On 2015 May 14, Suzanne Gage commented:

Power and colleagues¹ report evidence that individuals with genetic risk of schizophrenia smoke cannabis more, and in greater quantities, than the general population, arguing that these findings are evidence of gene-environment correlation, and suggest that part of the association between schizophrenia and cannabis use is due to shared genetic aetiology. However, an alternate explanation, which is hinted at in the discussion, is that genetic predisposition to schizophrenia may increase the risk of using cannabis. We discuss how causal associations from observational data can be assessed using Mendelian randomisation (MR) designs.

The study by Power and colleagues¹ is in some ways analogous to a MR design, whereby genetic information is used as a proxy for an exposure, in order to ascertain causal relationships from observational data.² In principle, this approach can overcome problems of confounding and reverse causality. While MR has more traditionally been utilised to study the consequence of modifiable risk factors, rather than disease processes, it can be utilised to study the consequences of disease-like states. For example, genetic variants related to hypercholesterolaemia are associated with higher prevalence of the use of cholesterol-lowering drugs (primarily statins).³ The findings of Power and colleagues are consistent with the interpretation that schizophrenia (or its prodrome) might increase the likelihood of cannabis use. The concept of a schizophrenia prodrome, whereby changes in brain chemistry may occur long before symptoms reach a clinical threshold, is increasingly accepted.⁴ Observational studies of cannabis use indicating an association with later psychosis, psychotic symptoms, or schizophrenia are not inconsistent with the self-medication hypothesis. This could be directly tested, using indicators that predict later transition to schizophrenia. For example, psychotic-like experiences might be indicative of prodromal status, and would therefore be expected to be associated with a genetic risk score for schizophrenia, if this predicts later transition to clinical psychosis. In this case, a MR design using a schizophrenia polygenic risk score as the exposure would provide some insight into whether the association between schizophrenia and cannabis use is due to self-medication.

The recent genome-wide association study (GWAS) by the Psychiatric Genomics Consortium (PGC) includes a polygenic risk score for schizophrenia risk which captures 18% of phenotypic variance,⁵ compared with 7% for the score used by Power and colleagues derived from an earlier study by the PGC.⁶ This is likely to be a suitable instrument for MR-style analyses, allowing further insight in to the relationship between cannabis use and schizophrenia through the use of bidirectional MR.⁷ However, an MR design using a schizophrenia risk score would also predict cannabis use if the association were due to genetic covariance (as Power and colleagues argue), rather than a causal effect on cannabis use. Fractionating the polygenic score could help here, as the hypothesis that a schizophrenia prodrome increases cannabis use would suggest that any combination of variants related to schizophrenia should relate to cannabis use.⁷ Genetic covariance would not be expected to exist for the whole gamut of the schizophrenia-related genetic variants. Furthermore, a more complex design combining risk scores and longitudinal data would be required to disentangle these possibilities and provide better evidence of the direction of causation.

Finally, it is well established that cannabis use is also heritable,⁸ and although no specific loci have yet been robustly identified, as larger GWAS of cannabis phenotypes are undertaken it is likely that they will be. MR studies of cannabis use as an exposure will then become possible, allowing the investigation of the causal association between cannabis use and schizophrenia (i.e., that schizophrenia is a consequence of cannabis use). Such studies have already been used to investigate causal associations of tobacco on various outcomes.^9,10 MR methods come with caveats, such as the need for large sample sizes, and may be invalid if certain assumptions are violated (e.g., pleiotropy, canalization, linkage disequilibrium and population stratification).^2,7 Nevertheless, they offer an opportunity to better understand the complex question of the nature of the relationship between cannabis use and schizophrenia.

Suzanne H. Gage, George Davey Smith and Marcus R. Munafò

References

1.Power RA, Verweij KJ, Zuhair M, Montgomery GW, Henders AK, Heath AC et al. Genetic predisposition to schizophrenia associated with increased use of cannabis. Molecular psychiatry 2014.

2.Davey Smith G, Ebrahim S. Mendelian randomization: prospects, potentials, and limitations. International journal of epidemiology 2004; 33(1): 30-42.

3.Benn M, Tybjaerg-Hansen A, Stender S, Frikke-Schmidt R, Nordestgaard BG. Low-density lipoprotein cholesterol and the risk of cancer: a mendelian randomization study. Journal of the National Cancer Institute 2011; 103(6): 508-519.

4.Dominguez MD, Wichers M, Lieb R, Wittchen HU, van Os J. Evidence that onset of clinical psychosis is an outcome of progressively more persistent subclinical psychotic experiences: an 8-year cohort study. Schizophrenia bulletin 2011; 37(1): 84-93.

5.Schizophrenia-Working-Group-of-the-Psychiatric-Genomics-Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 2014; 511(7510): 421-427.

6.Ripke S, O'Dushlaine C, Chambert K, Moran JL, Kahler AK, Akterin S et al. Genome-wide association analysis identifies 13 new risk loci for schizophrenia. Nature genetics 2013; 45(10): 1150-1159.

7.Davey Smith G, Hemani G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet in press.

8.Agrawal A, Lynskey MT. The genetic epidemiology of cannabis use, abuse and dependence. Addiction 2006; 101(6): 801-812.

9.Asvold BO, Bjorngaard JH, Carslake D, Gabrielsen ME, Skorpen F, Davey Smith G et al. Causal associations of tobacco smoking with cardiovascular risk factors: a Mendelian randomization analysis of the HUNT Study in Norway. International journal of epidemiology 2014.

10.Freathy RM, Kazeem GR, Morris RW, Johnson PC, Paternoster L, Ebrahim S et al. Genetic variation at CHRNA5-CHRNA3-CHRNB4 interacts with smoking status to influence body mass index. International journal of epidemiology 2011; 40(6): 1617-1628.

On 2015 May 14, Suzanne Gage commented:

Power and colleagues¹ report evidence that individuals with genetic risk of schizophrenia smoke cannabis more, and in greater quantities, than the general population, arguing that these findings are evidence of gene-environment correlation, and suggest that part of the association between schizophrenia and cannabis use is due to shared genetic aetiology. However, an alternate explanation, which is hinted at in the discussion, is that genetic predisposition to schizophrenia may increase the risk of using cannabis. We discuss how causal associations from observational data can be assessed using Mendelian randomisation (MR) designs.

The study by Power and colleagues¹ is in some ways analogous to a MR design, whereby genetic information is used as a proxy for an exposure, in order to ascertain causal relationships from observational data.² In principle, this approach can overcome problems of confounding and reverse causality. While MR has more traditionally been utilised to study the consequence of modifiable risk factors, rather than disease processes, it can be utilised to study the consequences of disease-like states. For example, genetic variants related to hypercholesterolaemia are associated with higher prevalence of the use of cholesterol-lowering drugs (primarily statins).³ The findings of Power and colleagues are consistent with the interpretation that schizophrenia (or its prodrome) might increase the likelihood of cannabis use. The concept of a schizophrenia prodrome, whereby changes in brain chemistry may occur long before symptoms reach a clinical threshold, is increasingly accepted.⁴ Observational studies of cannabis use indicating an association with later psychosis, psychotic symptoms, or schizophrenia are not inconsistent with the self-medication hypothesis. This could be directly tested, using indicators that predict later transition to schizophrenia. For example, psychotic-like experiences might be indicative of prodromal status, and would therefore be expected to be associated with a genetic risk score for schizophrenia, if this predicts later transition to clinical psychosis. In this case, a MR design using a schizophrenia polygenic risk score as the exposure would provide some insight into whether the association between schizophrenia and cannabis use is due to self-medication.

The recent genome-wide association study (GWAS) by the Psychiatric Genomics Consortium (PGC) includes a polygenic risk score for schizophrenia risk which captures 18% of phenotypic variance,⁵ compared with 7% for the score used by Power and colleagues derived from an earlier study by the PGC.⁶ This is likely to be a suitable instrument for MR-style analyses, allowing further insight in to the relationship between cannabis use and schizophrenia through the use of bidirectional MR.⁷ However, an MR design using a schizophrenia risk score would also predict cannabis use if the association were due to genetic covariance (as Power and colleagues argue), rather than a causal effect on cannabis use. Fractionating the polygenic score could help here, as the hypothesis that a schizophrenia prodrome increases cannabis use would suggest that any combination of variants related to schizophrenia should relate to cannabis use.⁷ Genetic covariance would not be expected to exist for the whole gamut of the schizophrenia-related genetic variants. Furthermore, a more complex design combining risk scores and longitudinal data would be required to disentangle these possibilities and provide better evidence of the direction of causation.

Finally, it is well established that cannabis use is also heritable,⁸ and although no specific loci have yet been robustly identified, as larger GWAS of cannabis phenotypes are undertaken it is likely that they will be. MR studies of cannabis use as an exposure will then become possible, allowing the investigation of the causal association between cannabis use and schizophrenia (i.e., that schizophrenia is a consequence of cannabis use). Such studies have already been used to investigate causal associations of tobacco on various outcomes.^9,10 MR methods come with caveats, such as the need for large sample sizes, and may be invalid if certain assumptions are violated (e.g., pleiotropy, canalization, linkage disequilibrium and population stratification).^2,7 Nevertheless, they offer an opportunity to better understand the complex question of the nature of the relationship between cannabis use and schizophrenia.

Suzanne H. Gage, George Davey Smith and Marcus R. Munafò

References

1.Power RA, Verweij KJ, Zuhair M, Montgomery GW, Henders AK, Heath AC et al. Genetic predisposition to schizophrenia associated with increased use of cannabis. Molecular psychiatry 2014.

2.Davey Smith G, Ebrahim S. Mendelian randomization: prospects, potentials, and limitations. International journal of epidemiology 2004; 33(1): 30-42.

3.Benn M, Tybjaerg-Hansen A, Stender S, Frikke-Schmidt R, Nordestgaard BG. Low-density lipoprotein cholesterol and the risk of cancer: a mendelian randomization study. Journal of the National Cancer Institute 2011; 103(6): 508-519.

4.Dominguez MD, Wichers M, Lieb R, Wittchen HU, van Os J. Evidence that onset of clinical psychosis is an outcome of progressively more persistent subclinical psychotic experiences: an 8-year cohort study. Schizophrenia bulletin 2011; 37(1): 84-93.

5.Schizophrenia-Working-Group-of-the-Psychiatric-Genomics-Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 2014; 511(7510): 421-427.

6.Ripke S, O'Dushlaine C, Chambert K, Moran JL, Kahler AK, Akterin S et al. Genome-wide association analysis identifies 13 new risk loci for schizophrenia. Nature genetics 2013; 45(10): 1150-1159.

7.Davey Smith G, Hemani G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet in press.

8.Agrawal A, Lynskey MT. The genetic epidemiology of cannabis use, abuse and dependence. Addiction 2006; 101(6): 801-812.

9.Asvold BO, Bjorngaard JH, Carslake D, Gabrielsen ME, Skorpen F, Davey Smith G et al. Causal associations of tobacco smoking with cardiovascular risk factors: a Mendelian randomization analysis of the HUNT Study in Norway. International journal of epidemiology 2014.

10.Freathy RM, Kazeem GR, Morris RW, Johnson PC, Paternoster L, Ebrahim S et al. Genetic variation at CHRNA5-CHRNA3-CHRNB4 interacts with smoking status to influence body mass index. International journal of epidemiology 2011; 40(6): 1617-1628.