Mendelian Randomization of Vitamin D for Preeclampsia: Embracing Complexity of Inference
Mendelian Randomization of Vitamin D for Preeclampsia: Embracing Complexity of Inference
Hooman Mirzakhani, M.D., MMSc., Ph.D.1; Scott T. Weiss, M.D., M.S.1,2
1Channing Division of Network Medicine, Department of Medicine, Brigham and Women's
Hospital, Harvard Medical School, Boston, MA, USA
1,2Partners Center for Personalized Medicine, Partners Health Care, Boston, MA, USA
[A letter to the editor regarding Mangus et al. Vitamin D and risk of pregnancy-related hypertensive disorders: mendelian randomisation study, BMJ 2018;361:k2167]; Word count: 1000
We read with interest the study by Mangus and colleagues who applied Mendelian randomization (MR) to examine the causal nature of the association between 25-hydroxyvitamin D (25OHD) and pregnancy-induced hypertension.1 We think that the accuracy and relevance of the results of MR studies should be interpreted in light of other sources of evidence and biological knowledge of this exposure-outcome relationship.
The individual result from a few vitamin D (VD) supplementation RCTs during pregnancy has not been supportive of VD supplementation to prevent preeclampsia and gestational hypertension. These studies varied in the dose (400–4,000 IU/day), time of initiation and in the baseline and attained VD levels. Except for one trial, all of the RCTs started the intervention after 20 weeks of gestation, the time that the two conditions generally become symptomatic.2 A recent systematic review by the Cochrane database on VD supplementation during pregnancy indicated a trend toward the reduction of preeclampsia risk in pregnant women supplemented with VD compared to no supplementation.3 The question remains whether we can expect to obtain further applicable insight from MR approaches or well-designed RCTs based on the prior knowledge are more applicable to conclusively investigate a preventive role of VD for adverse pregnancy outcomes.
To answer this question, we should first assess the plausibility of the MR assumptions in this context. The first assumption is the presence and strength of the relationship of VD variants with 25OHD levels. For continuous variables and a single instrument, statistical power for a fixed sample size is a function of the proportion of variation in the 25OHD levels explained by the variants and the magnitude of the association between the VD level and preeclampsia.4 According to GWAS studies, the identified SNPs explain less than 5% of the 25OHD varitaiton.5 6 Given this variation, a very large sample size is required to achieve sufficient statistical power for revealing 25OHD causal effects. This variability was less than 1.3% in Mangus and colleagues’ study. Thus, this report exemplifies a weak instrument bias due to using one or more genetic variants that only explain a small proportion of the variation in the risk factor, coupled with a small sample size.7 Preeclampsia has a heterogeneous phenotype with a low incidence in pregnancy (3-8% ) less than other common diseases such as cardiovascular disease. Therefore, the 25OHD effect size varies in different studies. This fact should also be considered in sample size calculation.
Another key assumption is the ‘no horizontal pleiotropy’, which requires that the VD SNPs acts on preeclampsia exclusively through the 25OHD. A violation of this principle can distort the result and obscure the conclusiveness. Some of the SNPs associated with 25OHD levels also influ¬ence VD binding protein (DBP) concentration and affinity for 25OHD; these SNPs might alter bioavailable 25OHD independently of total 25OHD concentration.8 Moy and colleagues showed that associated GC variants (including rs2282679 in Mangus and colleagues’ study) with lower 25OHD concentrations were strongly related to lower levels of DBP. Another finding is by Komangata and colleagues who showed the post-transcriptional regulation of CYP24A1 by miR-125b such that its increased expression resulted in the reduction of CYP24 protein and CYP24-mediated enzymatic activity.9 10. CYP24A1 encodes 24-hydroxylase, which initiates the degradation of 25OHD and 1,25(OH)2D.11 Furthermore, the adipocytokine profile has been shown to be different in preeclampsia and VD status in relation to VD genes.12 13 These observations potentially violate the second MR assumption in such studies. Additionally, VD is known to initiate down-stream expression of literally hundreds of genes making the use of a few SNPs a poor proxy for levels let alone the downstream effects of these levels.
Most GWAS studies have been conducted in non-pregnant individuals. Pregnancy is characterized by profound metabolic alterations to ensure adequate maternal-fetal physiological demands. VD metabolism is unique during pregnancy such that the conversion of 25OHD to 1,25(OH)2D in both mother and or fetus is not controlled by classic calcium homeostatic mechanisms. By 12 weeks of gestation, 1,25(OH)2D serum concentrations are more than twice that of a non-pregnant woman and continue to rise two- to threefold from the non-pregnant baseline.14 While, the heritability of maternal metabolic traits during pregnancy has not been examined, there is a greater possibility of gene-environment (e.g., pre-pregnancy BMI15) or gene-gene interactions. Furthermore, most instrumental variable methods assume a linear relationship between exposure and outcome of interest. As suggested by recent studies, it is reasonable to also examine a non-linear relationship of 25OHD with preeclampsia.2 14
MR studies cannot provide information on the dose-response relationship or the timing and duration of VD intervention. This information can only be obtained when the instrumental variable explains a high degree of variability in the exposure of interest throughout the lifespan. A defined optimal serum level of VD during pregnancy is desirable and has been suggested, but requires to be validated.5 More importantly, there might be a critical period for the vitamin D’s role in the prevention of preeclampsia at a preclinical stage of preeclampsia.2 After preeclampsia initiation, the progression to the clinical stage might be further under the influence of other factors. In a well-designed controlled clinical trial, the above factors can be considered, and relative influence of various VD SNPs on dose-response effect and the outcome further explored.
The MR approach has been shown to be complementary in acquiring biological insight on some risk factors of cardiovascular diseases.16 Thus far, three studies with large sample sizes have reported that a genetically lower VD level could influence the risk of their outcome of interest, i.e., all-cause and cancer mortality,17 ovarian cancer,18 and multiple sclerosis.19 The result of these studies should be evaluated under each study condition and exposure time to outcome. MR studies will continue to demand careful interpretation and results need to be placed in the appropriate biological and clinical context. Future research and RCTs are needed to investigate the critical periods for an optimal VD status and its maintenance throughout pregnancy. Finally, what effects VD has on genetic signatures that minimize the risk to the mother and developing fetus should be elucidated.
References:
1. Magnus MC, Miliku K, Bauer A, et al. Vitamin D and risk of pregnancy related hypertensive disorders: mendelian randomisation study. BMJ 2018;361:k2167. doi: 10.1136/bmj.k2167
2. Mirzakhani H, Litonjua AA, McElrath TF, et al. Early pregnancy vitamin D status and risk of preeclampsia. J Clin Invest 2016;126(12):4702-15. doi: 10.1172/JCI89031
3. De-Regil LM, Palacios C, Lombardo LK, et al. Vitamin D supplementation for women during pregnancy. The Cochrane database of systematic reviews 2016(1):CD008873. doi: 10.1002/14651858.CD008873.pub3
4. Brion MJ, Shakhbazov K, Visscher PM. Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 2013;42(5):1497-501. doi: 10.1093/ije/dyt179
5. Hiraki LT, Major JM, Chen C, et al. Exploring the genetic architecture of circulating 25-hydroxyvitamin D. Genet Epidemiol 2013;37(1):92-8. doi: 10.1002/gepi.21694
6. Jiang X, O'Reilly PF, Aschard H, et al. Genome-wide association study in 79,366 European-ancestry individuals informs the genetic architecture of 25-hydroxyvitamin D levels. Nat Commun 2018;9(1):260. doi: 10.1038/s41467-017-02662-2
7. Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ 2018;362:k601. doi: 10.1136/bmj.k601
8. Nielson CM, Jones KS, Bouillon R, et al. Role of Assay Type in Determining Free 25-Hydroxyvitamin D Levels in Diverse Populations. N Engl J Med 2016;374(17):1695-6. doi: 10.1056/NEJMc1513502
9. Komagata S, Nakajima M, Takagi S, et al. Human CYP24 catalyzing the inactivation of calcitriol is post-transcriptionally regulated by miR-125b. Mol Pharmacol 2009;76(4):702-9. doi: 10.1124/mol.109.056986
10. Zenata O, Vrzal R. Fine tuning of vitamin D receptor (VDR) activity by post-transcriptional and post-translational modifications. Oncotarget 2017;8(21):35390-402. doi: 10.18632/oncotarget.15697
11. Jones G, Prosser DE, Kaufmann M. 25-Hydroxyvitamin D-24-hydroxylase (CYP24A1): its important role in the degradation of vitamin D. Arch Biochem Biophys 2012;523(1):9-18. doi: 10.1016/j.abb.2011.11.003
12. Naruse K, Yamasaki M, Umekage H, et al. Peripheral blood concentrations of adiponectin, an adipocyte-specific plasma protein, in normal pregnancy and preeclampsia. Journal of reproductive immunology 2005;65(1):65-75. doi: 10.1016/j.jri.2004.09.004
13. Ruiz-Ojeda FJ, Anguita-Ruiz A, Leis R, et al. Genetic Factors and Molecular Mechanisms of Vitamin D and Obesity Relationship. Ann Nutr Metab 2018;73(2):89-99. doi: 10.1159/000490669
14. Hollis BW, Wagner CL. New insights into the vitamin D requirements during pregnancy. Bone Res 2017;5:17030. doi: 10.1038/boneres.2017.30
15. Hong X, Hao K, Ji H, et al. Genome-wide approach identifies a novel gene-maternal pre-pregnancy BMI interaction on preterm birth. Nat Commun 2017;8:15608. doi: 10.1038/ncomms15608
16. Thanassoulis G, O'Donnell CJ. Mendelian randomization: nature's randomized trial in the post-genome era. JAMA 2009;301(22):2386-8. doi: 10.1001/jama.2009.812
17. Afzal S, Brondum-Jacobsen P, Bojesen SE, et al. Genetically low vitamin D concentrations and increased mortality: Mendelian randomisation analysis in three large cohorts. BMJ 2014;349:g6330. doi: 10.1136/bmj.g6330
18. Ong JS, Cuellar-Partida G, Lu Y, et al. Association of vitamin D levels and risk of ovarian cancer: a Mendelian randomization study. Int J Epidemiol 2016;45(5):1619-30. doi: 10.1093/ije/dyw207
19. Mokry LE, Ross S, Ahmad OS, et al. Vitamin D and Risk of Multiple Sclerosis: A Mendelian Randomization Study. PLoS Med 2015;12(8):e1001866. doi: 10.1371/journal.pmed.1001866
Corresponding author:
Hooman Mirzakhani, M.D., MMSc., Ph.D.
Instructor in Medicine
Harvard Medical School
Brigham & Women’s Hospital
Department of Medicine, Channing Division of Network Medicine
181 Longwood Avenue, 530
Boston, MA, 02115, USA
Email: hooman.mirzakhani@channing.harvard.edu
Competing interests:
No competing interests
04 September 2018
Hooman Mirzakhani
Physician-scientist
Scott T. Weiss, MD, MS
Brigham and Women's Hospital-Harvard Medical School
Channing Division of Network Medicine, 181 Longwood Ave, 450, Boston, MA, 02115 USA
Rapid Response:
Mendelian Randomization of Vitamin D for Preeclampsia: Embracing Complexity of Inference
Mendelian Randomization of Vitamin D for Preeclampsia: Embracing Complexity of Inference
Hooman Mirzakhani, M.D., MMSc., Ph.D.1; Scott T. Weiss, M.D., M.S.1,2
1Channing Division of Network Medicine, Department of Medicine, Brigham and Women's
Hospital, Harvard Medical School, Boston, MA, USA
1,2Partners Center for Personalized Medicine, Partners Health Care, Boston, MA, USA
[A letter to the editor regarding Mangus et al. Vitamin D and risk of pregnancy-related hypertensive disorders: mendelian randomisation study, BMJ 2018;361:k2167]; Word count: 1000
We read with interest the study by Mangus and colleagues who applied Mendelian randomization (MR) to examine the causal nature of the association between 25-hydroxyvitamin D (25OHD) and pregnancy-induced hypertension.1 We think that the accuracy and relevance of the results of MR studies should be interpreted in light of other sources of evidence and biological knowledge of this exposure-outcome relationship.
The individual result from a few vitamin D (VD) supplementation RCTs during pregnancy has not been supportive of VD supplementation to prevent preeclampsia and gestational hypertension. These studies varied in the dose (400–4,000 IU/day), time of initiation and in the baseline and attained VD levels. Except for one trial, all of the RCTs started the intervention after 20 weeks of gestation, the time that the two conditions generally become symptomatic.2 A recent systematic review by the Cochrane database on VD supplementation during pregnancy indicated a trend toward the reduction of preeclampsia risk in pregnant women supplemented with VD compared to no supplementation.3 The question remains whether we can expect to obtain further applicable insight from MR approaches or well-designed RCTs based on the prior knowledge are more applicable to conclusively investigate a preventive role of VD for adverse pregnancy outcomes.
To answer this question, we should first assess the plausibility of the MR assumptions in this context. The first assumption is the presence and strength of the relationship of VD variants with 25OHD levels. For continuous variables and a single instrument, statistical power for a fixed sample size is a function of the proportion of variation in the 25OHD levels explained by the variants and the magnitude of the association between the VD level and preeclampsia.4 According to GWAS studies, the identified SNPs explain less than 5% of the 25OHD varitaiton.5 6 Given this variation, a very large sample size is required to achieve sufficient statistical power for revealing 25OHD causal effects. This variability was less than 1.3% in Mangus and colleagues’ study. Thus, this report exemplifies a weak instrument bias due to using one or more genetic variants that only explain a small proportion of the variation in the risk factor, coupled with a small sample size.7 Preeclampsia has a heterogeneous phenotype with a low incidence in pregnancy (3-8% ) less than other common diseases such as cardiovascular disease. Therefore, the 25OHD effect size varies in different studies. This fact should also be considered in sample size calculation.
Another key assumption is the ‘no horizontal pleiotropy’, which requires that the VD SNPs acts on preeclampsia exclusively through the 25OHD. A violation of this principle can distort the result and obscure the conclusiveness. Some of the SNPs associated with 25OHD levels also influ¬ence VD binding protein (DBP) concentration and affinity for 25OHD; these SNPs might alter bioavailable 25OHD independently of total 25OHD concentration.8 Moy and colleagues showed that associated GC variants (including rs2282679 in Mangus and colleagues’ study) with lower 25OHD concentrations were strongly related to lower levels of DBP. Another finding is by Komangata and colleagues who showed the post-transcriptional regulation of CYP24A1 by miR-125b such that its increased expression resulted in the reduction of CYP24 protein and CYP24-mediated enzymatic activity.9 10. CYP24A1 encodes 24-hydroxylase, which initiates the degradation of 25OHD and 1,25(OH)2D.11 Furthermore, the adipocytokine profile has been shown to be different in preeclampsia and VD status in relation to VD genes.12 13 These observations potentially violate the second MR assumption in such studies. Additionally, VD is known to initiate down-stream expression of literally hundreds of genes making the use of a few SNPs a poor proxy for levels let alone the downstream effects of these levels.
Most GWAS studies have been conducted in non-pregnant individuals. Pregnancy is characterized by profound metabolic alterations to ensure adequate maternal-fetal physiological demands. VD metabolism is unique during pregnancy such that the conversion of 25OHD to 1,25(OH)2D in both mother and or fetus is not controlled by classic calcium homeostatic mechanisms. By 12 weeks of gestation, 1,25(OH)2D serum concentrations are more than twice that of a non-pregnant woman and continue to rise two- to threefold from the non-pregnant baseline.14 While, the heritability of maternal metabolic traits during pregnancy has not been examined, there is a greater possibility of gene-environment (e.g., pre-pregnancy BMI15) or gene-gene interactions. Furthermore, most instrumental variable methods assume a linear relationship between exposure and outcome of interest. As suggested by recent studies, it is reasonable to also examine a non-linear relationship of 25OHD with preeclampsia.2 14
MR studies cannot provide information on the dose-response relationship or the timing and duration of VD intervention. This information can only be obtained when the instrumental variable explains a high degree of variability in the exposure of interest throughout the lifespan. A defined optimal serum level of VD during pregnancy is desirable and has been suggested, but requires to be validated.5 More importantly, there might be a critical period for the vitamin D’s role in the prevention of preeclampsia at a preclinical stage of preeclampsia.2 After preeclampsia initiation, the progression to the clinical stage might be further under the influence of other factors. In a well-designed controlled clinical trial, the above factors can be considered, and relative influence of various VD SNPs on dose-response effect and the outcome further explored.
The MR approach has been shown to be complementary in acquiring biological insight on some risk factors of cardiovascular diseases.16 Thus far, three studies with large sample sizes have reported that a genetically lower VD level could influence the risk of their outcome of interest, i.e., all-cause and cancer mortality,17 ovarian cancer,18 and multiple sclerosis.19 The result of these studies should be evaluated under each study condition and exposure time to outcome. MR studies will continue to demand careful interpretation and results need to be placed in the appropriate biological and clinical context. Future research and RCTs are needed to investigate the critical periods for an optimal VD status and its maintenance throughout pregnancy. Finally, what effects VD has on genetic signatures that minimize the risk to the mother and developing fetus should be elucidated.
References:
1. Magnus MC, Miliku K, Bauer A, et al. Vitamin D and risk of pregnancy related hypertensive disorders: mendelian randomisation study. BMJ 2018;361:k2167. doi: 10.1136/bmj.k2167
2. Mirzakhani H, Litonjua AA, McElrath TF, et al. Early pregnancy vitamin D status and risk of preeclampsia. J Clin Invest 2016;126(12):4702-15. doi: 10.1172/JCI89031
3. De-Regil LM, Palacios C, Lombardo LK, et al. Vitamin D supplementation for women during pregnancy. The Cochrane database of systematic reviews 2016(1):CD008873. doi: 10.1002/14651858.CD008873.pub3
4. Brion MJ, Shakhbazov K, Visscher PM. Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 2013;42(5):1497-501. doi: 10.1093/ije/dyt179
5. Hiraki LT, Major JM, Chen C, et al. Exploring the genetic architecture of circulating 25-hydroxyvitamin D. Genet Epidemiol 2013;37(1):92-8. doi: 10.1002/gepi.21694
6. Jiang X, O'Reilly PF, Aschard H, et al. Genome-wide association study in 79,366 European-ancestry individuals informs the genetic architecture of 25-hydroxyvitamin D levels. Nat Commun 2018;9(1):260. doi: 10.1038/s41467-017-02662-2
7. Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ 2018;362:k601. doi: 10.1136/bmj.k601
8. Nielson CM, Jones KS, Bouillon R, et al. Role of Assay Type in Determining Free 25-Hydroxyvitamin D Levels in Diverse Populations. N Engl J Med 2016;374(17):1695-6. doi: 10.1056/NEJMc1513502
9. Komagata S, Nakajima M, Takagi S, et al. Human CYP24 catalyzing the inactivation of calcitriol is post-transcriptionally regulated by miR-125b. Mol Pharmacol 2009;76(4):702-9. doi: 10.1124/mol.109.056986
10. Zenata O, Vrzal R. Fine tuning of vitamin D receptor (VDR) activity by post-transcriptional and post-translational modifications. Oncotarget 2017;8(21):35390-402. doi: 10.18632/oncotarget.15697
11. Jones G, Prosser DE, Kaufmann M. 25-Hydroxyvitamin D-24-hydroxylase (CYP24A1): its important role in the degradation of vitamin D. Arch Biochem Biophys 2012;523(1):9-18. doi: 10.1016/j.abb.2011.11.003
12. Naruse K, Yamasaki M, Umekage H, et al. Peripheral blood concentrations of adiponectin, an adipocyte-specific plasma protein, in normal pregnancy and preeclampsia. Journal of reproductive immunology 2005;65(1):65-75. doi: 10.1016/j.jri.2004.09.004
13. Ruiz-Ojeda FJ, Anguita-Ruiz A, Leis R, et al. Genetic Factors and Molecular Mechanisms of Vitamin D and Obesity Relationship. Ann Nutr Metab 2018;73(2):89-99. doi: 10.1159/000490669
14. Hollis BW, Wagner CL. New insights into the vitamin D requirements during pregnancy. Bone Res 2017;5:17030. doi: 10.1038/boneres.2017.30
15. Hong X, Hao K, Ji H, et al. Genome-wide approach identifies a novel gene-maternal pre-pregnancy BMI interaction on preterm birth. Nat Commun 2017;8:15608. doi: 10.1038/ncomms15608
16. Thanassoulis G, O'Donnell CJ. Mendelian randomization: nature's randomized trial in the post-genome era. JAMA 2009;301(22):2386-8. doi: 10.1001/jama.2009.812
17. Afzal S, Brondum-Jacobsen P, Bojesen SE, et al. Genetically low vitamin D concentrations and increased mortality: Mendelian randomisation analysis in three large cohorts. BMJ 2014;349:g6330. doi: 10.1136/bmj.g6330
18. Ong JS, Cuellar-Partida G, Lu Y, et al. Association of vitamin D levels and risk of ovarian cancer: a Mendelian randomization study. Int J Epidemiol 2016;45(5):1619-30. doi: 10.1093/ije/dyw207
19. Mokry LE, Ross S, Ahmad OS, et al. Vitamin D and Risk of Multiple Sclerosis: A Mendelian Randomization Study. PLoS Med 2015;12(8):e1001866. doi: 10.1371/journal.pmed.1001866
Corresponding author:
Hooman Mirzakhani, M.D., MMSc., Ph.D.
Instructor in Medicine
Harvard Medical School
Brigham & Women’s Hospital
Department of Medicine, Channing Division of Network Medicine
181 Longwood Avenue, 530
Boston, MA, 02115, USA
Email: hooman.mirzakhani@channing.harvard.edu
Competing interests: No competing interests