Reader Comments

Post a new comment on this article

Significant Problems in Study Design and Analyses

Posted by plosmedicine on 31 Mar 2009 at 00:13 GMT

Author: Virginia J. Vitzthum, PhD
Position: Bioanthropologist
Institution: Indiana University, Bloomington,and Max Planck Institute for Evolutionary Anthropology, Leipzig.
Submitted Date: September 06, 2007
Published Date: September 10, 2007
This comment was originally posted as a “Reader Response” on the publication date indicated above. All Reader Responses are now available as comments.

Serious limitations in methodology undermine this study's conclusions. Also, important descriptive statistics were omitted and previously published work was neglected.

Cross-sectional studies of progesterone (P), such as[1], suffer from selection bias[2-4]. Due to individual variation in fecundity[5], the most fecund women are more likely to be pregnant or lactating (participant exclusion criteria in[1]) leaving the least fecund (likely those with lower P) available for inclusion in the study. Thus, the population's true P level is systematically underestimated by an unknown amount. This bias is larger in samples from populations with high marriage rates and low contraceptive use (eg, SYL and ADU in[1]).

Fig 1 in[1] suggests that P has been underestimated. Peak P in WHI is ~75 pg/ml, only ~60% as high as peak P in similar samples assayed by the same method in the same lab[4,6,7]. This low estimate could arise from selection bias and/or the inclusion of anovulatory cycles[2,4]. Rather than using tested methods to identify ovulatory cycles[8], [1] "include[d] only the cycles for which an oestradiol midcycle peak and luteal progesterone rise were discernable." But the estradiol peak must be present AND of adequate height and duration to prompt the LH surge that triggers ovulation[3,9]. Nor can ovulation be inferred from a "discernable" P rise of unspecified height or duration.

The inclusion of anovulatory cycles biases the estimate of P downwards (more so in samples with higher anovulation rates) and obscures other sources of P variation. Lower acculturation among migrants can incur psychosocial and economic stress, factors associated with anovulation[4,10]. The very low P levels in SYL and ADU (luteal P barely rises above follicular P) resemble the characteristic profile of anovulatory cycles[4]. Also, although described as "well-nourished" women (anthropometrics were not given), P in SYL and ADU are markedly lower than P in ovulatory cycles of very poor Bolivians (assayed in the same lab[4]), again suggesting that SYL and ADU include anovulatory cycles.

Psychosocial stress is also associated with lower P in ovulatory cycles[10]. Yet, despite their statement that stress varies among the samples, [1] did not include this confounder in their analyses.

They also neglected the correlation between adult body size (an indicator of childhood environments[11]) and P levels in Bolivians[4]. Based on life history theory, I developed the Flexible Response Model (FRM) and proposed that chronically lower P reflects childhood conditions but need NOT be associated with impaired fecundity[12-14]. Lower P in poorer than in better-off Bolivians[4] and a longitudinal study[14] of P at conception in Bolivians supported the FRM. [1] dismissed this finding by noting that the Bolivians, unlike the US women, made no special effort to conceive. But this is irrelevant. Indeed, given Sylhet's high fertility[15], the FRM is also supported if P in SYL is low after accounting for bias and confounders.

In sum, the P estimates in [1] are likely biased to differing degrees, and can neither confirm nor refute a predicted association of childhood environments with adult P levels.


1.Núñez-de la Mora et al (2007) PLoS Med 4(5):e167.
2.ref [7,9,63] in [1].
3.Wood (1994) Dynamics of Human Reproduction. NY:Aldine.
4.Vitzthum, Bentley et al (2002) Hum Repro 17:1906.
5.Dunson et al (2002) Hum Repro 17:1399.
6.ref [49,64] in [1].
7.Gann et al (2001) Cancer Epi Biomark Prev 10:59.
8.Campbell and Rockett (2006) Paediatr Perinat Epidem 20 Sup 1:13.
9.Messinis (2006) Hum Repro Update 12:557.
10.Nepomnaschy et al (2004) Am J Hum Biol 16:523.
11.Komlos and Baten (1998) The Biological Standard of Living. Stuttgart:Verlag.
12.Vitzthum (1997) in Morbeck et al, The Evolving Female. Princeton U Press.
13.Vitzthum (2001) in Ellison, Reproductive Ecology and Human Evolution. NY:Aldine.
14.Vitzthum et al (2004) PNAS 101:1443.
15.Islam et al (2003) J Biosoc Sci 36:351

Competing interests declared: Drs. Bentley, Chatterton and I have long standing disputes regarding data analysis and interpretation, and other matters. I do not believe that these differences of opinion have influenced my scientific judgment of their article.