1. Preliminary analyses in the National
Collaborative Perinatal Project

2.  Levels of intervention
◦ Tertiary: Aimed at reducing the chronic effects of
established disease.
◦ Secondary: Aimed at curing disease.
◦ Primary: Preventing disease.

3.  (insert pie chart) which includes:
◦ Peri disease factors
◦ Young adult factors
◦ Childhood factors
◦ Genetic factors

4.  Ecologic
◦ High cardiovascular death rates in 1950 correlated with
high infant mortality in the early 1900’s.
 In an area of high infant mortality, “survivors” were still
comparatively stressed.
 Retrospective cohorts
◦ Nurses Health study, etc.
◦ Common feature- BW used as a surrogate for adequacy
of the intrauterine environment

5.  Heart attack
 Stroke
 Hypertension
 Diabetes
 Obesity, syndrome X
 Breast cancer
 Osteoporosis

6.  Does size matter?
◦ 2500 g BW 16g heart
◦ 3000 g 22 g heart
 (37.5% more heart)

7.  (insert chart)

8.  (insert chart)

9.  The thrifty phenotype?

10.  29621 singleton liveborn infants
 Standard placental measures
◦ Shape
◦ Larger and smaller diameters
◦ Thickness
◦ Distance from cord insertion to the nearest disk edge
◦ Umbilical cord length
◦ Placental weight
 Outcome variable: BW

11.  Disk shape
 Larger and smaller diameters
 Distance from cord insertion to disk edge
 Cord length
 Disk thickness
 Placental weight

12.  Birth weight
 BMI at ages 4 and 7 years

13.  (insert images)

14.  (insert images)

15.  Normal- 500 g
 Abnormal 1st-2nd trimester, 500 g
 Abnormal from 1st-2nd trimester, 250 g
 Abnormal 3rd trimester, 400 g

16.  Linear terms
◦ Higher order terms
◦ Interactions
 Local solutions (MARS 2.0)

17.  Indirect/Direct effects on BW
 Indirect/Direct effects on PW
 (insert image)

18.  Pick 10 different random seeds
 Get at least 9 different patterns of significant
higher order terms and/or interactions
 Inspection of distributions suggests
differences lie with outlier partitioning
 “significant” terms generally negative,
indicating floor/ceiling effects

19. And so to MARS 2.0

20.  (insert graphs)

21.  (insert graph)
 No evidence for placental “senescence”
 “Post maturity syndrome”
 “Fetal intolerance to labor”

22.  Fetal-placental weight ratio
◦ How many grams of baby supported by each gram of
placental
◦ ~7:1 at term
◦ Too low placental dysfunction
◦ Too high↓ placental reserves
 Ponderal index
◦ Fatness ratio (weight*100/length^3)
◦ Third trimester weight gain

23.  Fetoplacental weight ratio
 (insert graphs)

24.  (insert graphs)

25.  (insert table)

26.  (insert table)

27.  (insert graphs)
 Interpretation: the cord length-age 7 BMI
association is independent of PW

28.  (insert graphs)
 Cord length BW Age 7 BMI
 Cord length Age 7 BMI= only at extreme
cord lengths

29.  (insert regression results)
◦ N= 36
◦ N~ 1000
◦ N~ 25,000

30.  (insert regression results)
 N=240
 N~ 23,000
 N~ 3,000

31.  Recode BW into groups
◦ <2500 g: “Intrauterine growth restriction”
 Suggests chronic intrauterine deprivation
◦ >4000 g: Macrosomia
 Diabetic type metabolic pathology
◦ “Normal”
 Can placental growth classify pathologic fetal
growth that continues into childhood?

32.  Set penalties for misclassification
 Score data
 Incorrectly scored children are either bigger
or smaller than expected given their placental
dimensions
◦ They don’t “fit” their placentas
 Does “lack of fit” mark children’s growth
trajectory?

33. Class N Cases N % Error Cost
Mis-Classed
0 2,023 175 8.65 175
1 17,946 17,946 100.00 179,460
2 1,093 61 5.58 61

34.  (insert graph)

35. Observed <2500 2500-4000 >4000
Predicted
<2500 10.8 11.1 11.4
2500-4000 10.0 11.4 12.0
>4000 10.9 11.4 11.8

36.  (insert correlations)

37.  The placenta’s purpose is to “make a baby”
 Values of placental dimensions where the
placenta is bigger but the baby is not
“unbalanced”
 “Unbalanced” babies may be physiologically
vulnerable and may have different childhood
growth trajectories

38.  Your placenta provides all oxygen and
nutrients and genetics aside is the principal
determiner of fetal growth
 If you are bigger or smaller than your
placenta predicts, you don’t “fit”
 Children who don’t “fit” may be
physiologically vulnerable and have different
childhood development trajectories

39.  MARS and RandomForest allow
complementary and unique insights into how
placental growth is translated into fetal
growth
 Placental measures can be used to
characterize the fetal environment, with
physiologic and time-order inferences that
may be important to “fetal origins” research