I very much appreciate the thoughtful comments made by Drs. Brian and John Healy and, in particular, I am grateful for the time they took to read my paper carefully. However, while I agree with their theoretical construction of the various conditional probabilities in their letter, it seems from their comments that, possibly, I was not sufficiently clear in the paper about the fact that my construction was based less on pure theoretical considerations and more upon conclusions drawn from the available epidemiological evidence. In part, this may have been the result of dividing my logic between the text and the Appendix of the paper. In the following paragraphs, therefore, I have attempted to clarify my arguments and to address, hopefully, the concerns raised. The starting point of my analysis was that, in the general population, it should be possible to divide MS cases into two broad categories – those cases that developed MS through “genetic” pathways and those that developed MS through “non-genetic” pathways. In this context, the term “genetic” MS is used to indicate that the development of MS has occurred through a pathway requiring both a susceptible genotype (G) and specific environmental events (E). In this formulation, the terms (PG) and (PE) are conceptualized very broadly. Thus, (PG) refers to the probability of possessing any genotype that could possibly develop MS through the “genetic route” under some set of environmental exposures. Similarly, (PE) refers to the probability of experiencing any environmental exposure that could possibly produce MS under some selected set of genetic preconditions. The life-time probability that an individual will develop MS by the genetic route is termed (PMSG), whereas the corresponding probability for non-genetic MS is termed (PMSE). This pathway to non-genetic MS may include cases caused by either special environmental or purely stochastic events. Thus, my starting equation is: PMS = PMSG + PMSE – (PMSG)(PMSE) = 0.0015 (1a) where (0.0015) was chosen because it is the mid-point of the estimated prevalence range for MS in Canada (i.e., 0.1-0.2%). Because both (PMSE) and (PMSG) must be (≤ 0.0015), the cross-product term (the probability of getting MS through both routes) is negligible compared to the other two terms and can be ignored. If we define the proportion (p) as: PMSG = (p)(PMS) then Equation (1a) can be rewritten as: PMS = PMSG + PMSE = (p)(PMS) + (1-p)(PMS) = 0.0015 (2a) This equation is equivalent to Healy Equation (5), in which their first term represents (PMSG) and the sum of the last two terms represents (PMSE). However, I believe that only their assumption that P(MS|E1,G2)=0} is necessary for my construction. Their two further assumptions that {P(MS|E2,G1)=1 and P(MS|E2,G2)=1}seem (to me) unnecessary. My logic at this point in the analysis was to use actual epidemiological observations to estimate the value of (p). Two independent observations were used for this purpose. In considering the implications of these observations it is important to recognize that that essentially every case of concordant MS in monozygotic-twins (and also cases of concordant dizygotic-twins and siblings), represent individuals who are genetically susceptible. For example, because, as noted above, (PMSE < 0.0015), and because (PMSE │S ≈ PMSE) for the same reasons that lead to Equation (9a) below, the probability that a sibling (S) of an MS proband will get MS can be expressed as: PMS│S = 0.029 Therefore: = PMSG│S + PMSE│S < PMSG│S + 0.0015 PMSG│S > (0.95) (PMS│S) (3a) This percentage increases to much more than (95%) when either a more realistic estimate for (PMSE) is used or when considering the same circumstance for monozygotic (MZ) and dizygotic (DZ) twins where the observed concordance rates are considerably higher (0.25 and 0.054 respectively). Consequently, the conclusion that essentially all cases of concordant MS in monozygotic twins represent “genetic” MS is justified. In this circumstance, we can define (PMS*) as the probability of getting MS in the susceptible population (assuming, for the moment, that either the genetic profile of patients with clinical MS is similar to that of the susceptible population or that the average probability of getting MS is the same for the two groups), Equation (2a) can be written as: PMS│MZ = CRMZ = PMSG│MZ = (p)(PMS*) so that: PMS* = CRMZ / p (4a) The first epidemiological observation used to estimate (p) is the difference (Δ) in concordance rates for monozygotic twin-pairs, split according to whether or not they carry the HLA DRB1*1501 allele (Table). In the MS population of Canada (and in northern Europe), 55% of individual carry at least one copy of this allele compared to 24% in the general Canadian population (D. Sadovnick, personal communication). Because this particular allele has been clearly established as an MS-susceptibility allele, its excess in the MS population must be the result of “genetic” MS. Therefore this excess indicates that 31% (i.e., 55% – 24%) must represent a minimum value for (p) and, thus, that these individuals are genetically susceptible due, in part, to carrying the HLA DRB1*1501 allele. Presumably, some of the remaining 69% of MS cases are also genetically susceptible either due, in part, to this allele or for other reasons. Indeed, by reasoning similar to that leading to Equation (3a) above, the 28% proband-wise monozygotic-twin concordance rate for twin-pairs who lack this allele entirely (Table), clearly indicates the presence of other important genetic susceptibility factors. We will let (v) represent the proportion of MS patients who are not susceptible to MS due to possessing this allele, but who are still genetically susceptible for other reasons (0 ≤ v ≤ 1); we will let (r1) represent the probability of an individual possessing at least one copy of the HLA DRB1*1501 allele in the general population; and we will let (r2) represent the percent increase above the general population in the likelihood of an MS patient possessing this allele. Making the conservative assumption that HLA DRB1*1501 allele contributes to MS susceptibility only in those (r2) genotypes overrepresented in the MS population, leads to an estimate for (p) of: p = [(1 – r1 – r2)(v)] + [r2 + (r1)(v)] = [r2 + (1 – r2)(v)] (5a) Moreover, by Equation (4a), because all concordant MS in monozygotic twins is “genetic”, and assuming equal penetrance of different genotypes and using adjusted concordance rates (CRMZ(S)) as in Equation (14a) below, then: CRMZ(S)(HLA+) = [1/(r1 + r2)][r2 + (r1)(v)][CRMZ(S) / p] and: CRMZ(S)(HLA–) = [1/(1 – r1 – r2][ 1 – r1 – r2][v][CRMZ(S) / p] = [v][CRMZ(S) / p] Subtracting these two equations and substituting for (p) from Equation (5a) yields: Δ = [r2/(r1 + r2)][1– v][CRMZ(S)] / [r2 + (1 – r2)(v)] or: v = [(r2/{r1 + r2})(CRMZ(S)) – Δ(r2)] / [(r2/{r1 + r2})(CRMZ(S)) + Δ(1 – r2)] (6a) Because this is a linear system of two equations in two unknowns (v and p), it can be solved uniquely. As discussed below in the arguments leading to Equation (14a), we need to use adjusted values for CRMZ and Δ (i.e., CRMZ(S) = 0.156 and Δ = 0.01), and substitute these together with the observed values of (r1 = 0.24), and (r2 = 0.31) into Equation (6a), to get the point estimate for (v) of (v = 0.89). Substituting this value back into Equation (5a) gives the point estimate for (p) of: p = 0.93 This analysis considers only the possibility that the HLA DRB1*1501 allele contributes to MS susceptibility to the extent that persons with this allele are overrepresented in the MS population (i.e., in 31% of MS cases). If this allele, when present, contributes more often than this, the estimated of the proportion (p) will increase. The second method used to estimate the proportion (p) of genetic MS was based on a population-based survey of monozygotic twins in Finland. Using an unadjusted 25% concordance rate for monozygotic twins of an MS proband, together with the knowledge that essentially all twin pairs concordant for MS have “genetic” MS (see Equation 3a), permits the estimation of PMSG as four times the prevalence of concordant twin-pairs in the cohort. Although this analysis is based on only a small sample, using the data from this cohort, the estimated prevalence of “genetic” MS is (817/105 population). As pointed out in the original paper, this number greatly exceeds the reported prevalence of MS in the general population of Finland. It also exceeds the total estimated prevalence of MS (298/105 population) taken directly from this cohort. Consequently, both of these analyses of existing data suggest that the large majority of MS (perhaps all) occurs by the route of genetic susceptibility together with an appropriate environmental exposure and, therefore, suggest that the assumption (p ≈ 1) is a reasonable approximation. It was on this basis (i.e., assuming that: p ≈ 1) that my Equation (1) was developed and formulated. In this circumstance, the joint probability P(MS,G,E) should be equal to the marginal probability P(MS) and, thus, under these conditions, my Equation (1) and Healy Equations (1) and (5) are all the same. Moreover, even if (p < 1), PMSG would be smaller than I have assumed and, thus, the estimated proportion of the population who are genetically susceptible will be even less than that proposed. I agree with Drs. Healy, that my Equation (3) requires adjustment. However, contrary their assertion, I did provide the experimental data necessary to make this adjustment. Thus, as I pointed out in the paper, the 5.4% recurrence risk for MS in dizygotic twins (CRDZ) of an MS proband is higher than the 2.9% recurrence risk in nontwin siblings (CRS). Moreover, I argued that, because several experimental studies have failed to identify any differences in MS risk among adopted individuals, conjugal couples, brothers and sisters of different birth order, and in siblings and half-siblings raised together or apart, this difference between dizygotic twins and siblings must reflect the impact of a shared intra-uterine or early post-natal environment on MS risk. Consequently, my Equation (2) should have been expressed as: (PG│MZ)(PE│G, MZ)(PMS│G, MZ, E) = PMS│MZ which can be simplified to: (PE│MZ)(PMS│MZ, E) = CRMZ (7a) where (PE│MZ) is the conditional probability of exposure given that the individual is a monozygotic (MZ) twin of an MS proband and (PMS│MZ, E) is the conditional probability of developing MS given that the person is a monozygotic twin and has received a sufficient environmental exposure. The relationship between (PE│MZ)(PMS│MZ, E) and (PE│G)(PMS│G, E) needs to be determined from existing data regarding the impact of a shared intrauterine or early postnatal environment on the likelihood of MS. Fortunately, this impact can be estimated from: CRDZ = (CRDZ/CRS)(CRS) = (5.4 / 2.9) (CRS) = 1.86 (CRS) Because (PG) is the same for both twin and non-twin siblings, then: CRDZ = (PG)(PE│DZ)(PMS│DZ, E) = 1.86 (PG)(PE│S)(PMS│S, E) (8a) Where the different conditional probabilities are defined for the dizygotic (DZ) and sibling (S) cases in the same manner as terms of Equation (7a) were defined for the MZ case. Moreover, as discussed above, there seems to be no change in the risk of environmental exposure due to siblings sharing their childhood environment with the MS proband compared to the same risk in siblings growing up in an environment experienced by unrelated individuals in the general population. Similarly, there seems to be no change in the risk of environmental exposure due to an unrelated individual sharing their childhood environment with an MS proband compared to their risk growing up in the general population. Thus, it seems that the observed difference in MS risk between nontwin siblings and members of the general population is related to their genetic make-up (i.e., the PG term) and not their environmental exposure terms (other than the shared intrauterine effect noted above). Consequently, it seems reasonable to conclude that: (PE│S) = (PE│G) (9a) In my paper, I made the further assumption that: (PMS│S, E) ≈ (PMS│G, E) (10a) and therefore that: CRDZ = (PG)(PE│DZ)(PMS│DZ, E) ≈ 1.86 (PG)(PE│G)(PMS│G, E) (11a) Dividing out the common, but unknown, (PG) term from Equation (11a) yields: (PE│DZ)(PMS│DZ, E) ≈ 1.86 (PE│G)(PMS│G, E) (12a) Using this same estimate for the impact of a shared intra-uterine and early post-natal environment in monozygotic twins, yields: (PE│MZ)(PMS│MZ, E) = CRMZ ≈ 1.86 (PE│G)(PMS│G, E) (13a) Using Equation (13a), we can define an adjusted concordance rate (CRMZ(S)), removing the intrauterine/early postnatal effect, as: CRMZ(S) = CRMZ / 1.86 = 0.25 / 1.86 = 0.134 (14a) Using Equations (13a) and (14a) above, Equation (3) from the original paper should properly have been expressed as an adjusted rate of: PG ≈ (PMS / CRMZ(S)) = 0.0015 / 0.134 = 1.1% (15a) Nevertheless, I agree that my assumption in Equation (10a) requires justification. Specifically, the concern is that, given two different susceptible genetic profiles (G1 and G2), although the epidemiological observations leading to Equation (9a) suggest that: (PE│ G1) = (PE│ G2) they may not necessarily also suggest that: (PMS│ G1, E) = (PMS│ G2, E) In essence, I believe that this is same concern expressed by Drs Healy in their discussion surrounding the development of their equation Healy (3), in which they defined the set G as representing “the set of genetic profiles for the monozygotic twins.” Unlike the situation for dizygotic twins, however, the tendency to have monozygotic twins is not inherited, so that monozygotic twins should have the same distribution of genotypes (i.e., the same genetic profile) as the general population. Therefore, I suspect that Drs. Healy meant to define the set G as “the set of genetic profiles for individuals known to have a monozygotic twin with MS.” In this circumstance (i.e., if the penetrance term, [PMS│G, E], is different for different genotypes), then the genetic profile of such individuals will be different from that of the general population. To consider the impact of this possibility, I will let (x) represent the number of susceptibility loci (i.e., loci that harbor susceptibility alleles) in addition to the known HLA DRB1 locus. I will let (X) be the total number of possible combinations of alleles at the (x + 1) susceptibility loci, P(i) be the probability of the ith combination in the general population, and (PMSi*) be the conditional probability of developing MS given that the person has the ith genetic combination (Gi). Thus: PMSi* = (PE│Gi)(PMS│Gi, E) If each locus combines uniquely with other loci, then: x+1 X = Σ (x + 1)! / (x + 1 – i)! (i)! i=1 If, however, different susceptible states at one locus can combine with different susceptible states at other loci, then the number of possible combinations will be much greater. Regardless, letting (X) be the total number of possible combinations yields: X PG Σ [(P(i)] | (PMSi* > 0) = i=1 X PMS | G = (Σ [PMSi*][(P(i)] ) / (PG) (16a) i=1 Thus, (PMS | G) is a weighted average of the different conditional non-zero (PMSi) terms. Clearly, if all non-zero (PMSi) terms are approximately equal, then: PMS | G = PMSi* and estimate of (1.1%) from Equation (15a) will be approximately correct. However, if the different conditional (PMSi) terms are not equal, this may affect the estimated prevalence of genetic susceptibility. For example, consider the circumstance in which the different genotypes are grouped into two classes of susceptibility genotypes – the first (G1) being those genotypes with an expected penetrance higher than average and the second (G2) being those with an expected penetrance lower than average. For the purpose of this example, we will assume that the penetrance within each class is approximately the same for all genotypes. Again assuming that essentially all MS is “genetic”, we will define (PG1) and (PG2) as the probabilities of these two different classes of genotype in the general population so that: PMS ≈ PMSG1 + PMSG2 where: PMSG1 = (PG1)(PE│G1)(PMS│G1, E) and: PMSG2 = (PG2)(PE│G2)(PMS│G2, E) For simplicity, we will also define (PMSG1*) and (PMSG2*) respectively as the probabilities either that an individual in the G1 subset will develop MS when (PG1 = 1) or that an individual in the G2 subset will develop MS when (PG2 = 1). Thus: PMSG1* = (PE│G1)(PMS│G1, E) and: PMSG2* = (PE│G2)(PMS│G2, E) In the circumstance of a monozygotic twin, where (PG1) and (PG2) are the respective probabilities of class-membership in the general population, we will define (PG1app) and (PG2app) as these (apparent) probabilities in the population of monozygotic- twins in which the proband is known to have MS. Because (PMS ≈ PMSG) it is apparent that, under any circumstance, including those of Equation (3a), essentially all monozygotic twins of an MS proband are genetically susceptible (i.e., PGapp ≈ 1) and, therefore, that: PGapp = PG1app + PG2app = 1 and also that: PG1app = [(PMSG1*) (PG1)] / [(PMSG1*) (PG1) + (PMSG2*) (PG2)] PG2app = [(PMSG2*) (PG2)] / [(PMSG1*) (PG1) + (PMSG2*) (PG2)] or: PG1app = [(PMSG1*) (PG1)] / PMS (17a) PG2app = [(PMSG2*) (PG2)] / PMS From Equation (16a): PMS | G = [(PMSG1*)( PG1app) + (PMSG2*)( PG2app)] / (1) = CRMZ(S) where: (18a) 1 ≥ PMSG1* ≥ CRMZ(S) ≥ PMSG2* > 0 This leads to the linear system of two equations in two unknowns: (PMSG1*)(PG1app) + (PMSG2*)(PG2app) = 0.134 (19a) PG1app + PG2app = 1 In the case of two classes of susceptibility loci, this system can be solved uniquely as: and: PG1app = (0.134 – PMSG2*) / (PMSG1* – PMSG2*) (20a) PG2app = 1 – PG1app = (PMSG1* – 0.134) / (PMSG1* – PMSG2*) (21a) In the general population, where the estimated prevalence of MS is 0.15%, then: (PMSG1*)(PG1) + (PMSG2*)(PG2) = 0.0015 or: PG1 = [0.0015 – (PMSG2*)(PG2)] / PMSG1* (22a) and: PG2 = [0.0015 – (PMSG1*)(PG1)] / PMSG2* (23a) Equating the RHS of Equations (17a) and (20a), using Equation (23a) to substitute for (PG2), and solving for (PG1) yields: PG1 = [(0.0015)(0.134 – PMSG2*)] / [(PMSG1*)(PMSG1* – PMSG2*)] (24a) Equation (24a) cannot be solved uniquely but can still provide an estimate of how the spread of the expected penetrance between the two classes of susceptible genotypes might affect the total percentage of the population that is capable of developing MS (i.e., PG). For example, if the penetrance of either class is equal to CRMZ(S) (i.e., 0.134) the percentage of the population who are members of the other class is (0%) and the estimate of (PG = 1.1%) is accurate. When (PMSG1*) and (PMSG2*) are both removed from CRMZ(S), and also are somewhat removed from each other (e.g. PMSG1* = 0.2; PMS2 = 0.07), the estimated percentage of susceptible individuals is barely altered (i.e., PG = 1.5%). Even when they are removed by one order of magnitude (e.g., PMSG1* = 0.3; PMS2 = 0.03), the estimated percentage of susceptible individuals is still quite small (i.e., PG = 3.3%). It is only when (PMSG2*) approaches zero that the estimate for (PG) begins to increase more markedly. Thus, when (PMSG1*) and (PMSG2*) are removed from CRMZ(S) by two orders of magnitude (e.g., PMSG1* = 0.3; PMSG2* = 0.003), the estimated percentage of susceptible individuals becomes (PG = 28.2%). Nevertheless, there are also other important constraints on this system. Thus, because, by definition, PG1 and PG2 are mutually exclusive, and because (PMS ≈ PMSG), Equation (1a) becomes: PMS ≈ PMSG = (PG1)(PMSG1*) + (PG2)(PMSG2*) = PMSG1 + PMSG2 (25a) We will define (p) to be the proportion of susceptible individuals in the general population who belong to the G1 subset (0 ≤ p ≤ 1). Thus: PG = PG1 + PG2 = (p)(PG) + (1 – p)(PG) (26a) In the case of identical genotypes outside the IU environment, by Equations (17a), (25a), and (26a), Equation (18a) can be rewritten as the quadratic equation: (p)(PMSG1*)2(PG)/PMS + (1 – p)(PMSG2*)2 (PG)/PMS = CRMZ(S) = 0.134 (27a) The proportion of concordant monozygous twin-pairs that belong to the (G1) subset (PMZG1) is: PMZG1 = (p)(PMSG1*)2 / [(p)(PMSG1*)2 + (1 – p)(PMSG2*)2] (28a) Multiplying Equation (27a) by (PMZG1) yields: (p)(PMSG1)2(PG)/(PMS) = (PMZG1)(CRMZ(S)) = (PMZG1)(0.134) (29a) Also, as before, we can make use of the fact that the proband-wise monozygotictwin concordance-rates for probands who carry the HLA DRB1*1501 allele (CRMZ(S)[HLA+]) and for probands who do not carry this allele (CRMZ(S)[HLA–]) are substantially the same (0.16 and 0.15 respectively) and that the overall (i.e., combined) estimated proband-wise monozygotic-twin concordance-rate from this HLA-typed cohort is (CRMZ(S) = 0.156). Moreover, because HLA DRB1*1501 is the best-established and most consistently identified susceptibility allele in MS, because it is present in 55% of MS patients, and finally because genotypes with this allele have a penetrance at least as large as the average penetrance in the susceptible population (Table), those genotypes, in which this allele contributes to susceptibility, presumably belong to the G1 subset. Again, we will make use of the difference (Δ) in penetrance between genotypes with or with out the HLA DRB1*1501 allele; we will let (v) represent the proportion of MS patients who are not in the G1 subset due to having this allele, but who are, nonetheless, still in the G1 subset (0 ≤ v ≤ 1); we will let (r1) represent the probability of an individual possessing at least one copy of the HLA DRB1*1501 allele in the general population; and we will let (r2) represent the percent increase above the general population in the likelihood of an MS patient possessing this allele. Also, as before in Equation (5a): p = [(1 – r1 – r2)(v)] + [r2 + (r1)(v)] = [r2 + (1 – r2)(v)] (30a) Because, in the Canadian population, (r2 = 0.31), this requires that (p ≥ 0.31). Putting this constraint into Equation (28a) yields PMZG1 ≥ PMZG1min = (0.31)(PMSG1*)2 / [(0.31)(PMSG1*)2 + (0.69)(PMSG2*)2] where (PMZG1min) is the minimum value that (PMZG1) can take. Therefore, Equation (29a) can be rewritten as: or: (p)(PMSG1*)2(PG)/PMS ≥ (PMZG1min)(CRMZ(S)) (31a) PMSG1* ≥ [(PMZG1min)(CRMZ(S))(PMS) / (p)(PG)]1/2 (32a) Moreover, assuming approximately equal penetrance of the different genotypes within each of the two subsets, then: CRMZ(S)(HLA+) = [1/(r1 + r2)][r2 + (r1)(v)][ PMSG1*] + [r1 / (r1+ r2)][1–v][PMSG2*] and: CRMZ(S)(HLA–) = [1/(1 – r1 – r2][ 1 – r1 – r2][v][ PMSG1*] + [1-v][PMSG2*] = [v][ PMSG1*] + [1-v][PMSG2*] Subtracting these two equations yields: Δ = [r2/(r1 + r2)][1– v][PMSG1* – PMSG2*] (33a) As shown earlier using Equation (24a), if (PMSG1*) and (PMSG2*) are even close to each other in magnitude, the estimate for (PG) remains very small. It is only when (PMSG1* >> PMSG2*) that this estimate increases substantially. Nevertheless, in this circumstance, Equation 33a becomes: Δ ≈ [r2/(r1 + r2)][1– v][PMSG1*] Letting: R = r2/(r1 + r2) and: X = (PMZG1min)CRMZ(S))(PMS / PG) so that: (34a) PMSG1* = [X / p]1/2 and substituting both the value of (p) from Equation (30a) and the minimum value for PMSG1* from Equation (32a), Equation (34a) can be rearranged into the quadratic equation: [(R2)(1/ Δ2)(X)] v2 – [(1 – r2) + (2)(R2)(1/ Δ2)(X)] v – (r2) + [R2][1/ Δ2] [X] = 0 (35a) Solving Equation (34a), using the assigned values of (PMSG1* and PMSG2*), and substituting the observed values of (Δ = 0.01), (CRMZ(S) = 0.156), (r1 = 0.24), (r2 = 0.31), and (PMS = 0.0015), together with the value of (PG) from Equations (23a) and (24a) yields a minimum estimated (v) and, after substituting this back into Equation (30a), a minimum estimated (p) for any two pairs of penetrance values (PMSG1* and PMSG2*). This derived minimum expected value of (p) can then be compared to the actual value of (p) determined from Equations (23a and 24a) to assess whether the proposed pair of values is plausible. For example, in circumstances considered earlier (i.e., PMSG1* = 0.3 and PMSG2* = 0.03), the approximation of Equation (35a) holds, and Equations (30a) and (35a) yield an estimated minimum value for (p) of (p = 0.86) compared to the actual value of (p = 0.06) for this pair of penetrance values. This is obviously an implausible circumstance. In fact, for (PMSG1* ≥ 0.2), there are no values for (PMSG2*) over its entire possible range of (0 < PMSG2* ≤ 0.134) for which the actual value of (p) even reaches its minimum value of (p = 0.31). In this circumstance (i.e., PMSG1* = 0.2), the best fit is at (PMSG2* = 0.07), although, even here, the actual (p) is only 25%, which is still lower than its minimum value of (p ≥ 0.31). It is only when (PMSG1*) approaches CRMZ(S) that the two values of (p) approximate each other and, even then, only when the magnitude of (PMSG2*) is quite small (e.g., PMSG1* = 0.135 and PMSG2* = 0.01; p > 0.9 for both estimates,; and PG = 1.2%). Moreover, in all of these circumstances, the estimated (PG) from Equations (23a) and (24a) for the optimal fit is never exceeds (1.5%). Consequently, in any situation, either Equation (35a) approximately holds or (PMSG1*) and (PMSG2*) are close enough in magnitude so that that the estimate of (PG) is quite stable. Therefore, the approximation of Equation (10a) is justified and the conclusion that only a very small fraction of the general population is genetically susceptible to MS is secure. Moreover, this percentage can be approximated as: PG ≈ PMS / CRMZ(S) Also, presuming the impact of a shared intrauterine/early postnatal environment on twin concordance is similar in different locations, the prevalence of MS susceptibility in different regions of the world can be compared using unadjusted concordance rates: PMS / CRMZ Table. Concordance rates for MS in Monozygotic Twins of HLA DRB1*1501-positive (ZH+) and HLA DRB1*1501-negative (ZH-) Probands *. Monozygotic Twins of MS Probands HLA DRB1*1501 HLA DRB1*1501 Positive Negative Concordant (C) 9 11 20 Discordant (D) 31 42 73 Totals 40 53 93 ZH+ = (9/40) = 23% ZH- = (11/53) = 21% ZH+ = 30% ZH- = 28% ZH+ = 16% ZH- = 15% Pair-wise Totals Concordance† Proband-wise Concordance†† Adjusted Probandwise Concordance††† Data derived from: Willer et al., Proc Natl Acad Sci (USA) 2003;100:12877-12882. † Pair-wise rates calculated as (Z = C/(C + D). †† Proband-wise concordance rates calculated as (Z = 2C/(2C + D) adjusted for the overall probability of doubly ascertaining an affected twin (70%) in the Willer, et al., 2003 study. ††† See Text, Equation (14a)