0.004). Data were precise, but heterogeneity remained considerable (I2=90%, P<0.001). These findings show moderate quality evidence of reduced BDNF levels in all patients compared with healthy controls. This evidence has been gained from observational studies (assumed to be of low quality), which we have upgraded to moderate due to it being precise, with a large overall sample, but we did not upgrade to high due to considerable heterogeneity, as per GRADE guidelines.
Meta-regressions on the effects of age and medication levels
Using data from 16 studies (that is, excluding Buckley et al.16), we observed a significant association of mean age with blood BDNF levels using meta-regression on SMDs between patients and controls (β=−0.01749, Z=−3.78348, P<0.001). This finding represents a greater reduction in BDNF levels with increasing age in patients compared with the normal reduction in age-matched controls. Inspection of this regression slope showed another putative outlier29 with a reverse association with BDNF and age; exploration of the effects of this study on the overall association between age and BDNF levels were undertaken by removing this study from the analysis; however, the meta-regression showed similar overall results (β=−0.02444, Z=−5.20102, P<0.001).
Eight studies provided data on medication dosage. Using meta-regression on mean chlorpromazine equivalents, we found no effects of medication dosage levels on differences in blood BDNF levels between patients and controls (β=0.00026, Z=0.89705, P=0.370). This analysis may be underpowered, having only 8 studies, as for meta-regression a minimum of 10 studies is recommended.53 However, the low strength of the (nonsignificant) association suggests that the inclusion of two more studies may not have substantially changed the observed effect.
Meta-analyses for medicated and drug-naïve patients
Some studies provided separate data for drug-naïve and medicated patients and are therefore included in both analyses. Data from seven studies reporting BDNF levels for 233 drug-naïve patients and 284 healthy control participants provided evidence of a significant, moderate reduction in BDNF serum/plasma levels compared with control data (g=−0.446, s.e.=0.206, 95% CI −0.850 to −0.043, P=0.030; see Table 2); these data can be considered as precise with considerable heterogeneity (I2=75%, P<0.001) (Note that with the inclusion of Buckley et al.16, reporting drug-naïve patient data only, this subgroup analysis remains significant g=−0.817, s.e.=0.306, 95% CI −0.416 to −0.218, P=0.008), but heterogeneity increases (I2=89%, P<0.001)). Data from 13 studies of 839 medicated patients and 839 healthy control participants also provided precise evidence, but with considerable heterogeneity (I2=91%, P<0.001), of reduced BDNF compared with control samples (g=−0.401, 95% CI −0.765 to −0.036, P=0.031; see Table 2). Inspection of the forest plot (Figure 3) shows one study29 as a putative outlier, being the only study reporting significant results in the opposite direction to the quantitative analysis of all results (similar to the age regression). With Reis et al.29 removed, the finding remained moderate (g=−0.561, s.e.=0.154, 95% CI −0.864 to 0.258, P<0.001); data were still precise, but heterogeneity remained considerably high (I2=85%, P<0.001). Pairwise comparison of SMDs derived from these subgroup meta-analyses could not be undertaken owing to overlapping control samples.50
Meta-analyses for gender
Data from 9 studies reporting BDNF levels for 295 female patients provided evidence of a significant reduction in BDNF serum/plasma levels compared with control data (g=−0.450, s.e.=0.169, 95% CI −0.781 to −0.118, P=0.008; see Table 2); data were precise with considerable heterogeneity (I2=70%; significant P=0.001). In contrast, data from 10 studies of 588 male patients showed no significant reduction in BDNF compared with control samples (g=−0.227, s.e.=0.2586 95% CI −0.728 to −0.274, P=0.375; See Table 2); these data were precise but with considerable heterogeneity (I2=91%, P<0.001). Forest plots for meta-analyses of female and male subgroups are presented in Figure 4. Again, one study contributing data to the male subgroup only29 appeared as a putative outlier (∼8 s.e. from group mean, with increased BDNF levels). With this study29 removed from the analysis, the male subgroup showed a significant reduction in BDNF levels compared with control participants that was similar in magnitude to the females (g=−0.446, s.e.=0.190, 95% CI=−0.818 to −0.074, P=0.019); data were precise, but with considerable heterogeneity (I2=82%, P<0.001). Removing Reis et al.,29 thus provides consistent results of reduced BDNF in both males and females. Pairwise contrasts to compare the degree of reduction in blood BDNF levels in male and female patients were conducted with χ2 between subgroup mixed effects model, but revealed no significant difference between the gender subgroups, either with the Reis et al.29 study included (QB=0.528, P=0.467) or excluded (QB=0.000, P=0.989).
Unexplained heterogeneity: post hoc analyses
Given the considerable heterogeneity observed in the meta-analysis of all patient studies (I2=90%), which was borne out also in both drug-naïve (I2=75%) and medicated patient (I2=91%) subgroup analyses, we conducted post hoc analyses to examine the extent to which methodological differences in the blood compartment analysed (that is, serum vs plasma), or the units used to measure BDNF (that is, pg ml−1 vs ng ml−1) contributed to unexplained heterogeneity. For studies reporting BDNF levels in plasma (underpowered with only three studies21, 28, 50), the random effects estimate was nonsignificant (g=−0.317, P=0.333) and considerable heterogeneity remained (I2=84%). For the 13 studies reporting BDNF levels in serum (see Table 3), the random effects estimate was significant (g=−0.490, P=0.008) but considerable heterogeneity remained (I2=91%). For 11 studies reporting BDNF levels in ng ml−1 (see Table 3), the random effects estimate was significant (g=−0.644, P<0.001) but considerable heterogeneity remained (I2=86%). For studies reporting BDNF levels in pg ml−1 (underpowered with only four studies after removing Reis et al.29 as an outlier), the random effects estimate was nonsignificant (g=−0.392, P=0.306), with considerable heterogeneity remaining (I2=87%).
This systematic review includes data from 17 case–control studies assessing 1144 schizophrenia patients and 970 healthy control subjects (with inclusion of the study with the largest effect size16). Using meta-analysis, we compared BDNF blood levels between schizophrenia patients and healthy controls and found moderate quality evidence of a moderate reduction in peripheral BDNF levels in schizophrenia. Meta-regression on these data for mean age and medication dosage revealed a greater reduction in peripheral BDNF in schizophrenia with increasing age, but no significant effects of medication dosage. Taken together, these two results suggest that antipsychotic medication is not effective at reversing the reductions in blood BDNF levels in patients with schizophrenia. Additional meta-analyses of subgroups defined by gender and medication exposure (drug-naïve vs medicated) revealed moderate quality evidence of reduced BDNF in both drug-naïve and medicated patient groups, and also provided moderate quality evidence for a reduction of BDNF in both male and female patients, although the results for males was dependent on the exclusion of one putative outlier.29
Overall findings for all patients
The main findings confirm our hypothesis that peripheral BDNF levels would be reduced in schizophrenia when all studies’ data were combined; however, there was considerable heterogeneity in the results. We were unable to account for this heterogeneity with subgroup analyses (see below), or by post hoc exploration of differences between studies in BDNF measurement (for example, BDNF measurement scale or blood component from which BDNF measurements were derived). We speculate on the source of the heterogeneity in the discussion of limitations, below.
The moderating effects of age on these data, in which there were greater reductions in blood BDNF in schizophrenia with increasing age, relative to controls, is consistent with evidence of age-related decline in brain BDNF levels in schizophrenia.10, 12 These findings can be considered in the context of a recent comparison of blood BDNF levels in healthy children, adolescents and adults that suggests that there is progressive reduction in blood BDNF from childhood to adulthood.31 Convergent evidence of age-related decline in blood and brain BDNF in schizophrenia might therefore reflect an accelerated decline in normal age-related BDNF reduction, which may be occurring alongside disease progression in schizophrenia. Alternatively, schizophrenia may be characterized by a significant reduction in BDNF levels beginning early in development and continuing to lag behind those of healthy individuals with increasing age. Longitudinal investigation of blood and brain BDNF levels in schizophrenia are required to clarify these issues; that is, whether BDNF reduction occurs early in neurodevelopment and contributes to the onset of disease or represents a biomarker of disease progression during later stages of neurodevelopment, during the period of disease onset in adolescence and early adulthood.
The meta-regression of eight studies reporting usable data on antipsychotic medication dosage revealed no significant association of medication with differences in BDNF levels in schizophrenia patients. This analysis may have been underpowered owing to the limited number of studies included. Furthermore, different effects of typical and atypical antipsychotics may differentially affect blood17 and brain.4, 54 BDNF levels in schizophrenia, but we were unable to analyse the effect of antipsychotic subtype owing to lack of sufficiently detailed information in reporting on types of medication. Strong evidence for the effects of antidepressant medication on increasing blood BDNF levels55, 56, 57 suggests antidepressant medication dosage ought to be accounted for in studies of peripheral BDNF levels in schizophrenia. Future research should consider assessing antidepressant type and dosage to help determine their effect on blood BDNF levels in schizophrenia.
Subgroup analyses by medication exposure and gender
Contrary to our initial hypothesis, meta-analyses of patient subgroups according to gender provided evidence of a significant reduction in blood BDNF levels in female schizophrenia patients but not males, relative to controls. However, with a putative outlier,29 removed from the male subgroup meta-analysis, a significant reduction in BDNF levels was also revealed for the male subgroup, compared with control participants. Removing the Reis et al.29 study thus provided evidence of consistently reduced BDNF in both male and female schizophrenia patients compared with healthy controls. This still did not support the hypothesis that males with schizophrenia would have a greater reduction in blood BDNF than females with schizophrenia, and instead suggests that the magnitude of the blood BDNF reduction may be similar across gender.
A potential source of bias in any review is a failure to retrieve a comprehensive sample of studies. This problem applies to the present meta-analysis, where our search strategy may have missed some studies, particularly unpublished studies. However, the results of our Fail-safe N and Egger's test analyses suggest that this review is unlikely to be subject to publication bias. Another limitation is the absence of direct, pairwise comparisons of SMDs between subgroups defined by medication exposure, which we were unable to perform owing to overlapping control samples. In addition, characteristics such as medication type may be important, with studies of chronic patients more likely to involve typical rather than atypical antipsychotic medication; for example, Grillo et al.17 reports nonsignificantly higher levels of serum BDNF in schizophrenia patients taking clozapine compared with typical antipsychotics. The issue of whether changes in BDNF levels are specific to certain types of antipsychotic treatments (for example, typical vs atypical) could not be addressed in this study because antipsychotic medication types (and their mean dosages) were not always reported by individual studies. Other patient characteristics not available for consideration in the present meta-analyses include symptom severity and duration of medication. If antidepressant data were available, this may have helped to quantify the effects of medication on BDNF levels in schizophrenia more precisely. Finally, owing to lack of other psychiatric comparison groups in this analysis, the main finding of reduced BDNF levels in schizophrenia must be considered alongside findings of altered BDNF levels associated with several other neuropsychiatric disorders, including unipolar depression,57, 58 bipolar disorders30, 50, 59 and neurodegenerative disease.60 In this context, the status of reduced BDNF or as risk-indicator (biomarker) for schizophrenia26 remains to be clarified with regard to disease specificity; notably, BDNF has been proposed as a biomarker of disease development in Alzheimer's disease,61 and of treatment response in affective disorders.56, 62 Longitudinal studies may alternatively delineate a causal role for altered BDNF by its effects on neuronal signalling during neurodevelopment that may represent an intermediate phenotype on the path to psychotic illness.63
Peripheral BDNF levels are reduced in drug-naïve and medicated schizophrenia samples when compared with age-matched healthy controls, and this difference increases with age, but is not moderated by antipsychotic medication dosage. These data are precise but significant heterogeneity across study results remains unexplained. There is a need for direct pairwise comparisons of effect sizes for subgroups of patients taking different types of medication, and investigation of the effects of symptom type and severity, as well as antidepressant use on peripheral blood levels of BDNF in schizophrenia. More comprehensive reporting of all relevant data in future case–control studies of peripheral BDNF concentrations in schizophrenia would enable clarification of the role of reduced BDNF in the disease process.
Numakawa T, Suzuki S, Kumamaru E, Adachi N, Richards M, Kunugi H . BDNF function and intracellular signaling in neurons. Histol Histopathol 2010; 25: 237–258.
Bekinschtein P, Cammarota M, Katche C, Slipczuk L, Rossato JI, Goldin A et al. BDNF is essential to promote persistence of long-term memory storage. Proc Natl Acad Sci USA 2008; 105: 2711–2716.
Yamada K, Nabeshima T . Brain-derived neurotrophic factor/TrkB signaling in memory processes. J Pharmacol Sci 2003; 91: 267–270.
Lipska BK, Lerman DN, Khaing ZZ, Weinberger DR . The neonatal ventral hippocampal lesion model of schizophrenia: effects on dopamine and GABA mRNA markers in the rat midbrain. Eur J Neurosci 2003; 18: 3097–3104.
van Haren NEM, Cahn W, Hulshoff Pol HE, Kahn RS . Schizophrenia as a progressive brain disease. Eur Psychiatry 2008; 23: 245–254.
Toro CT, Hallak JE, Dunham JS, Leite JP, Sakamoto AC, Guarnieri R et al. The NR1 N-methyl-D-aspartate subunit and brain-derived neurotrophic factor in temporal lobe epilepsy hippocampus: a comparison of patients with and without coexisting psychiatric symptoms. Epilepsia 2007; 48: 2352–2356.
Tanaka T, Saito H, Matsuki N . Inhibition of GABAA synaptic responses by brain-derived neurotrophic factor (BDNF) in rat hippocampus. J Neurosci 1997; 17: 2959–2966.
Mattson MP . Glutamate and neurotrophic factors in neuronal plasticity and disease. Ann N Y Acad Sci 2008; 1144: 97–112.
Durany N, Michel T, Zochling R, Boissl KW, Cruz-Sanchez FF, Riederer P et al. Brain-derived neurotrophic factor and neurotrophin 3 in schizophrenic psychoses. Schizophrenia Res 2001; 52: 79–86.
Hashimoto T, Bergen SE, Nguyen QL, Xu B, Monteggia LM, Pierri JN et al. Relationship of brain-derived neurotrophic factor and its receptor TrkB to altered inhibitory prefrontal circuitry in schizophrenia. J Neurosci 2005; 25: 372.
Mellios N, Huang HS, Baker SP, Galdzicka M, Ginns E, Akbarian S . Molecular determinants of dysregulated GABAergic gene expression in the prefrontal cortex of subjects with schizophrenia. Biol Psychiatry 2009; 65: 1006–1014.
Weickert CS, Hyde TM, Lipska BK, Herman MM, Weinberger DR, Kleinman JE . Reduced brain-derived neurotrophic factor in prefrontal cortex of patients with schizophrenia. Mol Psychiatry 2003; 8: 592–610.
Takahashi M, Shirakawa O, Toyooka K, Kitamura N, Hashimoto T, Maeda K et al. Abnormal expression of brain-derived neurotrophic factor and its receptor in the corticolimbic system of schizophrenic patients. Mol Psychiatry 2000; 5: 293–300.
Wong J, Hyde TM, Cassano HL, Deep-Soboslay A, Kleinman JE, Shannon Weickert C . Promoter specific alterations of BNDF mRNA in schizophrenia. Neuroscience 2010; 169: 1071–1084.
Huang EJ, Reichardt LF . Neurotrophins: roles in neuronal development and function. Annu Rev Neurosci 2001; 24: 677–736.
Buckley PF, Pillai A, Evans D, Stirewalt E, Mahadik S . Brain derived neurotropic factor in first-episode psychosis. Schizophrenia Res 2007; 91: 1–5.
Grillo RW, Ottoni GL, Leke R, Souza DO, Portela LV, Lara DR . Reduced serum BDNF levels in schizophrenic patients on clozapine or typical antipsychotics. J Psychiatric Res 2006; 41: 31–35.
Ikeda Y, Yahata N, Ito I, Nagano M, Toyota T, Yoshikawa T et al. Low serum levels of brain-derived neurotrophic factor and epidermal growth factor in patients with chronic schizophrenia. Schizophrenia Res 2008; 101: 58–66.
Jindal RD, Pillai AK, Mahadik SP, Eklund K, Montrose DM, Keshavan MS . Decreased BDNF in patients with antipsychotic naïve first episode schizophrenia. Schizophrenia Res 2010; 119: 47–51.
Jockers-Scherubl MC, Danker-Hopfe H, Mahlberg R, Selig F, Rentzsch J, Schurer F et al. Brain-derived neurotrophic factor serum concentrations are increased in drug-naive schizophrenic patients with chronic cannabis abuse and multiple substance abuse. Neurosci Lett 2004; 371: 79–83.
Lee BH, Kim YK . Increased plasma brain-derived neurotropic factor, not nerve growth factor-beta, in schizophrenia patients with better response to risperidone treatment. Neuropsychobiology 2009; 59: 51–58.
Mackin P, Gallagher P, Watson S, Young A, Ferrier IN . Changes in brain-derived neurotrophic factor following treatment with mifepristone in bipolar disorder and schizophrenia. Australian and New Zealand J Psychiatry 2007; 41: 321–326.
Pirildar S, Gonul AS, Taneli F, Akdeniz F . Low serum levels of brain-derived neurotrophic factor in patients with schizophrenia do not elevate after antipsychotic treatment. Prog Neuropsychopharmacol Biol Psychiatry 2004; 28: 709–713.
Shimizu E, Hashimoto K, Watanabe H, Komatsu N, Okamura N, Koike K et al. Serum brain-derived neurotrophic factor (BDNF) levels in schizophrenia are indistinguishable from controls. Neurosci Lett 2003; 351: 111–114.
Toyooka K, Asama K, Watanabe Y, Muratake T, Takahashi M, Someya T et al. Decreased levels of brain-derived neurotrophic factor in serum of chronic schizophrenic patients. Psychiatry Res 2002; 110: 249–257.
Vinogradov S, Fisher M, Holland C, Shelly W, Wolkowitz O, Mellon SH . Is serum brain-derived neurotrophic factor a biomarker for cognitive enhancement in schizophrenia? Biol Psychiatry 2009; 66: 549–553.
Xiu MH, Hui L, Dang YF, Hou TD, Zhang CX, Zheng YL et al. Decreased serum BDNF levels in chronic institutionalized schizophrenia on long-term treatment with typical and atypical antipsychotics. Prog Neuropsychopharmacol Biol Psychiatry Clin Neurosci 2009; 33: 1508–1512.
Yoshimura R, Hori H, Sugita A, Ueda N, Kakihara S, Umene W et al. Treatment with risperidone for 4 weeks increased plasma 3-methoxy-4-hydroxypnenylglycol (MHPG) levels, but did not alter plasma brain-derived neurotrophic factor (BDNF) levels in schizophrenic patients. Prog Neuropsychopharmacol Biol psychiatry 2007; 31: 1072–1077.
Reis HJ, Nicolato R, Barbosa IG, Teixeira do Prado PH, Romano-Silva MA, Teixeira AL . Increased serum levels of brain-derived neurotrophic factor in chronic institutionalized patients with schizophrenia. Neurosci Lett 2008; 439: 157–159.
Gama CS, Andreazza AC, Kunz M, Berk M, Belmonte-de-Abreu PS, Kapczinski F . Serum levels of brain-derived neurotrophic factor in patients with schizophrenia and bipolar disorder. Neurosci Lett 2007; 420: 45–48.
Katoh-Semba R, Wakako R, Komori T, Shigemi H, Miyazaki N, Ito H et al. Age-related changes in BDNF protein levels in human serum: differences between autism cases and normal controls. Int J Dev Neurosci 2007; 25: 367–372.
Brozek J, Oxman A, Schünemann H . GRADEpro. Version 32 for Windows 2008. Computer program.
Grade WG . Grading quality of evidence and strength of recommendations. Br Med J 2004; 328: 1490.
Borenstein M, Hedges L, Higgins J, Rothstein H . Comprehensive Meta-Analysis. Version 2 Biostat: Engelwood, NJ, 2005.
Higgins J, Green S . Cochrane Handbook for Systematic Reviews of Interventions. The Cochrane Collaboration 2008. Version 501 2008.
Egger M, Davey Smith G, Schneider M, Minder C . Bias in meta-analysis detected by a simple, graphical test. Br Med J 1997; 315: 629–634.
Chen DC, Wang J, Wang B, Yang SC, Zhang CX, Zheng YL et al. Decreased levels of serum brain-derived neurotrophic factor in drug-naïve first-episode schizophrenia: relationship to clinical phenotypes. Psychopharmacology 2009; 207: 375–380.
Tan LY, Zhou DF, Zhang XY . Decreased plasma brain-derived neurotrophic factor levels in schizophrenic patients with tardive dyskinesia: association with dyskinetic movements. Schizophrenia Res 2005; 74: 263–270.
Tan YL, Zhou DF, Zhang XY . Decreased plasma brain-derived neurotrophic factor levels in schizophrenic patients with tardive dyskinesia: association with dyskinetic movements. Neurosci Lett 2005; 382: 27–32.
Yun LT, Dong FZ, Lian YC, Yi ZZ, Xiang YZ . Decreased BDNF in serum of patients with chronic schizophrenia on long-term treatment with antipsychotics. Neurosci Lett 2005; 382: 27–32.
Zhang XY, Zhou DF, Wu GY, Cao LY, Tan YL, Haile CN et al. BDNF levels and genotype are associated with antipsychotic-induced weight gain in patients with chronic schizophrenia. Neuropsychopharmacology 2008; 33: 2200–2205.
Zhang XY, Tan YL, Zhou DF, Cao LY, Wu GY, Xu Q et al. Serum BDNF levels and weight gain in schizophrenic patients on long-term treatment with antipsychotics. J Psychiatric Res 2007; 41: 997–1004.
Guimaraes LR, Jacka FN, Gama CS, Berk M, Leitao-Azevedo CL, Belmonte de Abreu MG et al. Serum levels of brain-derived neurotrophic factor in schizophrenia on a hypocaloric diet. Prog Neuropsychopharmacol Biol Psychiatry 2008; 32: 1595–1598.
Hedges S, El-Mallakh RS, Carvey PM, Suddath RL, Wyatt RL . Dopamine-responsive neurotrophic factor in schizophrenia. Schizophrenia Res 1995; 18: 83–84.
Hori H, Yoshimura R, Yamada Y, Ikenouchi A, Mitoma M, Ida Y et al. Effects of olanzapine on plasma levels of catecholamine metabolites, cytokines, and brain-derived neurotrophic factor in schizophrenic patients. Inter Clin Psychopharmacol 2007; 22: 21–27.
Huang TL, Hung YY . Lorazepam reduces the serum brain-derived neurotrophic factor level in schizophrenia patients with catatonia. Prog Neuropsychopharmacol Biol Psychiatry 2009; 33: 158–159.
Lee JG, Shin BS, Lee YC, Park SW, Kim YH . Clinical effectiveness of the Kampo medicine kamishoyosan for adjunctive treatment of tardive dyskinesia in patients with schizophrenia: a 16-week open trial. Psychiatry Clin Neurosci 2007; 61: 509–514.
Gama CS, Berk M, Andreazza AC, Kapczinski F, Belmonte-de-Abreu P . Serum levels of brain-derived neurotrophic factor and thiobarbituric acid reactive substances in chronically medicated schizophrenic patients: a positive correlation. Revista Brasileira de Psiquiatria 2008; 34: 337–340.
Cannon TD, Yolken R, Buka S, Torrey EF . Decreased neurotrophic response to birth hypoxia in the etiology of schizophrenia. Biol Psychiatry 2008; 64: 797–802.
Palomino A, Vallejo-Illarramendi A, Gonzalez-Pinto A, Aldama A, Gonzalez-Gomez C, Mosquera F et al. Decreased levels of plasma BDNF in first-episode schizophrenia and bipolar disorder patients. Schizophrenia Res 2006; 86: 321–322.
Huang TL, Lee CT . Associations between serum brain-derived neurotrophic factor levels and clinical phenotypes in schizophrenia patients. J Psychiatric Res 2006; 40: 664–668.
Rizos EN, Rontos I, Laskos E, Arsenis G, Michalopoulou PG, Vasilopoulos D et al. Investigation of serum BDNF levels in drug-naive patients with schizophrenia. Prog Neuropsychopharmacol Biol Psychiatry 2008; 32: 1308–1311.
Borenstein M, Hedges LV, Higgins JPT, Rothstein HR . Introduction to Meta-Analysis. John Wiley & Sons, Ltd.: West Sussex, UK, 2009.
Chlan-Fourney J, Ashe P, Nylen K, Juorio AV, Li XM . Differential regulation of hippocampal BDNF mRNA by typical and atypical antipsychotic administration. Brain Res 2002; 954: 11–20.
Brunoni AR, Lopes M, Fregni F . A systematic review and meta-analysis of clinical studies on major depression and BDNF levels: implications for the role of neuroplasticity in depression. Int J Neuropsychopharmcol 2008; 11: 1169–1180.
Russo-Neustadt AA, Beard RC, Huang YM, Cotman CW . Physical activity and antidepressant treatment potentiate the expression of specific brain-derived neurotrophic factor transcripts in the rat hippocampus. Neuroscience 2000; 101: 305–312.
Shimizu E, Hashimoto K, Okamura N, Koike K, Komatsu N, Kumakiri C et al. Alterations of serum levels of brain-derived neurotrophic factor (BDNF) in depressed patients with or without antidepressants. Biol Psychiatry 2003; 54: 70–75.
Ozan E, Okur H, Eker C, Eker OD, Gonul AS, Akarsu N . The effect of depression, BDNF gene val66met polymorphism and gender on serum BDNF levels. Brain Res Bull 2010; 81: 61–65.
Tseng M, Alda M, Xu L, Sun X, Wang J-F, Grof P et al. BDNF protein levels are decreased in transformed lymphoblasts from lithium-responsive patients with bipolar disorder. J Psychiatry Neurosci 2008; 33: 449–453.
Zuccato C, Cattaneo E . Brain-derived neurotrophic factor in neurodegenerative diseases. Nat Rev Neurol 2009; 5: 311–322.
Peng S, Wuu J, Mufson EJ, Fahnestock M . Precursor form of brainderived neurotrophic factor and mature brain-derived neurotrophic factor are decreased in the pre-clinical stages of Alzheimer's disease. J Neurochem 2005; 93: 1412–1421.
Gass P, Hellweg R . Peripheral brain-derived neurotrophic factor (BDNF) as a biomarker for affective disorders? Int J Neuropsychopharmcol 2010; 13: 1–4.
Kendler KS, Neale MC . Endophenotype: a conceptual analysis. Mol Psychiatry 2010; 15: 789–797.
This project was supported by research funding from the Australian National Health and Medical Research Council (NHMRC: Project Grant 630471), the Australian Research Council (Future Fellowship: FT0991511) and the Schizophrenia Research Institute, using an infrastructure grant from the NSW Department of Health. We acknowledge Amy Sparks for assistance with production of the paper, and the following authors for the provision of raw data for inclusion in the meta-analysis: Peter Buckley, Clarissa Gama, Alex Genevsky, Ana Maria González-Pinto, TL, Huang, Yumiko Ikeda, Flávio Kapczinski, Yong-Ku Kim, Diogo Rizzato Lara, Carlos Matute, E Rizos, Eiji Shimizu, Amaia Ugarte Ugarte, Sophia Vinogradov, Reiji Yoshimura and Xiang Yang Zhang.
The authors declare no conflict of interest.
About this article
Cite this article
Green, M., Matheson, S., Shepherd, A. et al. Brain-derived neurotrophic factor levels in schizophrenia: a systematic review with meta-analysis. Mol Psychiatry 16, 960–972 (2011). https://doi.org/10.1038/mp.2010.88
- blood serum
Alterations in DNA methylation rates of brain-derived neurotrophic factor in patients with schizophrenia
The European Journal of Psychiatry (2021)
Pharmacology & Therapeutics (2021)
BDNF affects the mediating effect of negative symptoms on the relationship between age of onset and cognition in patients with chronic schizophrenia
Frontiers in Psychiatry (2020)