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

back

Table A and Table B  show data from the same patients from the same paper. One presentation of the data shows 446 low birth weight infants and another shows 565 low birth weight infants for the same births which indicates the data has been miscounted.

Table A        Number of Low Birth Weight Infants for Nonsmokers  (5,400 Nonsmokers)1

                                                                                            <2,500 grams                          >2,500 grams

No Alcohol                                                            342 (11.3%)                                 2,690 (88.7%)
Light Alcohol (1-3 Drinks Per Mo. )                       70 (11.0%)                                   566 (89%)
Moderate Alcohol (1-13 Drinks Per Week)            30 (13.0%)                                   201 (87%)
Heavy Alcohol (>14 Drinks Per Week)                    4 (30.8%)                                       9 (69.27%)   
                                            Total:                      446 low birth weight infants

 

Data for the same 5,400 nonsmokers with a slightly different alcohol stratification from the same paper.  Note that there is a different number of low birth weight infants which is clearly an error. 
 

Table B   Number of Low Birth Weight Infants for Nonsmokers (same 5,400 Nonsmokers)1                                     

                                                                     <2,500 grams                        >2,500 grams

No Alcohol                                                          342                                        2,690
1-10 Drinks Per Week                                           86                                           384      
11-20 Drinks Per Week                                       106                                           340      
>21 Drinks Per Week                                            31                                            75         
                                                             Total       565     low birth weight infants  

1data from:  Virji. Relationship between alcohol consumption during pregnancy and infant birth weight. Acta Obstet Gynecol Scand 1991; 70: 303-308

 

The author1 made the serious error of miscounting data points regarding the primary outcome, i.e. low birth weight infants, which the trial was evaluating.

  One presentation of the data says there are 446 low birth weight infants in nonsmokers and the other table for the same group of infants shows 565 low birth weight infants for nonsmokers. Obviously, both can not be correct, (There were two tables showing the data. The first table showed the patients with the alcohol intake divided among light, moderate, and heavy. The next table was divided by alcohol intake of no alcohol, 1 to 10 drinks, 11 to 20 drinks, and greater than 20 drinks per week. The data for the nonsmokers is for 5,400 patients. However, the two different tables get completely different numbers for the number of low birth weight infants occurring in the nonsmoking group.  The upper table shows that there are 446 nonsmokers with low birth weight infants. The same patients presented in a slightly different fashion in the second table indicate that there are 565 low birth weight infants in the nonsmokers who had low birth weight. 

These numbers should be identical which means that the author miscounted the data points involved. Since the data is corrupt and not consistent, it cannot be determined with certainty which table, the first or second, is most likely to accurately reflect the true results.  However, there were 5,400 nonsmoking patients being analyzed. Adding every nonsmoker in the first table gives a total of 5,401 patients.  In the second table, adding every nonsmoking patient in this table gives 5,430 patients, which is clearly erroneous.  This study did not reliably count the data points of its primary outcome.

1. Virji. Relationship between alcohol consumption during pregnancy and infant birth weight. Acta Obstet Gynecol Scand 1991; 70: 303-308

back