HuGENet™ Case Study
Glucocerebrosidase Gene Mutations and Parkinson Disease
What other information about the cases and controls would be important to know? Are there any specific concerns about the healthy controls selected for this study?
A common deficiency in case-control studies of gene-disease associations is the lack of comparability of cases and controls (i.e., controls may not come from the same population from which cases were selected). This could lead to possible selection bias and inability to generalize from the results. In this study a significant deficiency in the provided information is the age of the study groups. The healthy controls were ascertained from persons undergoing DNA mutation testing for certain recessive genetic disorders that are more prevalent in Ashkenazi Jews (Tay-Sachs disease, cystic fibrosis, etc.); most persons seeking such a test would do so because they are making reproductive plans and would like to assess their risk of having a child with one of the recessive conditions. Therefore, the healthy controls are most likely single persons planning a marriage or couples planning a pregnancy; persons in this category will be much younger than the typical person in the Parkinson Disease group. Some persons in the healthy control group may develop Parkinson Disease when they are older, and, therefore, would not be appropriate for inclusion in the control group. It would also have been helpful to know the gender distribution of the healthy controls in order to confirm that the study groups have been comparable in this respect.
Are the observed carrier frequencies for the N370S and 84GG mutations significantly different from the carrier estimates reported in the literature?
For the N370S mutation, the observed carriers and noncarriers were 92 and 1451 (carrier frequency of 1 in 16.7). For the previously reported carrier frequency of approximately 1 in 17.5, the expected number of carriers and noncarriers in the control group would be approximately 88 and 1455, respectively. The observed and expected are compared by the Yates Chi-Square, corrected for continuity:
Chi-square = 0.05, df = 1, p = 0.8231
Thus, the observed and expected carrier frequencies for the N370S mutation are not significantly different. For the 84GG mutation, the observed carriers and noncarriers were 3 and 1540 (carrier frequency of approximately 1 in 514). For the previously reported carrier frequency of 1 in 400, the expected number of carriers and noncarriers in the control group would be approximately 3.9 and 1539.1, respectively. The observed and expected are shown in a 2 x 2 table:
Note that for a 2 x 2 table, the Yates Chi-Square test is valid only if all expected cell frequencies are equal to or greater than 5. Since this requirement is not met for the data shown here, the Pearson Chi-Square, uncorrected for continuity is calculated: Chi-square = 0.14, df = 1, p = 0.7083
The observed and expected carrier frequencies for the 84GG mutation are also not significantly different.
Is the definition that the authors proposed for a “carrier” for Gaucher disease appropriate for this analysis?
By grouping together homozygotes, heterozygotes, and compound heterozygotes for GBA mutations, the authors perhaps have assumed that the risk of developing Parkinson Disease for persons with each genotype is similar. However, this assumption is not valid. In fact, because the N370S mutation is considered a “mild” mutation, while the 84GG is considered a “severe” mutation, carriers for these mutations may have quite different risks for developing Parkinson Disease, depending upon their genotypes. In addition, even though the proportion of homozygous persons in the Parkinson Disease group is only 9.7% (3 N370S homozygotes divided by 31 total “carriers”), the risk for homozygotes or compound heterozygotes to develop Parkinson Disease may be much higher than among simple heterozygotes, which could influence the overall risk assessment for “carriers.” Therefore, by grouping all genotypes with GBA mutations into a single category it is not possible to determine if the genotype-specific risks for developing Parkinson Disease are different, each with perhaps quite different clinical and public health consequences.
Based on these findings, what is the odds ratio (estimated relative risk) for Parkinson Disease among “carriers” of Gaucher disease? Is the rate of “carriage” of a GBA mutation among patients with Alzheimer's disease increased compared with healthy controls?
Odds ratios are calculated from the 2 x 2 tables as follows:
|Parkinson Disease||Alzheimer's Disease||Odds Ratio||Parkinson Disease||Healthy Controls||Odds Ratio|
The authors concluded that patients with Parkinson Disease had significantly greater odds for being “carriers” of Gaucher disease than did patients with Alzheimer's disease (odds ratio, 10.8; 95% confidence interval, 3.0-46.6; p<0.001) or control subjects (odds ratio, 6.9; 95% confidence interval, 4.2-11.4; p<0.001).
The odds ratio for GBA mutation in Alzheimer's disease was:
|Alzheimer's Disease||Healthy Controls||Odds Ratio|
At first glance, finding an odds ratio less than one might suggest that “carriers” for a GBA gene mutation have a lower risk for developing Alzheimer's disease. However, the rate of “carriage” of a GBA gene mutation among patients with Alzheimer's disease did not differ significantly from that in healthy controls (95% confidence interval 0.2-2.2, p=0.62).
What additional analysis of the data should be performed in order to determine the validity of the authors' second conclusion?
The authors' second conclusion is not rigorously supported unless an analysis is performed to evaluate the effects of specific alleles, or the contribution of homozygosity for a mutant allele versus heterozygosity, on the risk for developing Parkinson Disease in this patient population. Although the authors chose not to do this analysis, perhaps because of small sample sizes, a determination of the odds ratios for developing Parkinson Disease among persons with each GBA genotype should be performed since the results will be helpful for determining the risk for persons with each genotype of developing Parkinson Disease, relative to the prevalence of Parkinson Disease in the population.
Since there were no N370S homozygotes among the healthy controls, how can the odds ratio be calculated for this genotype?
One way to obtain the expected number of N370S homozygotes among the controls is to assume that the GBA alleles in the control population are in Hardy-Weinberg equilibrium. The calculated expected number of controls that are homozygous for the N370S mutation is 1.3713, which is then used to estimate the odds ratio.
Are the observed numbers for the genotypes that comprise the healthy controls significantly different from Hardy-Weinberg equilibrium? Is it valid to have used 1.3713 as the number of healthy controls with the N370S/N370S genotype in order to calculate the odds ratio?
The numbers of observed and expected controls with each genotype can be compared by the Chi-square test. The result for these data is: Chi-square = 1.5566; df = 3; p = 0.500-0.750. Based on Hardy-Weinberg equilibrium, the observed frequencies for each genotype in the healthy controls do not differ significantly from those expected for all three alleles: +, N370S, and 84GG. Therefore, it is valid to utilize the expected number of controls for the N370S/N370S genotype in order to calculate the odds ratio.
How strong is the association between each of the GBA gene genotypes and Parkinson Disease?
All calculated odds ratios were statistically significant, with p values < 0.001. The analyses showed that (1) N370S homozygotes were approximately 47 times more likely than people without this genotype to have Parkinson Disease, but accounted for only 3 of 98 cases, (2) N370S heterozygotes were approximately 5 times more likely to have Parkinson Disease, and (3) 84GG heterozygotes were approximately 28 times more likely to have Parkinson Disease, but accounted for only 4 out of 98 cases. Therefore, of all of the “carrier” classes defined by the authors, the most important contribution at the population level to risk for developing Parkinson Disease comes from the N370S heterozygotes which are 5 times more likely to have Parkinson Disease.
Are the results of this analysis valid? If the result is correct, what is then suggested about the role of the GBA gene genotypes in Parkinson Disease at the population level?
The Parkinson Disease and healthy control study groups were all of Ashkenazi Jewish heritage and presumably sampled from the same population (by referral to a tertiary medical center for northern Israel ). To estimate attributable fraction for the Ashkenazi Jewish population, we would have to assume that the cases are representative of all Ashkenazi Jewish persons with Parkinson Disease and that the controls are also a representative (unbiased) sample of the Ashkenazi Jewish population. The authors did not present the age distribution of the Parkinson Disease patients or the age or gender distribution of healthy controls, who were probably much younger on average than the Parkinson Disease patients, a factor that would tend to bias the study against finding an association with GBA genotype. If results of the analysis are correct, homozygosity for N370S and heterozygosity for 84GG are strong risk factors for Parkinson Disease but account for only a small proportion of cases. Heterozygotes for N370S were much more prevalent than homozygotes among cases, with an attributable fraction of almost 20%. Although all the studied GBA genotypes taken together accounted for 26% of the Parkinson Disease cases, most people with Parkinson Disease in the study population did not have one of the 6 analyzed mutations in the GBA gene.
Based on these genotype frequencies, and assuming that the total population risk of Parkinson Disease is 1 per 100 people (1%), what is the estimated absolute risk of Parkinson Disease by genotype?
Assume that the odds ratio is an approximation of the relative risk, and that the total population risk is 0.01. The population wide incidence (sum of the genotype specific incidence rates) and the absolute risk are calculated by summing the products of the genotype frequencies and the odds ratios, where “x” is the absolute risk of Parkinson Disease in people without any of the GBA gene mutations (wild type) :
Prevalence = 0.01 = 1.3426x
x = 0.00744
The risk for developing Parkinson Disease for persons with each genotype is then determined by multiplying the baseline absolute risk (x) by the odds ratio:
|Genotype||Incidence/1000||Risk of Developing Parkinson Disease With the Genotype|
Would it beneficial to perform population-based screening of Ashkenazi Jews for the N370S mutation--the mutation with the highest incidence in the population--in order to identify those persons with an increased risk of developing Parkinson Disease?
Screening 1000 Ashkenazi Jews will identify approximately 59 persons who are N370S heterozygotes or homozygotes:
(1000)(0.05778) + (1000)(.00089)
57.8 + 0.9 = approximately 59 persons
Based on the risks that persons with each genotype will develop Parkinson Disease, of the 59 identified persons, approximately 3 would be expected to develop the condition:
(1000)(.05778)(0.040) + (1000)(.00089)(0.347)
2.31 + 0.31 = approximately 3 persons
Therefore, 95% of identified persons with the N370S mutation are not expected to develop Parkinson Disease. This finding is in contrast to the authors' conclusion that “mutations in the GBA gene emerge as a strong genetic determinant predisposing people to Parkinson Disease.” This statement may relate to the fact that mutations in the GBA gene, particularly in the Ashkenazi Jewish population, are much more common than mutations in other genes implicated in the genetic susceptibility to Parkinson Disease (parkin, SNCA, SNCAIP, etc.). However, the individual risk of developing Parkinson Disease for carriers of a mutation in the GBA gene (particularly the N370S mutation), is still very low. Because there is no way to distinguish the N370S heterozygotes or homozygotes who will develop Parkinson Disease from those who won't, and there are no clear lifestyle modifications or medical interventions to lower the risk, N370S carrier identification would cause needless anxiety about the uncertain likelihood of developing a major future health problem without providing a clear health benefit.
- Page last reviewed: June 15, 2009
- Page last updated: March 9, 2010
- Content source: