(www.improvingmedicalstatistics.com)                                          
      home   author    contact   web site dedication
  

back

The initial study was published in JAMA (Mills)1 in 1984 titled “Maternal Alcohol Consumption and Birth Weight – How Much Drinking During Pregnancy Is Safe”. The data that the authors presented seemed to suggest a different conclusion than what the authors proposed. 

The authors suggested that low level alcohol ingestion resulted in low birth weight infants. A  letter2 was written and published in JAMA  suggesting an alternative interpretation of the data noting that there appeared to be no significant difference in the average birth weight for nonsmokers for low levels of alcohol intake.

                                            JAMA Study Data:

Alcohol intake,             Mean Birth Weight (nonsmokers)   Smokers ( 1/2 pack/day)
        0     (drinks/day)              3,468±534(S.D.)                     3,330±548               
      <1                                     3,500±527                              3,386±522
       1-2                                   3,452±582                              3,253±564
      >3                                     3,138±512                              3,085±626            


 The original authors divided their data, at least for part of the analysis, between smokers and nonsmokers. Their raw data suggested that nonsmokers did not have a significant decrease in the mean birth weight of the infants for either less than one drink per day or one to two drinks per day.  This was not a small subgroup of the women studied.  The majority of the women studied were nonsmokers (22,485 pregnancies in nonsmokers and 8,619 pregnancies in smokers) so there were more than adequate numbers of nonsmokers to analyze.

The raw data for mean birth weight in nonsmokers shows no decrease in birth weight for alcohol ingestion of less than one drink for day. 

(In fact, there was a insignificant increase in mean birth weight of infants of very light drinkers of 32g, (1 oz.=28g.)).  For alcohol ingestion  of 1-2 drinks per day in nonsmokers, there was an insignificant decrease in mean birth weight of 14grams  (1 oz. = 28g.).  Hence, for alcohol intake of 2 or less drinks per day in nonsmokers there was no significant change  demonstrated.

Smokers had lower birth weight infants.  

The authors when questioned about this lack of effect, answered in JAMA that after removing the effect of smoking by multiple regression analysis (letter)3, they were able to show that this effect existed even in nonsmokers.  This may overestimate the power of multiple regression analysis to reliably balance out confounding factors. 

The crude data suggests there is no meaningful difference between the crude infant birth weight and the crude mean infant birth weight of nonsmokers who drank zero drinks per day versus less than one drink per day or those that drank one to two drinks per day.  For three or more drinks, the birth weight trended downward in the nonsmokers.

If the raw data from this large study still exists it would be quite interesting if this data was given to several different groups of physicians and statisticians to reanalyze the relation of low level alcohol ingestion in nonsmokers to birth weight and pregnancy outcomes.

1. Mills JL, Graubard BI, Harley EE, et al.  Maternal alcohol consumption and birth weight: How much drinking during pregnancy is safe? JAMA 1984:252:1875-1879

2. Roehm E.  Letter, Maternal alcohol consumption and birth weight. JAMA 1985:253:3551

3.  Mills JL, et al.  Letter, Maternal alcohol consumption and birth weight. JAMA 1985:253:3351

back