Free
Genetics  |   October 2014
Direct-to-Consumer Personal Genome Testing for Age-Related Macular Degeneration
Author Affiliations & Notes
  • Gabriëlle H. S. Buitendijk
    Department of Ophthalmology, Erasmus Medical Center, Rotterdam, The Netherlands
    Department of Epidemiology, Erasmus Medical Center, Rotterdam, The Netherlands
  • Najaf Amin
    Department of Epidemiology, Erasmus Medical Center, Rotterdam, The Netherlands
  • Albert Hofman
    Department of Epidemiology, Erasmus Medical Center, Rotterdam, The Netherlands
    Netherlands Consortium for Healthy Aging, Netherlands Genomics Initiative, The Hague, The Netherlands
  • Cornelia M. van Duijn
    Department of Epidemiology, Erasmus Medical Center, Rotterdam, The Netherlands
  • Johannes R. Vingerling
    Department of Ophthalmology, Erasmus Medical Center, Rotterdam, The Netherlands
    Department of Epidemiology, Erasmus Medical Center, Rotterdam, The Netherlands
  • Caroline C. W. Klaver
    Department of Ophthalmology, Erasmus Medical Center, Rotterdam, The Netherlands
    Department of Epidemiology, Erasmus Medical Center, Rotterdam, The Netherlands
  • Correspondence: Caroline C.W. Klaver, Department of Ophthalmology, Erasmus Medical Centre, P.O. Box 2040, NL-3000 CA Rotterdam, The Netherlands; c.c.w.klaver@erasmusmc.nl
Investigative Ophthalmology & Visual Science October 2014, Vol.55, 6167-6174. doi:10.1167/iovs.14-15142
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Gabriëlle H. S. Buitendijk, Najaf Amin, Albert Hofman, Cornelia M. van Duijn, Johannes R. Vingerling, Caroline C. W. Klaver; Direct-to-Consumer Personal Genome Testing for Age-Related Macular Degeneration. Invest. Ophthalmol. Vis. Sci. 2014;55(10):6167-6174. doi: 10.1167/iovs.14-15142.

      Download citation file:


      © 2017 Association for Research in Vision and Ophthalmology.

      ×
  • Supplements
Abstract

Purpose.: Genetic testing may be the next step in clinical medicine for a more personalized approach in determining risk of disease. Direct-to-consumer (DTC) personal genome tests may fulfill this role. We explored the practicability and predictive value of DTC tests from four companies (23andMe, deCODEme, Easy DNA, Genetic Testing Laboratories) for AMD.

Methods.: Body specimens of three individuals were collected and sent to four companies for DNA genotyping and disease risk estimation. In addition, DNA was also genotyped using Illumina HumanOmniExpress 12v1 array in the Rotterdam Study laboratory, and risk estimates of AMD were calculated using the validated prediction model from the population-based Three Continent AMD Consortium.

Results.: Genotyped results of the four DTC tests matched genotyping performed by the Rotterdam Study laboratory. The estimated risks provided by the companies varied considerably in the tested individuals, from a 1.6-fold difference for overall relative risk to an up to 12-fold difference for lifetime risk. The lifetime risks for the individuals ranged from 1.4% to 16.1% in the DTC tests, while they varied from 0.5% to 4.2% in the validated prediction model. Most important reasons for the differences in risks were the testing of only a limited set of genetic markers, the choice of the reference population, and the methodology applied for risk calculation.

Conclusions.: Direct-to-consumer personal genome tests are not suitable for clinical application as yet. More comprehensive genetic testing and inclusion of environmental risk factors may improve risk prediction of AMD.

Introduction
Genetic studies of AMD have elucidated a major proportion of its genetic background. Currently, genome-wide studies (GWAS) have identified associations with greater than 30 genetic loci for this disease, explaining a large part of the heritability of AMD (Fritsche LG, et al. IOVS 2014;55:ARVO E-Abstract 4570). 1 Subsequently, these genomic findings have been incorporated into prediction models, many of which provide a greater than 80% discriminative accuracy for late AMD. 221 This high predictive ability makes AMD particularly suitable for genetic testing, which may be the next step to a more personalized approach in clinical medicine. 
Direct-to-consumer (DTC) personal genome tests had been made available for consumers and thousands have purchased these tests via the internet to determine a personal disease risk. Recently, methods of three DTC-tests have been examined and compared for several diseases. 22 Age-related macular degeneration was the disease for which each test obtained the best predictive ability. Several companies offered genetic tests for AMD and implementation of these tests in the clinic could help identify individuals at risk of developing the disease to apply risk-dependent patient care and surveillance strategies. Therefore, the accuracy of the risk estimates will be a great concern, and will determine whether such tests will be meaningful in the clinic. 
In this study, we evaluated the results of AMD prediction tests provided by four major companies. We sent biosamples from three individuals to these companies to test proof of principle, and reviewed the sampling process, the type of analysis, the genotyping, and the risk information. In addition, we compared results with a validated prediction model based on population studies. 
Methods
Experimental Design
Evaluation of test methodology. 
Study Participants
Three investigators (GHSB, JRV, CCWK) agreed to voluntarily participate in the study, and signed informed consent. The study was adhered to the tenets of the Declaration of Helsinki. 
DTC-Tests for AMD
We searched for internet-based DTC tests for AMD using a web search engine and the word groups “genetic testing for age-related macular degeneration,” “genetic prediction of age-related macular degeneration,” and “genetic tests for age-related macular degeneration.” Only companies available for European citizens and testing more than one single nucleotide polymorphism (SNP) were eligible, and of these, four companies were selected: 23andMe, deCODEme, Easy-DNA, and Genetic Testing Laboratories. 
23andme.
https://www.23andme.com/ (in the public domain). 
This privately-held American company was founded in 2006 with the intention to empower individuals in accessing their own genetic information and to stimulate a way into more personalized medicine. One can order a single ‘spit' kit for $99 (shipping costs of $14.95–$118.95) from the website on the internet, and a sample collection kit will be sent by mail, with instructions how to provide a saliva sample, and details for returning the sample. An assisted collection kit for persons having trouble to spit can be ordered together with the DTC kit for an additional $25, requiring only one-half the amount of saliva. The returned saliva sample will arrive at the contracted LabCorp's Clinical Laboratory Improvement Amendments (CLIA) certified laboratory, where DNA will be isolated from cells in the saliva and processed on an Illumina HumanOmniExpress array customized by 23andMe (>1 million SNPs, call rate above 98%). These SNPs provide information about traits, carrier status, and risks for over 100 diseases, including AMD. The risk for developing AMD is estimated based on the risk in the reference population and an overall relative risk (RR) representing risks of five SNPs: CFH rs1061147; C2 rs547154; LOC387715/ARMS2 rs3750847; C3 rs2230199; TIMP3 rs9621532. 10,2335 Risk of AMD in the reference population differed for males and females and was 6.5% and 7%, respectively. Methods of risk calculation have been described in a white paper, 36 accessible after login to the 23andMe website. No health reports including risk prediction and carrier status are currently provided for new customers. 
DeCODE.
http://www.decodeme.com/ (in the public domain). 
DeCODE was founded in 1996 and the headquarters are located in Reykjavik, Iceland. This company developed the deCODEme test, which provides results for 47 conditions and traits. Unfortunately, new tests are no longer offered by the company. Costs were $1100 per test, with no extra costs for shipping. After purchasing the test from the internet, a buccal swab kit was sent in the mail with instructions on how to collect and return the sample. The samples were processed at a CLIA certified lab, the deCODE laboratory in Reykjavik, for DNA isolation. Genotyping was performed on an Illumina Human 1M Beadchip (Illumina, Inc., San Diego, CA, USA), which determines more than 1 million SNPs. Validation occurred by bidirectional Sanger sequencing and independent SNP genotyping platforms. 
An overall RR for developing AMD was calculated based on six risk variants: ARMS2/HTRA1 rs3750847, C2/FB rs9332739 and rs547154, C3 rs230199, and CFH rs1061147 and rs1329428. 26,37 Subsequently, for the tested individual a lifetime risk was calculated based on the overall RR and the AMD risk in the reference population, which was set at 8%. A white paper 38 describing the risk calculation is available after login to the deCODEme website. 
Easy-DNA.
Easy-DNA is an international company, which provides a genetic DNA predisposition test on 25 conditions and diseases. This test can be purchased from the internet for €299/$299/£299, respectively, including shipping costs. A kit will be sent by mail for collection of a blood sample, and includes submission forms, instructions for collecting the blood sample from a punctured finger, the sample collection kit, and a self-addressed envelope. This company does not provide information on the genotyping method, but states that results are provided for CFH rs1061170 and C2 rs800292. 39,40 Risk estimates are presented as lifetime and overall RR of AMD. Risk of AMD in the reference population was set at 8%. Methods for risk calculation was not provided by the company. 
Genetic Testing Laboratories (GTL).
This company provides a DNA predisposition test, which will reveal the genetic and environmental predisposition for 25 diseases and conditions including AMD. The DNA predisposition test costs $285 with additional costs of $45 for shipping outside the Contiguous United States. After purchasing the kit from the internet, it will be sent to your own physician or a professional collector agency appointed by GTL to collect the sample, which can be a buccal or a blood sample. The sample will be processed by a CLIA accredited laboratory. As for Easy-DNA, this company also is unclear on genotyping method, but states that results are provided for CFH rs1061170 and C2 rs800292. 39,40 Lifetime and overall RR are provided for each tested person. Risk of AMD in the reference population was set at 8%.The risk calculation method of this company was not available for consumers or professionals. 
We followed each company's instructions for the collection of biosamples used for DNA extraction. We sent the samples to the various laboratories associated with the companies, and awaited the results. 
Genotyping in Rotterdam
Genotyping for the three individuals was also performed at the Rotterdam Study Laboratory: Genetic Laboratory of Internal Medicine at the Erasmus Medical Center in Rotterdam, The Netherlands. Genomic DNA was extracted from peripheral leukocytes and all participants were genotyped using the Illumina HumanOmniExpress 12v1_J microarray (Illumina, Inc.). Call rate for the genotyping was greater than 97.5%.We imputed genotype data to Hapmap 3 release 2 and 1000 genomes phase I V3. 
Assessment of Covariates
The covariates age, length, weight, smoking status, and family history regarding AMD were obtained by interview. Body mass index (BMI) was calculated by dividing weight (kg) by the height squared (m2). Age-related macular degeneration phenotype was evaluated by standard ophthalmologic examination including fundus photography (Topcon TRC-50EX fundus camera; Topcon Optical Co., Tokyo, Japan; and Sony DXC-950P digital camera; Sony Corporation, Tokyo, Japan) after pharmacological mydriasis. Images were graded according to the Wisconsin Age-Related Maculopathy Grading 41 and the modified international classification system 42 by graders from the Rotterdam Study. 
Risk Score Three Continent AMD Consortium Prediction Model and DTC-Tests
The Three Continent AMD Consortium (3CC) developed a validated prediction model including a total risk score based on 31 variables; 26 genetic variants associated with AMD, age, sex, smoking, BMI, and AMD phenotype. The prediction model had 87% discriminative accuracy for incident late AMD. 21 For each individual in this study this summary risk score was calculated. Based on the risk score, lifetime risks could be assessed for each individual. 
Ancestry Assessment
Ancestry of the three individuals was determined using multidimensional scaling (MDS) protocol from ENIGMA 43 using Hapmap 3 release 2 as the reference. 
Statistical Analysis
Test results included predicted risks for several diseases from four companies. For the purpose of this study, we only evaluated the predicted risks for AMD. 23andMe provided odds ratios (OR) and the other companies RRs per SNP per genotype, but all were adjusted for the average risk of the SNP in the population, and will be referred to as OR and RR, respectively. Genotype frequency, risks per genotype, overall RR, lifetime population risk, and lifetime risk of the tested individual were obtained from the test results. 
Minor allele frequencies were not provided by the companies, but calculated using the formula    
with p representing the major allele and q the minor allele. For the different genotypes, frequencies could be calculated after applying this information; homozygous for major alleles = p 2, heterozygous = 2pq and homozygous for minor alleles = q 2
All analyses were performed using SPSS version 20.0 (SPSS, Inc., Chicago, IL, USA) except for the MDS-analysis, which was performed using R software. 44  
Results
Demographic characteristics of the three study subjects are provided in Table 1. All three were younger than the average age of AMD onset, and none had any features of AMD, as determined by grading of fundus photographs. One had a history of smoking, and one had a positive family history for late AMD. All three were Caucasian and had Northern/Western European ancestry (Supplementary Fig. S1). 
Table 1
 
Descriptives of the Participants
Table 1
 
Descriptives of the Participants
Variable Individual 1 Individual 2 Individual 3
Age, y 45 29 51
Sex Female Female Male
Ethnicity Caucasian Caucasian Caucasian
Ancestry Northern/Western European Northern/Western European Northern/Western European
BMI, kg/m2 22.7 20.2 24.3
Smoking Never Never Past
AMD phenotype None None None
Family history of AMD Grandmother None None
DTC Tests
Details of the DTC tests are given in Table 2. Tests differed considerably in price, the most costly being 11× more expensive than the cheapest test. Sampling methods varied from saliva, to buccal swap, to blood from a finger prick. One participant particularly had difficulty to deliver the saliva specimen of 2.5 mL for 23andMe, which required approximately 1 hour of sampling time. Genetic Testing Laboratories required for all participants and Easy-DNA only for US residents a physician or another health professional assigned by the company to collect the blood sample and only the collectors obtained the test results. However, the forms for requesting the test from GTL were open access. Delivery time for test results ranged from 2 to 4 weeks for most tests; results from one Easy-DNA test were delayed up to 8 weeks without notice or explanation. 
Table 2
 
Overview Genetic Testing Companies
Table 2
 
Overview Genetic Testing Companies
Company Name Website Costs Per Kit DNA Source Easy To Collect? Additional Notes
23andMe https://www.23andme.com $99/€74 Saliva Difficult in  1 participant Street address is needed to deliver DTC-test
deCODEme* https://www.decodeme.com $1100/€821 Buccal Yes
Easy-DNA http://www.easygenetictest.com $299/€299 Blood Yes For US residents: sample needs to be collected by physiscian or professional collector
Genetic Testing Laboratories http://www.gtldna.com/ $285/€213 Blood Yes Sample needs to be collected by physiscian or professional collector
In contrast to the statement of Easy-DNA and GTL, the SNP rs800292 is located in the CFH gene, not in C2 (Table 3). Thus, these two companies only tested risk variants in CFH. DeCODEme and 23andMe covered four and five AMD loci, respectively. The tested SNPs varied among tests, however, there was considerable overlap. Individual genotypes at these SNP locations are shown in Table 3. Risk increasing as well as decreasing variants were present in all three individuals. The effect estimates of these variants showed the largest range in individual 2, in particular for the risks predicted by 23andMe and deCODEme. The lifetime AMD population risk used by the companies varied from 6.5% to 8%, and varied for sex in the 23andMe calculations. For 23andMe and deCODEme the ancestry of the reference populations was European, for GTL and Easy-DNA this was European Tuscan. Only for individual 1 did the Easy-DNA test list European ancestry as the reference population. Genotypes identified by the DTC tests were identical to those determined at the Rotterdam Study laboratory in all three individuals. 
Table 3
 
Risks of the Tested Variants, Overall Risk, and Lifetime Risk Per Company for Each Individual
Table 3
 
Risks of the Tested Variants, Overall Risk, and Lifetime Risk Per Company for Each Individual
AMD Individual 1
Gene SNP Number 23andMe deCODEme Easy-DNA* GTL
Genotype OR Genotype RR Genotype RR Genotype RR
CFH rs1061147 AC 0.97 AC 1.56||
CFH rs1329428 GG
CFH rs1061170 CT 1.26 CT 1.60
CFH rs800292 CC 0.67 CC 0.63
C2 rs547154 GG 1.07 CC 1.10
C2 rs9332739 GG 1.06
LOC387715/ARMS2 rs3750847 CC 0.47 GG 0.46
C3 rs2230199 CG 1.37 CG 1.29
TIMP3 rs9621532 AA 1.02
Overall RR‡ 0.70 1.01 0.85 1.00
Lifetime population risk, % 7.0 8.0 8.0 8.0
Lifetime risk, %§ 4.9 8.6 6.8 8.1
Table 3
 
Extended
Table 3
 
Extended
Individual 2 Individual 3
23andMe deCODEme Easy-DNA GTL 23andMe deCODEme Easy-DNA GTL
Genotype OR Genotype RR Genotype RR Genotype RR Genotype OR Genotype RR Genotype RR Genotype RR
CC 0.34 CC 0.21|| AC 0.97 AC 1.56||
AA GG
TT 0.64 TT 0.64 CT 1.60 CT 1.60
CT 1.26 CT 1.26 CT 1.26 CT 1.26
GG 1.07 CC 1.10 GT 0.57 AC 0.58
GG 1.06 GG 1.06
CT 1.63 AG 1.59 CC 0.47 GG 0.46
CG 1.37 CG 1.29 GG 0.79 CC 0.76
AA 1.02 AA 1.02
0.70 0.50 0.81 0.81 0.22 0.34 2.01 2.01
7.0 8.0 8.0 8.0 6.5 8.0 8.0 8.0
5.9 4.0 6.5 6.5 1.4 2.7 16.1 16.1
The inter-test variability of the overall relative and lifetime risks was large in all three individuals, but most profoundly in individual 3 (Table 3). For this person, these risks were lower and higher than the population risk, depending on the test. Lifetime risks between lowest and highest estimate differed by factor 1.7, 1.6, and 11.5 for individuals 1, 2, and 3, respectively. 
Risk Prediction Based on Three Continent AMD Consortium
The prediction model developed by the population-based 3CC consists of 31 variables, which were represented in a total risk score indicating the risk of developing late AMD. 21 For each individual the total risk score was calculated (Table 4) and used to assess lifetime risks. Lifetime population risk for developing late AMD was 17.4% at a life expectancy of 90 years in the 3CC cohort. Lifetime risks for all three individuals were also calculated using the 3CC risk score, and were 4.2%, 0.5%, and 0.5%, respectively (Table 4). Although the population risk in the 3CC cohort was much higher than for the DTC tests, lifetime risks for the three individuals were considerably lower than the lifetime risks provided by the companies (4.9–8.6; 4.0–6.5; 1.4–16.1, Table 3). 
Table 4
 
Risk Estimates From the Three Continent AMD Consortium Prediction Model
Table 4
 
Risk Estimates From the Three Continent AMD Consortium Prediction Model
Variable Code Risk Per Code Individual 1 Individual 2 Individual 3
ARMS2 rs10490924 GG = 0/GT = 1/TT = 2 0/0.779/1.720 0 0.779 0
ADAMTS9 rs6795735 CC = 0/TC = 1/TT = 2 0/0.130/0.424 0 0.424 0.424
SLC16A8 rs8135665 CC = 0/TC = 1/TT = 2 0/0.313/0.648 0.313 0 0.313
Sex M = 0/F = 1 0/0.320 0.320 0.320 0
CETP rs3764261 CC = 0/CA = 1/AA = 2 0/0.215/0.478 0.215 0 0
CFH rs1061170 TT = 0/TC = 1/CC = 2 0/0.175/0.278 0.175 0 0.175
Smoking Never = 0/past = 1/current = 2 0/0.164/0.651 0 0 0.164
MYRIP rs2679798 AA = 0/AG = 1/GG = 2 0/0.059/0.156 0.059 0.156 0
VEGFA rs943080 CC = 0/TC = 1/TT = 2 0/0/0.098 0 0 0.098
TNFRSF10A rs13278062 TT = 0/TG = 1/GG = 2 0/0.093/0.196 0.093 0 0
TGBR1 rs334353 TT = 0/TG = 1/GG = 2 0/0.039/−0.336 0.039 0.039 0
IER3/DDR1 rs3130783 AA = 0/AG = 1/GG = 2 0/0.029/0.166 0 0.029 0.029
SKIV2L rs429608 GG = 0/GA = 1/AA = 2 0/0.027/0.590 0 0 0.027
Age, y = <65 = 0/65–75 = 1/75+ = 2 0/1.558/2.433 0 0 0
AMD baseline grade Level 10 = 0/level 20 = 1/  level 30 = 2/level 40 = 3 0/1.458/2.560/3.398 0 0 0
BMI, kg/m2 = <25 = 0/25+ = 1 0/0.007 0 0 0
C2/CFB rs4151667 TT = 0/TA or AA = 1 0/−1.245 0 0 0
B3GALTL rs9542236 TT = 0/TC = 1/CC = 2 0/−0.231/−0.169 0 0 0
LIPC rs12912415 AA = 0/AG or GG = 1 0/−0.098 0 0 0
COL8A1 rs13081855 GG = 0/GT = 1/TT = 2 0/0.223/0.890 0 0 0
TIMP3 rs5749482 GG = 0/GC or CC = 1 0/−0.357 0 0 0
C3 rs2230199 CC = 0/GC = 1/GG = 2 0/−0.033/0.755 −0.033 −0.033 0
ABCA1 rs1883025 CC = 0/TC = 1/TT = 2 0/−0.046/0.076 −0.046 −0.046 0
LPL rs256 CC = 0/TC or TT = 1 0/−0.048 0 −0.048 −0.048
CFI rs10033900 CC = 0/TC = 1/TT = 2 0/−0.070/−0.223 0 −0.070 −0.070
C3 rs433594 GG = 0/GA = 1/AA = 2 0/−0.110/−0.591 −0.110 −0.110 0
FRK/COL10A1 rs3812111 TT = 0/TA = 1/AA = 2 0/−0.278/−0.118 0 0 −0.118
RAD51B rs8017304 AA = 0/AG = 1/GG = 2 0/−0.414/−0.138 0 0 −0.414
C2/CFB rs641153 GG = 0/GA or AA = 1 0/−0.592 0 0 −0.592
CFH rs800292 GG = 0/GA = 1/AA = 2 0/−0.899/−1.614 0 −0.899 −0.899
CFH rs12144939 GG = 0/GT = 1/TT = 2 0/−0.947/−1.195 0 −0.947 0
Total risk score 1.025 −0.406 −0.911
Lifetime risk, % 4.2 0.5 0.5
Discussion
Until recently, anyone could order a DTC test and get a personal risk estimate for common diseases. Interpretation of the test results and evaluation of their validity has been difficult, even for professionals. Our study shows that predicted risks of AMD vary considerably among DTC tests, and none may represent the true disease risk. 
We examined four DTC tests in three individuals, and compared test results with predicted risks from a validated model developed in the large population-based 3CC. 21 Predicted risks varied widely within each individual, and differences between highest and lowest estimates for lifetime risk were up to 12-fold. Within the same person, overall RRs could be increased as well as decreased, depending on which test was used. All tests provided higher estimates for lifetime risk than the 3CC model. Several key points explain these differences. 
First, the DTC tests genotyped only two to six SNPs to calculate the risk of AMD. These risks were often based on case-control studies instead of population-based studies, which often comprise lower risks. 21 Recent reports show that greater than 30 loci have been associated by GWAS studies (Fritsche LG, et al. IOVS 2014;55:ARVO E-Abstract 4570). 1 Not testing a comprehensive set of SNPs may lead to imbalance of harmful and protective SNPs, and provide a very different overall risk estimate. For example, individual 2 had several important risk-increasing as well as -decreasing variants (Table 4), and not testing these hampered accurate risk profiling (Table 3). This was also acknowledged for the population at large; inclusion of an extended set of variants increased risk prediction in three population-based studies. 21 We expect that even more common and rare variants will be identified for AMD in the near future, and inclusion of these variants will further refine personalized risk prediction. 
Second, the lifetime population risk and reference population differed among the DTC tests. The lifetime population risk used by 23andMe was lower than that used by the other companies, and differed for men and women. Which population had been used as reference for the calculation of the lifetime AMD population risk was not specified by any of the companies. They were all lower than the lifetime population risk estimate in 3CC (6.5%–8% vs. 17.4%, respectively). Lifetime population risks were based on life expectancy of 79 years for 23andMe and 90 years for 3CC. No information was provided on life expectancy by the other companies. The average life expectancy is currently above 80 years in Western Europe and 79 years in the United States. 45 Life expectancy increases once a certain age has been reached: For instance, persons who reached the age of 80 years during 2008 to 2010 in France still had an average life expectancy of 8.3 years for men and 10.6 years for women. 46 In these persons, a life expectancy of 90 years is not unrealistic. Ancestry also influences the risk estimates. All companies asked the applicant for their ethnicity and used questionnaire data for analysis. However, calculation of ancestry is more accurate using MDS analysis with genotype data. In GTL and Easy DNA, all results were based on European Tuscan ancestry, although European ethnicity was stated by the individuals at application. Multidimensional scaling analysis with genotype data from all three individuals confirmed their Northern/Western European ancestry comparable with their appearance (Supplementary Fig. S1). Why a Tuscan ancestry was chosen for these individuals is unclear and incorrect. The choice of two different ancestries (European Tuscan and European) in one individual (Table 3) in these tests is presumably an unintended error. 
The conversion to a different ancestry can lead to an alteration of the risk, since the frequency of genotypes may differ among ethnicities. The minor allele frequency (MAF) for the CFH rs1061170 variant in the Easy-DNA and GTL tests was set at 17% for those with Tuscan ancestry. Minor allele frequency for this variant varies among ethnicities: approximately 36% in Europeans and Africans, approximately 17% in Latinos/Hispanics, and approximately 10% to 15% in Asians. 47 Tuscans cluster more closely with Northern/Western Europeans than with Latinos/Hispanics (Supplementary Fig. S1), and literature indicates that the actual MAF of the CFH rs1061170 variant in an Italian population is also 36%. 48 Therefore, these companies should have used a MAF of 36% rather than 17% for European Tuscans. Not using the correct MAF resulted in higher risks since all risks per SNP have been adjusted for the average risk of the SNP in the population, which can be calculated using the risk per genotype and genotype frequency. This effect is particularly visible in the risks for individual 1 (Table 3); risks provided by Easy-DNA used the European ancestry as reference population and a MAF of 36% resulting in an RR of 1.26, while GTL used the European Tuscan ancestry with a MAF of 17% resulting in a higher RR of 1.60. For carriers of the CFH rs1061170 CC-genotype this difference in risk will be even more extreme. In summary, an incorrect reference population was assigned to the three individuals and to this reference population (Tuscans) an incorrect MAF for the CFH rs1061170 SNP was assigned. In this particular case the largest effect on risk prediction of AMD was the incorrect assigned MAF. This most likely influenced the risk prediction for the other diseases predicted by the companies as well. 
Third, there were mistakes in assignment of an AMD risk variant. Easy-DNA and GTL stated that the tested SNP rs800292 was located in the C2 gene, when in fact this particular rs-number is located in the CFH gene. 49 Apart from the incorrect gene, the direction of the risk for this variant was opposite of that reported in 3CC 21 ; in the tests from Easy-DNA and GTL the T allele was set as the risk variant, increasing the risk of AMD, while in 3CC this allele decreased the risk of AMD. 
Fourth, the DTC tests lacked inclusion of nongenetic risk factors. Only 23andMe took age and sex into account in their risk calculation. Age is the most important nongenetic factor associated with AMD known to date, and it is therefore prudent to incorporate this factor in risk predictions of AMD. 50 None of the companies included environmental factors in their risk prediction. We recommend inclusion of smoking since this factor is an important environmental risk factor for AMD, 51 which also shows interaction with genetic risk variants. 39 Inclusion of nongenetic risk factors can improve the predictive ability of the test. 21  
Lastly, the companies applied different methods for their risk calculation. A recent study examined and compared the methods from three DTC tests (23andMe, deCODEme, and Navigenics) for several diseases including AMD. 22 The authors showed that the formulas used by deCODEme can lead to a predicted risk exceeding 100% in high risk cases. The formulas used by 23andMe followed the Bayes' theorem preventing risks to exceed 100%, leading to more realistic risk estimates. Unfortunately, methods for risk calculation were not provided by Easy-DNA or GTL, and could therefore not be evaluated. 
Recently, many companies stopped offering DTC tests. Several issues played a role. First, the Food and Drug Administration questioned the evidence of the safety and efficacy of these prediction tests. 52 Second, it was unclear what actions the individual will take when made aware of his/her genetic profile. Third, healthcare professionals lacked guidelines for counseling and patient management after genetic profiling. Do these issues apply to DTC tests for AMD? Our study encountered no genotyping errors. Nevertheless, predictions were inaccurate based on methodology. It is indeed unclear what an individual should do when diagnosed with a high genetic risk of AMD, and what a clinician should advise such patients. Cessation of smoking and lowering BMI is advice, which applies to all persons. However, it is likely that individuals who have been made aware of a high genetic risk after testing will be more motivated to make drastic life style changes than persons who are ignorant. 
Although genetic testing for prediction of disease risk is the next step to personalized medicine, the current state of the art is that most DTC tests are accurate at genotyping, but not at risk prediction. Improvement can be achieved by incorporation of a more comprehensive set of genetic markers with population-based risks. Inclusion of nongenetic risk factors, a more adequate choice of the reference population, and implementation of valid methodology for risk calculation will further improve these tests. Only then will these genetic tests become suitable for clinical practice. 
Supplementary Materials
Acknowledgments
The authors thank Sven J. van der Lee and Maarten Kooijman for their assistance in the MDS analysis, Michiel Koolhaas for logistics of the DTC-tests, and members of the Three Continent AMD Consortium for providing the risk score calculation from their developed prediction model. 
Supported by grants from the Stichting Nederlands Oog Onderzoek (SNOO) Rotterdam, The Netherlands; stichting UITZICHT, The Netherlands; Netherlands Organization for Scientific Research, The Hague; Swart van Essen, Rotterdam, the Netherlands; Bevordering van Volkskracht, Rotterdam, The Netherlands; Rotterdamse Blindenbelangen Association, Rotterdam, The Netherlands; Algemene Nederlandse Vereniging ter Voorkoming van Blindheid, Doorn, The Netherlands; Oogfonds Nederland, Utrecht, The Netherlands; MDFonds, Utrecht, The Netherlands; Vereniging Trustfonds Erasmus Universiteit Rotterdam, Rotterdam, The Netherlands; and Lijf en Leven, Krimpen aan de IJssel, The Netherlands. An unrestricted grant was obtained from Topcon Europe BV, Capelle aan den IJssel, The Netherlands. 
Disclosure: G.H.S. Buitendijk, None; N. Amin, None; A. Hofman, None; C.M. van Duijn, None; J.R. Vingerling, None; C.C.W. Klaver, None 
References
Fritsche LG Chen W Schu M Seven new loci associated with age-related macular degeneration. Nat Genet . 2013; 45: 433–439. [CrossRef] [PubMed]
Chen Y Zeng J Zhao C Assessing susceptibility to age-related macular degeneration with genetic markers and environmental factors. Arch Ophthalmol . 2011; 129: 344–351. [CrossRef] [PubMed]
Gibson J Cree A Collins A Lotery A Ennis S. Determination of a gene and environment risk model for age-related macular degeneration. Br J Ophthalmol . 2010; 94: 1382–1387. [CrossRef] [PubMed]
Gold B Merriam JE Zernant J Variation in factor B (BF) and complement component 2 (C2) genes is associated with age-related macular degeneration. Nat Genet . 2006; 38: 458–462. [CrossRef] [PubMed]
Grassmann F Fritsche LG Keilhauer CN Heid IM Weber BH. Modelling the genetic risk in age-related macular degeneration. PLoS One . 2012; 7: e37979. [CrossRef] [PubMed]
Hageman GS Gehrs K Lejnine S Clinical validation of a genetic model to estimate the risk of developing choroidal neovascular age-related macular degeneration. Hum Genomics . 2011; 5: 420–440. [CrossRef] [PubMed]
Jakobsdottir J Conley YP Weeks DE Ferrell RE Gorin MB. C2 and CFB genes in age-related maculopathy and joint action with CFH and LOC387715 genes. PLoS One . 2008; 3: e2199. [CrossRef] [PubMed]
Jakobsdottir J Gorin MB Conley YP Ferrell RE Weeks DE. Interpretation of genetic association studies: markers with replicated highly significant odds ratios may be poor classifiers. PLoS Genet . 2009; 5: e1000337. [CrossRef] [PubMed]
Klein ML Francis PJ Ferris FL III Hamon SC Clemons TE. Risk assessment model for development of advanced age-related macular degeneration. Arch Ophthalmol . 2011; 129: 1543–1550. [CrossRef] [PubMed]
McKay GJ Dasari S Patterson CC Chakravarthy U Silvestri G. Complement component 3: an assessment of association with AMD and analysis of gene-gene and gene-environment interactions in a Northern Irish cohort. Mol Vis . 2010; 16: 194–199. [PubMed]
Perlee LT Bansal AT Gehrs K Inclusion of genotype with fundus phenotype improves accuracy of predicting choroidal neovascularization and geographic atrophy. Ophthalmology . 2013; 120: 1880–1892. [CrossRef] [PubMed]
Reynolds R Hartnett ME Atkinson JP Giclas PC Rosner B Seddon JM. Plasma complement components and activation fragments: associations with age-related macular degeneration genotypes and phenotypes. Invest Ophthalmol Vis Sci . 2009; 50: 5818–5827. [CrossRef] [PubMed]
Seddon JM Reynolds R Maller J Fagerness JA Daly MJ Rosner B. Prediction model for prevalence and incidence of advanced age-related macular degeneration based on genetic, demographic, and environmental variables. Invest Ophthalmol Vis Sci . 2009; 50: 2044–2053. [CrossRef] [PubMed]
Seddon JM Reynolds R Yu Y Daly MJ Rosner B. Risk models for progression to advanced age-related macular degeneration using demographic, environmental, genetic, and ocular factors. Ophthalmology . 2011; 118: 2203–2211. [CrossRef] [PubMed]
Sharma NK Sharma SK Gupta A Prabhakar S Singh R Anand A. Predictive model for earlier diagnosis of suspected age-related macular degeneration patients. DNA Cell Biol . 2013; 32: 549–555. [CrossRef] [PubMed]
Spencer KL Olson LM Schnetz-Boutaud N Using genetic variation and environmental risk factor data to identify individuals at high risk for age-related macular degeneration. PLoS One . 2011; 6: e17784. [CrossRef] [PubMed]
Ying GS Maguire MG. Development of a risk score for geographic atrophy in complications of the age-related macular degeneration prevention trial. Ophthalmology . 2011; 118: 332–338. [CrossRef] [PubMed]
Yu Y Reynolds R Rosner B Daly MJ Seddon JM. Prospective assessment of genetic effects on progression to different stages of age-related macular degeneration using multistate Markov models. Invest Ophthalmol Vis Sci . 2012; 53: 1548–1556. [CrossRef] [PubMed]
Zanke B Hawken S Carter R Chow D. A genetic approach to stratification of risk for age-related macular degeneration. Can J Ophthalmol . 2010; 45: 22–27. [CrossRef] [PubMed]
Hughes AE Orr N Patterson C Neovascular age-related macular degeneration risk based on CFH, LOC387715/HTRA1, and smoking. PLoS Med . 2007; 4: e355. [CrossRef] [PubMed]
Buitendijk GH Rochtchina E Myers C Prediction of age-related macular degeneration in the general population: the Three Continent AMD Consortium. Ophthalmology . 2013; 120: 2644–2655. [CrossRef] [PubMed]
Kalf RR Mihaescu R Kundu S de Knijff P Green RC Janssens AC. Variations in predicted risks in personal genome testing for common complex diseases. Genet Med . 2013; 16: 85–91. [CrossRef] [PubMed]
Conley YP Jakobsdottir J Mah T CFH, ELOVL4, PLEKHA1 and LOC387715 genes and susceptibility to age-related maculopathy: AREDS and CHS cohorts and meta-analyses. Hum Mol Genet . 2006; 15: 3206–3218. [CrossRef] [PubMed]
Johnson PT Betts KE Radeke MJ Hageman GS Anderson DH Johnson LV. Individuals homozygous for the age-related macular degeneration risk-conferring variant of complement factor H have elevated levels of CRP in the choroid. Proc Natl Acad Sci U S A . 2006; 103: 17456–17461. [CrossRef] [PubMed]
Laine M Jarva H Seitsonen S Y402H polymorphism of complement factor H affects binding affinity to C-reactive protein. J Immunol . 2007; 178: 3831–3836. [CrossRef] [PubMed]
Maller J George S Purcell S Common variation in three genes, including a noncoding variant in CFH, strongly influences risk of age-related macular degeneration. Nat Genet . 2006; 38: 1055–1059. [CrossRef] [PubMed]
Skerka C Lauer N Weinberger AA Defective complement control of factor H (Y402H) and FHL-1 in age-related macular degeneration. Mol Immunol . 2007; 44: 3398–3406. [CrossRef] [PubMed]
Sofat R Casas JP Webster AR Complement factor H genetic variant and age-related macular degeneration: effect size, modifiers and relationship to disease subtype. Int J Epidemiol . 2012; 41: 250–262. [CrossRef] [PubMed]
Chen W Stambolian D Edwards AO Genetic variants near TIMP3 and high-density lipoprotein-associated loci influence susceptibility to age-related macular degeneration. Proc Natl Acad Sci U S A . 2010; 107: 7401–7406. [CrossRef] [PubMed]
Neale BM Fagerness J Reynolds R Genome-wide association study of advanced age-related macular degeneration identifies a role of the hepatic lipase gene (LIPC). Proc Natl Acad Sci U S A . 2010; 107: 7395–7400. [CrossRef] [PubMed]
Gold B Merriam JE Zernant J Variation in factor B (BF) and complement component 2 (C2) genes is associated with age-related macular degeneration. Nat Genet . 2006; 38: 458–462. [CrossRef] [PubMed]
Kanda A Chen W Othman M A variant of mitochondrial protein LOC387715/ARMS2, not HTRA1, is strongly associated with age-related macular degeneration. Proc Natl Acad Sci U S A . 2007; 104: 16227–16232. [CrossRef] [PubMed]
Park KH Fridley BL Ryu E Tosakulwong N Edwards AO. Complement component 3 (C3) haplotypes and risk of advanced age-related macular degeneration. Invest Ophthalmol Vis Sci . 2009; 50: 3386–3393. [CrossRef] [PubMed]
Yates JR Sepp T Matharu BK Complement C3 variant and the risk of age-related macular degeneration. N Engl J Med . 2007; 357: 553–561. [CrossRef] [PubMed]
Spencer KL Hauser MA Olson LM Protective effect of complement factor B and complement component 2 variants in age-related macular degeneration. Hum Mol Genet . 2007; 16: 1986–1992. [CrossRef] [PubMed]
Macpherson MNB Hsu A Mountain J. White paper 23-01: estimating genotype-specific incidence for one or several loci. 2007. Available at: https://23andme.https.internapcdn.net/res/pdf/HIC-SXIYiYqXreldAxO5yA_23-01_Estimating_Genotype_Specific_Incidence.pdf. Accessed October 6, 2013.
Maller JB Fagerness JA Reynolds RC Neale BM Daly MJ Seddon JM. Variation in complement factor 3 is associated with risk of age-related macular degeneration. Nat Genet . 2007; 39: 1200–1201. [CrossRef] [PubMed]
deCODEme. Risk calculation. Available at: http://www.decodeme.com/health-watch-information/risk-calculation . Accessed October 6, 2013.
Despriet DD Klaver CC Witteman JC Complement factor H polymorphism, complement activators, and risk of age-related macular degeneration. JAMA . 2006; 296: 301–309. [CrossRef] [PubMed]
Jylhava J Eklund C Jylha M Complement factor H 402His variant confers an increased mortality risk in Finnish nonagenarians: the Vitality 90+ study. Exp Gerontol . 2009; 44: 297–299. [CrossRef] [PubMed]
Bird AC Bressler NM Bressler SB An international classification and grading system for age-related maculopathy and age-related macular degeneration. The International ARM Epidemiological Study Group. Surv Ophthalmol . 1995; 39: 367–374. [CrossRef] [PubMed]
Klein R Davis MD Magli YL Segal P Klein BE Hubbard L. The Wisconsin age-related maculopathy grading system. Ophthalmology . 1991; 98: 1128–1134. [CrossRef] [PubMed]
ENIGMA2 GENETICS Support Team. ENIGMA2 1KGP Cookbook (v3). The Enhancing Neuroimaging Genetics through Meta-Analysis (ENIGMA) consortium. Available at: http://enigma.ini.usc.edu/wp-content/uploads/2012/07/old/ENIGMA2_1KGP_v3_old.pdf. Accessed December 17, 2013.
R Development Core Team. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2010.
WHO. Life expectancy at birth - 2011. Available at: http://gamapserver.who.int/gho/interactive_charts/mbd/life_expectancy/atlas.html . Accessed December 19, 2013.
Prioux F Barbieri M. Recent demographic developments in France: relatively low mortality at advanced ages. Population (Engl Ed) . 2012; 67.
Nonyane BA Nitsch D Whittaker JC An ecological correlation study of late age-related macular degeneration and the complement factor H Y402H polymorphism. Invest Ophthalmol Vis Sci . 2010; 51: 2393–2402. [CrossRef] [PubMed]
Ricci F Zampatti S D'Abbruzzi F Typing of ARMS2 and CFH in age-related macular degeneration: case-control study and assessment of frequency in the Italian population. Arch Ophthalmol . 2009; 127: 1368–1372. [CrossRef] [PubMed]
NCBI. dbSNP. Available at: http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=800292 . Accessed December 20, 2013.
van Leeuwen R Klaver CC Vingerling JR Hofman A de Jong PT. The risk and natural course of age-related maculopathy: follow-up at 6 1/2 years in the Rotterdam study. Arch Ophthalmol . 2003; 121: 519–526. [CrossRef] [PubMed]
Vingerling JR Hofman A Grobbee DE de Jong PT. Age-related macular degeneration and smoking. The Rotterdam Study. Arch Ophthalmol . 1996; 114: 1193–1196. [CrossRef] [PubMed]
FDA: US Food and Drug Administration. Medical Devices . US Department of Health and Human Services. 2014. Available at: http://www.fda.gov/MedicalDevices/ProductsandMedicalProcedures/InVitroDiagnostics/default.htm. Accessed June 30, 2014.
Table 1
 
Descriptives of the Participants
Table 1
 
Descriptives of the Participants
Variable Individual 1 Individual 2 Individual 3
Age, y 45 29 51
Sex Female Female Male
Ethnicity Caucasian Caucasian Caucasian
Ancestry Northern/Western European Northern/Western European Northern/Western European
BMI, kg/m2 22.7 20.2 24.3
Smoking Never Never Past
AMD phenotype None None None
Family history of AMD Grandmother None None
Table 2
 
Overview Genetic Testing Companies
Table 2
 
Overview Genetic Testing Companies
Company Name Website Costs Per Kit DNA Source Easy To Collect? Additional Notes
23andMe https://www.23andme.com $99/€74 Saliva Difficult in  1 participant Street address is needed to deliver DTC-test
deCODEme* https://www.decodeme.com $1100/€821 Buccal Yes
Easy-DNA http://www.easygenetictest.com $299/€299 Blood Yes For US residents: sample needs to be collected by physiscian or professional collector
Genetic Testing Laboratories http://www.gtldna.com/ $285/€213 Blood Yes Sample needs to be collected by physiscian or professional collector
Table 3
 
Risks of the Tested Variants, Overall Risk, and Lifetime Risk Per Company for Each Individual
Table 3
 
Risks of the Tested Variants, Overall Risk, and Lifetime Risk Per Company for Each Individual
AMD Individual 1
Gene SNP Number 23andMe deCODEme Easy-DNA* GTL
Genotype OR Genotype RR Genotype RR Genotype RR
CFH rs1061147 AC 0.97 AC 1.56||
CFH rs1329428 GG
CFH rs1061170 CT 1.26 CT 1.60
CFH rs800292 CC 0.67 CC 0.63
C2 rs547154 GG 1.07 CC 1.10
C2 rs9332739 GG 1.06
LOC387715/ARMS2 rs3750847 CC 0.47 GG 0.46
C3 rs2230199 CG 1.37 CG 1.29
TIMP3 rs9621532 AA 1.02
Overall RR‡ 0.70 1.01 0.85 1.00
Lifetime population risk, % 7.0 8.0 8.0 8.0
Lifetime risk, %§ 4.9 8.6 6.8 8.1
Table 3
 
Extended
Table 3
 
Extended
Individual 2 Individual 3
23andMe deCODEme Easy-DNA GTL 23andMe deCODEme Easy-DNA GTL
Genotype OR Genotype RR Genotype RR Genotype RR Genotype OR Genotype RR Genotype RR Genotype RR
CC 0.34 CC 0.21|| AC 0.97 AC 1.56||
AA GG
TT 0.64 TT 0.64 CT 1.60 CT 1.60
CT 1.26 CT 1.26 CT 1.26 CT 1.26
GG 1.07 CC 1.10 GT 0.57 AC 0.58
GG 1.06 GG 1.06
CT 1.63 AG 1.59 CC 0.47 GG 0.46
CG 1.37 CG 1.29 GG 0.79 CC 0.76
AA 1.02 AA 1.02
0.70 0.50 0.81 0.81 0.22 0.34 2.01 2.01
7.0 8.0 8.0 8.0 6.5 8.0 8.0 8.0
5.9 4.0 6.5 6.5 1.4 2.7 16.1 16.1
Table 4
 
Risk Estimates From the Three Continent AMD Consortium Prediction Model
Table 4
 
Risk Estimates From the Three Continent AMD Consortium Prediction Model
Variable Code Risk Per Code Individual 1 Individual 2 Individual 3
ARMS2 rs10490924 GG = 0/GT = 1/TT = 2 0/0.779/1.720 0 0.779 0
ADAMTS9 rs6795735 CC = 0/TC = 1/TT = 2 0/0.130/0.424 0 0.424 0.424
SLC16A8 rs8135665 CC = 0/TC = 1/TT = 2 0/0.313/0.648 0.313 0 0.313
Sex M = 0/F = 1 0/0.320 0.320 0.320 0
CETP rs3764261 CC = 0/CA = 1/AA = 2 0/0.215/0.478 0.215 0 0
CFH rs1061170 TT = 0/TC = 1/CC = 2 0/0.175/0.278 0.175 0 0.175
Smoking Never = 0/past = 1/current = 2 0/0.164/0.651 0 0 0.164
MYRIP rs2679798 AA = 0/AG = 1/GG = 2 0/0.059/0.156 0.059 0.156 0
VEGFA rs943080 CC = 0/TC = 1/TT = 2 0/0/0.098 0 0 0.098
TNFRSF10A rs13278062 TT = 0/TG = 1/GG = 2 0/0.093/0.196 0.093 0 0
TGBR1 rs334353 TT = 0/TG = 1/GG = 2 0/0.039/−0.336 0.039 0.039 0
IER3/DDR1 rs3130783 AA = 0/AG = 1/GG = 2 0/0.029/0.166 0 0.029 0.029
SKIV2L rs429608 GG = 0/GA = 1/AA = 2 0/0.027/0.590 0 0 0.027
Age, y = <65 = 0/65–75 = 1/75+ = 2 0/1.558/2.433 0 0 0
AMD baseline grade Level 10 = 0/level 20 = 1/  level 30 = 2/level 40 = 3 0/1.458/2.560/3.398 0 0 0
BMI, kg/m2 = <25 = 0/25+ = 1 0/0.007 0 0 0
C2/CFB rs4151667 TT = 0/TA or AA = 1 0/−1.245 0 0 0
B3GALTL rs9542236 TT = 0/TC = 1/CC = 2 0/−0.231/−0.169 0 0 0
LIPC rs12912415 AA = 0/AG or GG = 1 0/−0.098 0 0 0
COL8A1 rs13081855 GG = 0/GT = 1/TT = 2 0/0.223/0.890 0 0 0
TIMP3 rs5749482 GG = 0/GC or CC = 1 0/−0.357 0 0 0
C3 rs2230199 CC = 0/GC = 1/GG = 2 0/−0.033/0.755 −0.033 −0.033 0
ABCA1 rs1883025 CC = 0/TC = 1/TT = 2 0/−0.046/0.076 −0.046 −0.046 0
LPL rs256 CC = 0/TC or TT = 1 0/−0.048 0 −0.048 −0.048
CFI rs10033900 CC = 0/TC = 1/TT = 2 0/−0.070/−0.223 0 −0.070 −0.070
C3 rs433594 GG = 0/GA = 1/AA = 2 0/−0.110/−0.591 −0.110 −0.110 0
FRK/COL10A1 rs3812111 TT = 0/TA = 1/AA = 2 0/−0.278/−0.118 0 0 −0.118
RAD51B rs8017304 AA = 0/AG = 1/GG = 2 0/−0.414/−0.138 0 0 −0.414
C2/CFB rs641153 GG = 0/GA or AA = 1 0/−0.592 0 0 −0.592
CFH rs800292 GG = 0/GA = 1/AA = 2 0/−0.899/−1.614 0 −0.899 −0.899
CFH rs12144939 GG = 0/GT = 1/TT = 2 0/−0.947/−1.195 0 −0.947 0
Total risk score 1.025 −0.406 −0.911
Lifetime risk, % 4.2 0.5 0.5
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×