Sudden infant death syndrome (SIDS) victims were regarded as normal as a matter of definition (Beckwith 1970) until 1952 when Kinney and colleagues argued for elimination of the clause, “unexpected by history.” They argued that “not all SIDS victims were normal,” and referred to their hypothesis that SIDS results from brain abnormalities, which they postulated “to originate in utero and lead to sudden death during a vulnerable postnatal period.” Bergman (1970) argued that SIDS did not depend on any “single characteristic that ordains a infant for death,” but on an interaction of risk factors with variable probabilities. Wedgwood (1972) agreed and grouped risk factors into the first “triple risk hypothesis” consisting of general vulnerability, age-specific risks, and precipitating factors. Raring (1975), based on a bell-shaped curve of age of death (log-transformed), concluded that SIDS was a random process with multifactorial causation. Rognum and Saugstad (1993) developed a “fatal triangle” in 1993, with groupings similar to those of Wedgwood, but included mucosal immunity under a vulnerable developmental stage of the infant. Filiano and Kinney (1994) presented the best known triple risk hypothesis and emphasized prenatal injury of the brainstem. They added a qualifier, “in at least a subset of SIDS,” but, the National Institute of Child Health and Development SIDS Strategic Plan 2000, quoting Kinney’s work, states unequivocally that “SIDS is a developmental disorder. Its origins are during fetal development.” Except for the emphasis on prenatal origin, all 3 triple risk hypotheses are similar.

Interest in the brainstem of SIDS victims began with Naeye’s 1976 report of astrogliosis in 50% of all victims. He concluded that these changes were caused by hypoxia and were not the cause of SIDS. He noted an absence of astrogliosis in some older SIDS victims, compatible with a single, terminal episode of hypoxia without previous hypoxic episodes, prenatal or postnatal.

Kinney and colleagues (1983) reported gliosis in 22% of their SIDS victims. Subsequently, they instituted studies of neurotransmitter systems in the brainstem, particularly the muscarinic (1995) and serotenergic systems (2001). The major issue is when did the brainstem abnormalities, astrogliosis, or neurotransmitter changes occur and whether either is specific to SIDS. There is no published method known to us of determining the time of origin of these markers except that the injury causing astrogliosis must have occurred at least 4 days before death (Del Bigio and Becker, 1994). Because the changes in neurotransmitter systems found in the arcuate nucleus in SIDS victims were also found in the chronic controls with known hypoxia, specificity of these markers for SIDS has not been established. It seems likely that the “acute control” group of Kinney et al (1995) died too quickly to develop gliosis or severe depletion of the neurotransmitter systems. We can conclude that the acute controls had no previous episodes of severe hypoxia, unlike SIDS or their “chronic controls.” Although the average muscarinic cholinergic receptor level in the SIDS victim was significantly less than in the acute controls, the difference was only 27%, and only 21 of 41 SIDS victims had values below the mean of the acute controls. The study of the medullary serotonergic network by Kinney et al (2001) revealed greater reductions in the SIDS victims than in acute controls, but the questions of cause versus effect of the abnormalities, and whether they occurred prenatally or postnatally, remain unanswered.

Hypoplasia of the arcuate nucleus was stated to occur in 5% of their SIDS cases by Kinney et al (2001), but this is a “primary developmental defect” according to Matturri et al (2002) with a larger series, many of whom were stillbirths. These cases should not be included under the rubric of SIDS, by definition.

There are difficulties with Filiano and Kinney’s (1994) explanation of the age at death distribution of SIDS. They postulate that the period between 1 and 6 months represents an unstable time for virtually all physiologic systems. However, this period demonstrates much less instability than does the neonatal period, when most deaths from congenital defects and severe maternal anemia occur. We present data for infants born to mothers who were likely to have suffered severe anemia as a consequence of placenta previa, abruptio placentae, and excessive bleeding during pregnancy; these infants presumably are at increased risk of hypoxia and brainstem injury. The total neonatal mortality rate in these 3 groups of infants is 4 times greater than the respective postneonatal mortality, and in the postneonatal period the non-SIDS mortality rate is between 14 and 22 times greater than the postneonatal SIDS rate in these 3 groups. A preponderance of deaths in the neonatal period is also found for congenital anomalies, a category that logically should include infants who experienced prenatal hypoxia or ischemia; this distribution of age of death is very different from that for SIDS, which mostly spares the first month and peaks between 2 and 3 months of age.

Finally, evidence inconsistent with prenatal injury as a frequent cause of SIDS comes from prospective studies of ventilatory control in neonates who subsequently died of SIDS; no significant respiratory abnormalities in these infants have been found (Waggener et al 1990; Schectman et al 1991).

We conclude that none of the triple risk hypotheses presented so far have significantly improved our understanding of the cause of SIDS. Bergman’s and Raring’s concepts of multifactorial causation with interaction of risk factors with variable probabilities is less restrictive and more in keeping with the large number of demonstrated risk factors and their varying prevalence. If prenatal hypoxic damage of the brainstem occurred, it seems likely that the infant so afflicted would be at risk for SIDS, but it is even more likely that their death would occur in the neonatal period, as we have demonstrated in infants who have known maternal risk factors that involve severe anemia. This is in contrast to the delay until the postneonatal period of most SIDS deaths. A categorical statement that the origin of SIDS is prenatal is unwarranted by the evidence. Brainstem abnormalities have not been shown to cause SIDS, but are more likely a nonspecific effect of hypoxia.

## Comments

## Gender: a missing factor in the triple risk models for SIDS

Guntheroth and Spiers (2002) do not note that all the triple-risk models they cite are without specific gender-dependent risk factors, so they cannot predict the global 50% male excess of SIDS per 1000 live births of each gender [e.g., USA, 57,624 male, 37,880 female, male fraction = 0.603 (CDC, 2002); Albania, 151 male, 94 female, male fraction = 0.616 (Godo et al., 2002)]. Consequently any triple-risk hypothesis to explain SIDS that only has gender-independent risk factors must be incomplete (Mage and Donner, 1996; 2002).

The authors also discuss triple-risk models for SIDS in relation to the form of the SIDS age distribution and uncritically accept a 2- parameter lognormal model for the SIDS age distribution to describe it (Raring, 1975). But they present it as a paradigm of a triple-risk “multifactorial causation” for SIDS which by definition must contain 3 or more adjustable age-parameters, at least one for each of the three independent risk factor that varies with age.

By definition (Hahn and Shapiro, 1967), any probability distribution for a SIDS risk factor, P(x) at age x, must satisfy the integral (Int) condition,

Int [P(x) dx] = 1 (0 < x < infinity). (1)

Raring's lognormal model for the SIDS age distribution has two adjustable parameters, with the age appearing in the denominator as 1/x (Aitchison and Brown, 1972). Because it would go to infinity at birth as x goes to zero, a risk factor with only an age term of 1/x would violate Equation 1. However, the numerator of the lognormal distribution has another age term in a negative exponential function of the logarithm of the age squared, that goes to zero faster than the denominator as x goes to zero. Thus, it must be part of the same risk factor with the 1/x term which leads to P(x) = 0 at birth. Therefore this lognormal model with only two independent terms involving age may apply to a single risk factor for SIDS that rises rapidly from zero at birth and then falls slowly towards zero with age, without any upper age limit, and thus it cannot be a product of three independent risk factors. The authors correctly noted that Raring’s lognormal model for SIDS ages plots as a straight line on probability paper, but did not recognize that such a line incorrectly predicts SIDS can occur at any age throughout life. Had they done so the lognormal age model could have been rejected immediately on its face because no valid SIDS age model should predict SIDS occurring beyond infancy into adulthood. For any probability model to represent a “triple- risk” it would require at least three age parameters, one or more for each of the three hypothesized independent risk distributions that all must satisfy Equation 1.

The authors are apparently unaware that this antinomy of irreconcilable contradiction between a two-parameter SIDS age model and a triple-risk SIDS causation model can be resolved by use of a Johnson SB model to describe the age distribution (Johnson, 1949; Hahn and Shapiro, 1967). The SB probability model for SIDS ages has four adjustable parameters that allow for modeling of three age-dependent risk factors satisfying Equation 1, that together may lead to a fatal cerebral anoxia. They might vary as increasing with age (e.g., risk of infection), decreasing with age (e.g., risk of neurological prematurity), and rising and falling with age (e.g., risk of infant anemia as adult hemoglobin slowly replaces rapidly depleted fetal hemoglobin), and together they predict zero SIDS beyond an age of 3.5 years(Mage, 1996; 2001).

In conclusion, a gender-related risk factor needs to be added to the triple-risk models the authors discussed in order to predict the 50% preponderance of male SIDS in the global data base, and the 2-parameter lognormal distribution is an invalid model for the SIDS age distribution.

References: Aitchison J, Brown JAC. The Lognormal Distribution, Cambridge, Cambridge University Press, 1972: 8.

CDC. Infant mortality data, 1979-1998, http://wonder.cdc.gov, Accessed November 10, 2002.

Godo A, Pano A, Veveca E, Kuli Gj, Caushi N, Cenko F. Osservazioni epidemiologici sulla SIDS in Albania (in Italian). Abstract, The Seventh SIDS International Conference, Firenze, August 31 - September 4, 2002., Conference Handbook, 14-15.

Guntheroth WG, Spiers PS. The triple risk hypotheses in sudden infant death syndrome. Pediatrics 2002; 110: e64.

Hahn GJ, Shapiro SS. Statistical Models in Engineering, New York, John Wiley, 1967: 33, 213.

Johnson, NL. Systems of frequency curves generated by methods of translation. Biometrika 1949; 36: 149 - 182.

Mage DT. A probability model for the age distribution of SIDS. Journal of Sudden Infant Death Syndrome and Infant Mortality 1996; 1: 13- 32 (1996).

Mage DT, Donner M. A genetic basis for the sudden infant death syndrome sex ratio. Medical Hypotheses 1997; 48: 137-142.

Mage DT. Antinomy and the SB model for SIDS. Epidemiology. 2001; 12: 471.

Mage DT, Donner M. Is SIDS an X-linked recessive condition? A four- factor risk model. Abstract, The Seventh SIDS International Conference, Firenze, August 31 - September 4, 2002., Conference Handbook, p. 122.

Raring RH. Crib Death: Scourge of Infants - Shame of Society. Hicksville NY: Exposition Press; 1975: 93-97.