SEARCH

SEARCH BY CITATION

Keywords:

  • Disease association;
  • transmission/disequilibrium test;
  • TDT;
  • case control;
  • sickle cell;
  • hemoglobin;
  • β-globin;
  • malaria;
  • plasmodium falciparum;
  • segregation distortion

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Electronic-Database Information
  9. Acknowledgments
  10. References

There has been much debate about the relative merits of population- and family-based strategies for testing genetic association, yet there is little empirical data that directly compare the two approaches. Here we compare case-control and transmission/disequilibrium test (TDT) study designs using a well-established genetic association, the protective effect of the sickle-cell trait against severe malaria. We find that the two methods give similar estimates of the level of protection (case-control odds ratio = 0.10, 95% confidence interval 0.03–0.23; family-based estimate of the odds ratio = 0.11, 95% confidence interval 0.04–0.25) and similar statistical significance of the result (case-control: χ2= 41.26, p= 10−10, TDT: χ2= 39.06, p= 10−10) when 315 TDT cases are compared to 583 controls. We propose a family plus population control study design, which allows both case-control and TDT analysis of the cases. This combination is robust against the respective weaknesses of the case-control and TDT study designs, namely population structure and segregation distortion. The combined study design is especially cost-effective when cases are difficult to ascertain and, when the case-control and TDT results agree, offers greater confidence in the result.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Electronic-Database Information
  9. Acknowledgments
  10. References

There is growing enthusiasm for the use of association rather than linkage analysis to investigate the genetic bases of complex diseases. This is based largely on its power to detect common susceptibility alleles, which is particularly important when searching for multiple genetic determinants whose individual contributions to overall disease susceptibility may be modest (Risch & Merikangas, 1996). However, there remains some uncertainty about the most effective epidemiological strategy to test for genetic association.

There are two approaches, both with potential drawbacks. The case-control method requires cases and controls to be drawn from exactly the same genetic pool, to avoid artifactual associations due to ethnic admixture. This can be exceedingly difficult to ensure in urban and peri-urban populations, and statistical methods have been proposed to detect and control for population admixture (Bacanu et al. 1998). The alternative to case-control is a family-based approach, such as the transmission/disequlibrium test (TDT), which effectively compares allele frequencies in affected individuals with those predicted from their parental genotypes (Spielman et al. 1993; Ewens & Spielman, 1995). The TDT is unaffected by population stratification but is liable to a different source of artifact, known as segregation distortion, at any locus that influences survival in early life. For example, an allele that is highly favourable for successful fertilization may have a higher frequency amongst live births than standard Mendelian ratios would predict, thereby confounding the TDT. Such artifacts can be excluded by comparing TDT data from affected families against TDT data from families unselected for disease status (Naumova et al. 1995; Eaves et al. 1999), but control families are not always available and alternative security measures are needed. One potential solution is to use both case-control and TDT analyses, and to check that associations are replicated. This raises the issue of whether case-control and TDT analyses are expected to give identical results.

The purpose of this study was to compare the statistical output of a case-control analysis and the TDT for a known protective allele in a complex disease. Plasmodium falciparum kills over a million African children each year, but a much larger number of children survive repeated infections with this parasite. A combination of environmental, parasite genetic and host genetic factors is thought to determine an individual's susceptibility to severe malaria (Kwiatkowski, 2000), defined here as life-threatening complications such as cerebral malaria or severe malarial anaemia, which in The Gambia occur in approximately 1% of infections (Greenwood et al. 1987). A previous case-control analysis showed the sickle cell trait (i.e., heterozygous carriage of the HbS allele) to be a strong protective factor against severe malaria in this population (Hill et al. 1991).

We genotyped 315 Gambian children with severe malaria. To analyze them by the TDT we genotyped their 630 parents; for case-control analysis, we genotyped 583 population controls. Given the theoretical evidence that TDT and case-control methods have roughly similar power (Risch, 2000), we set out to provide empirical evidence that both methods of analysis should yield a similar p value for the protective effect of the HbS allele. It has also been implied that, in Hardy-Weinberg populations, the nontransmitted alleles among parents of affected children should equal the allele frequency among the general population (Spielman et al. 1993; Thomson, 1995). We hope to demonstrate empirically that this is true in our study population. If the nontransmitted alleles among parents do equal the allele frequency of the general population, then the odds ratio of allele frequency in transmitted versus non-transmitted chromosomes should provide an estimate of the population relative risk associated with the HbS allele, of similar magnitude to that estimated by the odds ratio in a case-control analysis. Finally, we hope to provide the first evidence from family-based association methods supporting the protective role of the sickle cell trait in severe malaria.

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Electronic-Database Information
  9. Acknowledgments
  10. References

Subjects

Cases of severe malaria were identified on the children's ward of the Royal Victoria Hospital in Banjul, The Gambia, during the 1997-98 rainy seasons. Cerebral malaria was defined as Blantyre coma score <3, persisting for at least 30 minutes after correction of hypoglycaemia or treatment of convulsions, with asexual forms of P. falciparum on blood film and no other evident cause of coma (Molyneux et al. 1989). Severe malarial anaemia was defined as packed cell volume ≤15% (equivalent to hemoglobin ≤5 g/dl) with asexual forms of P. falciparum on blood film. DNA was extracted from blood samples drawn from each child and both parents. For estimates of population allele frequency, DNA was extracted from umbilical cord blood collected from perinatal clinics in the Gambia. All samples were taken with informed parental consent, and the study was approved by the Gambian Government/Medical Research Council Joint Ethical Committee.

Genotyping

The sickle cell mutation is an A to T transition in codon six of β-globin (Hb S [MIM 141900.0243]). Allele-specific PCR primers were designed to amplify either the HbS allele ( 5′GGC AGT AAC GGC AGA CTT CTC CA 3′) or the normal HbA allele ( 5′ GGC AGT AAC GGC AGA CTT CTC CT 3′) when paired with a common reverse primer ( 5′ GCC AGT GCC AGA AGA GCC AA 3′). Separate PCR reactions were set up for the HbA and HbS primer pairs, and each reaction also contained control primers ( 5′ TGC CAA GTG GAG CAC CCA A 3′ and 5′ GCA TCT TGC TCT GTG CAG AT 3′) to generate a 796 bp amplicon from the HLA-DRB1 gene as a positive control. Genotypes were scored from photographs of ethidium-bromide stained gels, having randomized DNA samples to conceal their identity. Ten percent of samples were genotyped in duplicate to serve as an internal control, no discrepancies were found.

Statistical Analysis

Odds ratios estimating the protective effect of carrying one or two copies of the sickle cell allele (genotype A/S or S/S), versus carrying zero copies of the sickle cell allele on developing severe malaria (i.e., cerebral malaria or severe malarial anaemia) were calculated by comparing cases and controls stratified by ethnic group using the Mantel-Haenszel method. To assess for an effect of the sickle cell allele on mortality, only fatal cases were compared against population controls, stratified by ethnic group. Transmissions of the HbS allele to children with malaria were analysed by the Transmission Disequilibrium Test (Spielman et al. 1993). Significance of the TDT was tested by the chi-squared test, except in the case of severe malarial anaemia, which was tested by Fisher's Exact Test. The frequency of HbS alleles among parents of children with severe malaria was compared against the frequency of HbS alleles among cord blood controls, and tested by chi-squared analysis. Parental alleles were then separated into two groups, transmitted and non-transmitted chromosomes, and each group was compared against the population controls using the chi-squared test. Transmitted chromosomes were then compared to non-transmitted chromosomes to calculate an odds ratio, with chi-squared analysis to measure the significance of the result.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Electronic-Database Information
  9. Acknowledgments
  10. References

Study Population

Index cases were 315 children with severe malaria, of whom 235 had cerebral malaria and 80 had severe malarial anaemia. Children with cerebral malaria tended to be older than those with severe malarial anaemia (median age 48 vs 26 months) and had a higher case fatality rate (67/235 vs 4/80). Both groups were 45% female and of similar ethnic composition. Maternal ethnicity was 40% Mandinka, 22% Jola, 11% Wollof, 11% Fula, and 15% other.

Population controls were 583 samples of umbilical cord blood obtained from maternity clinics in 6 different towns in the Gambia. Maternal ethnicity was 29% Mandinka, 15% Jola, 16% Wollof, 23% Fula and 17% other.

Case-Control Analysis

Table 1 shows genotypes stratified by ethnic group. Overall HbS allele frequency in population controls was 0.087. Among cases of severe malaria, overall HbS allele frequency was 0.011. We estimated the protective effect of carrying the HbS allele in either the heterozygous or homozygous state. By simple chi-squared test this yielded an uncorrected odds ratio of 0.10, 95% confidence interval 0.03 –0.22 (p= 2.7 × 10−11); or after correction for ethnic group by stratified Mantel-Haenszel test, a weighted odds ratio of 0.10, 95% confidence interval 0.03 –0.23 (p= 1.4 × 10−10).

Table 1.  Genotypes and allele frequencies in severe malaria cases and controls from 5 ethnic classifications
Ethnic groupSevere malaria casesPopulation controls
NA/SS/SHbS frequencyNA/SS/SHbS frequency
All315510.0115839530.087
Mandinka126100.0041733300.095
Jola70110.021881300.074
Fula36200.0281372700.099
Wollof35100.014921110.071
Other48000931120.081

Table 2 separates cases of cerebral malaria and severe malarial anaemia, and fatal cases from survivors. The confidence intervals of the odds ratios for the clinical subgroups of cerebral malaria and severe malarial anaemia broadly overlap, suggesting that the protective effect is similar in each subgroup. When fatal cases only were compared to population controls, the protective effect was similar in magnitude to that estimated from survivors, suggesting that the sickle trait does not affect mortality among those who develop severe malaria.

Table 2.  Case-control analysis
 NA/SS/SHbS frequencyOdds Ratioap
  1. aOdds ratio from Mantel-Haenszel analysis of cases versus population controls stratified by ethnic group. Exact 95% confidence intervals are provided in parentheses.

Population controls5839530.087
All severe malaria315510.0110.10 (0.03–0.23)1.4 × 10−10
Cerebral malaria235410.0130.11 (0.03–0.27)3.1 × 10−8
 Fatal cases67200.0150.16 (0.02–0.62)0.007
 Survivors168210.0120.10 (0.02–0.28)1.2 × 10−6
Severe malarial anaemia80100.0060.07 (0.00–0.43)6 × 10−4
 Fatal cases400   0… (0.00–7.1)0.78
 Survivors76100.0070.07 (0.00–0.40)7 × 10−4

Family-Based Analysis

To confirm the protective effect of HbS, we proceeded to genotype parents of severe malaria patients (Table 3). Sixty-four of the 630 parents of malaria patients were heterozygous (HbS = 5.1%), and this allele frequency is intermediate between the frequency among cases (1.1%) and controls (8.7%). Fifty-four of the 470 parents of cerebral malaria patients were heterozygous (HbS = 5.7%) and 10 of the 160 parents of severe anaemia patients were heterozygous (HbS = 3.1%). If HbS had no effect on severe malaria, then heterozygous parents (A/S) would be expected to transmit HbS and HbA with equal probability (50%/50%). However, there were only 7 transmissions of HbS and 57 transmissions of HbA from 64 heterozygous parents (11%/89%, χ2= 39.06, p= 4 × 10−10) (Table 4). Among the 54 heterozygous parents of cerebral malaria patients we observed 6 transmissions of HbS, and only 48 transmissions of HbA (11%/89%, χ2= 32.7, p= 10−8), a significant departure from random transmission. Among the 10 heterozygous parents of severe anaemia patients we observed 1 transmission of HbS and 9 transmissions of HbA (10%/90%, Fisher's Exact Test p= 0.07).

Table 3.  Genotypes of parents of severe malaria cases
Parental matingsNCases
A/AA/SS/S
  1. Note – No S/S parents were found in the sample population.

AA × AA254254
AA × AS58535
AS × AS3201
Table 4.  TDT for alleles HbS and HbA of β-globin in severe malaria: Data from 64 A/S parents
PhenotypeNumber of alleles transmittedχ2p
HbSHbATotal
  1. aFisher's Exact Test 1-tailed probability.

All severe malaria7576439.064 × 10−10
Cerebral malaria6485432.6710−8
Severe malarial anaemia1910 0.07a

Comparing Parental Allele Frequency Against Control Allele Frequency

If we first compared all 630 parents of children with severe malaria against the population controls (Table 5); there was a significant difference in allele frequency: 64/1260 (5.1%) versus 101/1166 (8.7%) (OR= 0.56, p= 0.0005). This effect was found in parents of both cerebral malaria patients (OR= 0.64, p= 0.01) and severe anaemia patients (OR= 0.33, p= 0.0008). When we counted the allele frequency among those parental chromosomes that were not transmitted to affected offspring, we found that the allele frequency 57/630 (9.0%) did not significantly differ from that of the population controls 101/1166 (8.7%), p= 0.78. A comparison of transmitted versus nontransmitted chromosomes from the parents (Table 6) gave an odds ratio OR= 0.11 (95% C.I. 0.04–0.25, p= 10−10) similar to a comparison of case genotypes versus population control genotypes OR= 0.10 (95% C.I. 0.03–0.23, p= 10−10).

Table 5.  Comparison of parental chromosomes against population controls
Case groupNumber of HbS allelesOdds ratioχ2p
CaseControl
Parents of severe malaria64/1260101/11660.56 (0.40–0.79)12.265 × 10−4
Transmitted chromosomes7/630101/11660.12 (0.05–0.26)41.2610−10
Non-transmitted chromosomes57/630101/11661.05 (0.73-1.49)0.080.78
Table 6.  Estimate of population odds ratio from family data
Parental chromosomesNumber of allelesTotal
HbS HbA
  1. inline image.

Transmitted7623630
Non-transmitted57573630
 
Total6411961260

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Electronic-Database Information
  9. Acknowledgments
  10. References

We have confirmed that the sickle-cell variant of β-globin (HbS) confers protection against severe malaria using both case-control and family-based analyses. The case-control study estimated that HbS protects against severe malaria with an odds ratio of 0.10 (95% C.I. = 0.03–0.23). Both clinical subgroups of severe malaria, cerebral malaria and severe malarial anaemia, gave similar estimates of protection: OR= 0.11 (95% C.I. = 0.03–0.27) and OR= 0.06 (95% C.I. = 0.00–0.37), respectively. The reduced transmission of the HbS allele from parents to children with severe malaria in the TDT confirmed the case-control association in each clinical category; furthermore, the probability associated with both the TDT and case-control statistics were of the same order of magnitude, p= 10−10.

These results from case-control and family based association analysis agree with previous evidence that the sickle-cell trait protects against severe disease in a case-control study (Hill et al. 1991), and against infection in a community survey (Allison, 1954). We chose the sickle-cell allele as an example of an unequivocally functional polymorphism for the purposes of this empirical comparison of case-control and family-based methods of association; however, this example does not address the relationship between family and population-based methods when using a non-functional disease marker or an allele with a more modest effect.

Estimating an Odds Ratio from Family Data

Although the TDT does not give an estimate of the disease risk associated with an allele, an odds ratio may be calculated by comparing transmitted and nontransmitted alleles. The equivalence of nontransmitted allele frequencies to population allele frequencies at the disease locus was evident in the original description of the TDT, although this property was not stated explicitly. Nontransmitted allele frequencies equal m+ (θδ/p) and 1 −m− (θδ/p) (Spielman et al. 1993) (where m is the frequency of a marker allele, θ is the recombination fraction between marker and disease loci, δ is the disequilibrium coefficient between marker and disease loci, and p is the frequency of the disease allele, which is in Hardy-Weinberg proportions). When θ= 0 between marker and disease locus (i.e., at the disease locus), the nontransmitted allele frequencies equal m and 1 −m, which are the population allele frequencies. (This is also a property of the AFBAC test when θ= 0 (Thomson, 1995)). The allele frequency among transmitted chromosomes will be affected by the ascertainment of severe malaria cases; however, the nontransmitted chromosomes were inherited by the parents as part of a random process that is unbiased by the study design. Thus, in the absence of segregation distortion, nontransmitted chromosomes can be used as population controls to estimate relative risk.

If we examine the allele frequencies of the parents' transmitted chromosomes only, they differ significantly from the population controls; however, the parents' nontransmitted chromosomes do not differ significantly from the population controls. Dissection of the parental chromosomes into transmitted and nontransmitted categories shows that the allele frequency of HbS among nontransmitted chromosomes is equivalent to the allele frequency of the population controls: both carry the HbS allele at a frequency of 9%. A comparison of transmitted versus nontransmitted chromosomes from the parents gives an odds ratio equivalent to a comparison of cases versus population controls.

Interestingly, a comparison of 630 parents of children with severe malaria against the population controls shows a significant difference in allele frequency. This effect is found in parents of both cerebral malaria patients and severe anaemia patients. One might expect that this reduction in HbS allele frequency in adults is due to mortality attributable to the HbSS genotype; however if the HbS allele is at equilibrium because of balancing selection, allele frequency in children and adults is expected to be equal, as AS heterozygotes rise in frequency and SS homozygotes fall. Upon closer inspection of the data, we discover that this significant reduction in HbS allele frequency is due primarily to the process of ascertaining children with severe malaria, who are less likely to carry the protective HbS allele.

Detecting Segregation Distortion

When segregation distortion favours the transmission of one allele over the other (i.e., because of selection at this locus prior to sampling), the null hypothesis of 50% transmission is no longer appropriate for the TDT statistic. By observing the transmission ratio in families unselected for disease, one can test for a significant departure from 50% transmission and develop an alternative null hypothesis for the TDT, as demonstrated by (Eaves et al. 1999). Unfortunately, it is often difficult and expensive to recruit complete families who represent the general population.

Because unrelated population controls are more easily recruited than families, it may be preferable to use population controls (age-matched to the TDT probands) to test for segregation distortion. Under the assumption of 50% transmission, nontransmitted allele frequencies among parents of affected offspring should equal the allele frequency among the population controls (Spielman et al. 1993; Thomson, 1995). When segregation distortion is active at a locus, it may manifest itself as a difference between nontransmitted parental allele frequency and allele frequency among the population controls. The underlying transmission ratio (T0) can be calculated using the formula inline image where a is the allele frequency in the population controls and b is the allele frequency among the nontransmitted chromosomes of parents of affected children. Our simulations of HWE populations with a locus showing segregation distortion (data not shown) suggest that the underlying transmission ratio can be estimated from nontransmitted parental alleles in a TDT study and population controls. The population controls should be age and ethnically matched to the probands of the TDT. In this study, population controls differed from TDT cases in ethnic composition; this potential confounder was corrected for in the stratified case-control analysis of the odds ratio, but remained a potential confounder of our estimate of the baseline transmission ratio.

Some have suggested using the transmission of alleles from parents to an unaffected sibling of an affected child as an estimate of the baseline transmission ratio (Spielman et al. 1993). There is debate over what the expected transmission ratio to unaffected siblings in TDT families should be (Lie et al. 2000; Eaves et al. 1999). This issue is further complicated by the challenge of identifying a sibling as “unaffected” (as opposed to “not-yet-affected”), and together make unaffected siblings an unreliable sentinel of segregation distortion. The TDT family plus age-matched population control study design may offer cost-effective security measures against the problem of segregation distortion in the TDT.

Conclusion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Electronic-Database Information
  9. Acknowledgments
  10. References

We have conducted a genetic association study investigating the protective effect of the sickle-cell trait against severe malaria in a multi-ethnic population in The Gambia. DNA samples from cases of severe malaria were compared against DNA from cord blood samples representing the general population in a case-control analysis stratified by ethnic group. The same cases of severe malaria were analysed by TDT using DNA samples from their parents. Our stratified case-control analysis confirms a strong protective effect of the sickle-cell trait against both cerebral malaria and severe malarial anaemia, with a ten-fold reduction in risk. TDT analysis provides the first family-based association data of the sickle-cell trait in severe malaria, and confirms the case-control findings with p < 10−9.

We report the allele frequency among nontransmitted parental chromosomes to show that, in this study, nontransmitted parental chromosomes carry the sickle cell allele at a frequency that is equal to the frequency in the general population. This equivalence provides two useful comparisons. First, an odds ratio can be estimated by comparison of transmitted and nontransmitted chromosomes among parents of affected offspring, a meaningful statistic for comparison with case-control studies. Second, a comparison of nontransmitted parental alleles against population controls allows detection of segregation distortion, a potential source of error for the TDT.

Analysis of the protective effect of the sickle-cell trait against severe malaria offers a comparison of case-control and family-based methods of detecting a genetic association in a complex disease; furthermore, this study demonstrates the potential of this sample collection to identify novel genetic factors which influence susceptibility to severe malaria.

Acknowledgments

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Electronic-Database Information
  9. Acknowledgments
  10. References

We thank the patients and their families, the staff on the Children's Ward of the Royal Victoria Hospital, Banjul; and Yaya Dibba, Momodou Saidykhan and Simon Correa for their expert assistance. We thank Dr. Melanie Newport for kindly providing the cord blood samples used in this study. This work was supported by the Medical Research Council U.K. and the Rhodes Trust.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Electronic-Database Information
  9. Acknowledgments
  10. References