The nourishment a baby receives from its mother and its exposure to infection after birth determine its susceptibility to disease in later life(1). Animal studies have shown that undernutrition in early life program the body’s structure and metabolism(1). Programming describes the process where a stimulus or insult at a critical period of development has lasting or lifelong significance (2). Fetal undernutrition  results in impaired growth in the womb and during infancy and it is marked by low birth weight, placenta size, length, and head circumference which are all factors examined in determining fetal origin of disease. Undernutrition causes the fetus to adapt by compromising development of certain organs or processes to ensure its continued survival and growth at the expense of longevity (1). Impaired growth in the womb and during infancy is followed by coronary heart disease, hypertension, non-insulin dependent diabetes, raised serum cholesterol and abnormal blood clotting later in life(1). 
 Most of studies on fetal origin of chronic disease have had to rely on crude markers such as low birth weight, placenta size and head circumference to determine maternal nutrition. Much of the research has arisen from isolated areas due to record unavailability. The purpose of this paper is to discuss how certain sampling techniques can be applied to a study of fetal origin of disease to enhance the research
 For a brief overview of the sampling topics please visit:
Nonprobability Sampling
Cluster Probability
Sample Selection Within a Household
Design Effects
Random Number Generation 
 

Links to Fetal Origin of Disease 

     Nonprobability sampling would be a useful way to approach studying the fetal origin of disease because it would enable the investigator to select individuals based on appropriate characteristics of interest.
      Specifically, purposive sampling would be a resourceful way to investigate the impact of low birthweight on chronic disease later in life. For example, one major criticism of linking the fetal origin of chronic disease to low birth weight postulate is that undernutrition in utero is linked with undernutrition in childhood and is strongly correlated with behavior in adulthood(3). Purposive sampling would provide an opportunity to select cases with the characteristics of interest. An experienced investigator could select individuals who were low birth weight but who had adequate nutrition during childhood and adulthood. The nutrition of the fetus reflects the nutrition of the mother throughout her life, including her own fetal life(1). Another example of purposive sampling in studying the fetal origin of disease postulate would be investigators could select only cases with available information of the mother’s health history prior to pregnancy, such as her birth weight, nutritional habits, her body mass index, and her weight gain during pregnancy. Therefore, it would be useful to use purposive sampling to study the effects of mothers’ nutrition throughout their lives. The thesis proposed by Barker is that great benefit will come from improving the health and nutrition of girls and young women and mothers during pregnancy and lactation(1). Using purposive sampling would enable investigators to select cases with the characteristics of interest, specifically the availability of lifelong nutritional history rather than rely on crude measures of fetal nutrition such as the most commonly used measurement of birth weight. 
      Convenience sampling would be another useful nonprobability sampling technique because the investigator could select cases based on the availability of information, such as ability to locate records or accessibility to records, and the capability of obtaining information from participants. Successfully tracing the fate of births can be problematic for a few reasons. Women are often difficult to trace because they may marry and change their names making it difficult to track their records, as shown in the Hertfordshire study (1, 4). Also, often the cause of death records are unavailable(4). Due to this difficulty in obtaining all the pertinent information in fetal origin of chronic disease studies, it would be advantageous to use convenience sampling and base cases on the availability of records. 
      Furthermore, the availability of information such as placenta weight, head circumference, and mother’s nutrition during pregnancy and throughout her entire life is minimal. Therefore, using convenience sampling to access information that is readily available, or more specifically in the case of fetal origin of chronic disease available at all, would be prudent.
     Cluster sampling, or area probability sampling, would be an advantageous sampling mechanism for fetal origin studies because it would provide an efficient way to sample across a broad geographic area while retaining the characteristics of a probability sample. Much of the previous research on the fetal origin of disease has been sampled from a small homogeneous area. A follow up study of women who were born in Helsinki University Central Hospital and who went to school in Helsinki found that coronary heart disease reflects both poor prenatal nutrition and consequent small body size at birth combined with improved postnatal nutrition and consequent small body size (5). Another study in Helsinki, Finland looked at men born in the same hospital in Helsinki and found the highest death rates from coronary heart disease occurred in boys who were thin at birth but whose weight caught up(5). Similarly, a study done among 1334 men born in Uppsala, Sweden found that birth weight showed a specific, inverse association with mortality from circulatory disease(6). 
      If researchers were approaching this with a study designed to obtain fetal birth weight and length, gestational age, and any other available pertinent information from birth records, multi-stage cluster sampling would be an affordable way to obtain a sample from many different regions. For example, if the study was done in the United States, the primary sampling unit could be a random sample of states. Then, the next stage could be counties, followed by hospitals within the counties, and then the enumeration unit could be a random sample of birth records in a given time period. 
      If researchers approached studying the fetal origin of chronic disease by interviewing individuals to ask about their birth weight and current health status, then how interviewers would select samples from within households would be very important. There are many different methods of selecting interviewees if more than one individual in the household is eligible for participation,  but they all are similar methods. The interviewer would determine the number of individuals in the household who are eligible, would rank them in order of gender and age, and then would refer to a chart to determine which individual to select. 
      The advantage of using this approach to examine current health status and its association with birthweight and other markers of fetal undernutrition would be to eliminate the bias associated with record inaccessibility and the bias associated with the difficulty in tracing women due to name changes. The main disadvantage of the interview approach is recall bias, which would be a major issue since investigators would no longer rely on records. 
      Interviewing mothers would be one of the most effective ways to obtain information on maternal nutrition, exposures, and body mass index throughout her life and during pregnancy. Then investigators can follow up on the health status of her progeny and link this information with her interview. Using this approach would be advantageous for studying the effects of low birth weight and fetal undernutrition on blood pressure and other health effects that can be seen in children because locating and examining the children would be much easier than tracking grown adults. 
 There would only be a problem with sample selection within households in this approach, (interviewing mothers), if there were more than one mother in the household. In this situation interviewers would refer to some sort of table, such as a Kish table.
     As discussed in Design Effects, efficient sample size calculations assume simple random samples(7). Therefore, sample designs other than simple random sampling have an impact, called design effects, on sampling variability. For example, variability increases when cluster sampling is used rather than simple random sampling(7). Since researchers might use cluster sampling  or stratification in studying the fetal origin of disease, they must use a design effect in determining efficient sample size as a direct way of addressing the impact of the design on variability.
      To determine the design effect, researchers must estimate the true variance under the actual design and also estimate the variance that would have been expected had it been simple random sampling of the same size(8). Then, the design effect can be multiplied by the expected sampling variance in the calculation of an efficient sample size to adjust for the impact of the design.
      Randomly selecting cases eliminates selection bias and is required to generate unbiased comparisons among study groups. Therefore, it is an essential component of researching the fetal origin of disease.  Researchers can randomly select cases, such as hospital records, by using either a computer or random number table to generate a random sample of cases. 
 For example, if the investigator was using multistage cluster sampling, a number would be assigned to each cluster unit at each phase, such as assigning a number to each state and then using a computer random number generator or a random number table to select a certain number of these cases. Then the investigator would do the same for each sampling cluster within the previous cluster, such as each county, hospital, and then birth record. 
      Researchers could generate sequences of random numbers using Random Number Generators, RNGs, which are deterministic algorithms that produce numbers with certain distribution properties. Random number generators require the user to specify an initial value, or seed.
 

Links:
The Fetal Origins of Type 2 Diabetes

Fetal Origins of Adult Disease

Newsweek: Shaped by Life in the Womb

Life in the Womb: the origin of health and disease

References
1. Barker DJP. Mothers, Babies, and Health Later in Life. Edinburgh: Churchill Livingstone, 1998.
2. Lucas A. Role of Nutritional Programming in Determining Adult Morbidity. Archives of Disease in Childhood 1994;71:288-290.
3. Koupilova I, Leon DA, Vagero D. Can confounding by sociodemographic and behavioural factors explain the association between size at birth and blood pressure at age 50 in Sweden? Journal of Epidemiology & Community Health 1997;51:14-8.
4. Leon DA. Fetal growth and adult disease. European Journal of Clinical Nutrition 1998;52:S72-8; discussion S78-82.
5. Eriksson JG, Forsen T, Tuomilehto J, Winter PD, Osmond C, Barker DJ. Catch-up growth in childhood and death from coronary heart disease: longitudinal study. Bmj 1999;318:427-31.
6. Koupilova I, Leon DA, McKeigue PM, Lithell HO. Is the effect of low birth weight on cardiovascular mortality mediated through high blood pressure? Journal of Hypertension 1999;17:19-25.
7. Henry GT. Practical Sampling. Newbury Park: Sage Publications, 1990.
8. Rosander A. Case Studies in Sample Design. New York: M. Dekker, 1977.
 

Return to Sampling