More details on the shifting goalposts for defining vitamin D deficiency and vitamin D insufficiency:

We searched in PubMed for studies on Vitamin D deficiency in humans published before 2000. Out of 100 papers we reviewed, 20 papers gave specific thresholds to define vitamin D deficiency or insufficiency (also called mild, marginal, subclinical, or borderline deficiency). These are summarized in the table below. They all date from the 1980s or 1990s. Note that 10 ng/ml was the most common threshold used to define vitamin D deficiency, whereas 12 ng/ml was the most common threshold used to define vitamin D insufficiency. The highest threshold used was 16 ng/ml.

 

PMID               

Citation

Words used to describe low vitamin D        

Threshold in ng/mL

1556361

J Am Geriatr Soc. 1992 Apr;40(4):343-7.        

definitely deficient

5

9718191

J Bone Miner Res. 1998 Aug;13(8):1238-42. 

severe deficiency

8

1883119

Ann Intern Med. 1991 Oct 1;115(7):505-12. 

below the normal range

8

7148741

Am J Clin Nutr. 1982 Dec;36(6):1225-33. 

borderline deficiency

8

7602040

J Am Geriatr Soc. 1995 Jul;43(7):822-8. 

deficiency

10

2615720

Miner Electrolyte Metab. 1989;15(6):332-7.

below the normal range

10

9418761

J Am Acad Dermatol. 1997 Dec;37(6):942-7. 

below the normal range

10

7669375

J Paediatr Child Health. 1995 Jun;31(3):180-4.         

deficiency

10

2349922

Am J Clin Nutr. 1990 Jun;51(6):1075-81. 

deficiency

10

7474272

JAMA. 1995 Dec 6;274(21):1683-6. 

deficiency

10

8574630

Rev Rhum Engl Ed. 1995 Oct;62(9):576-81.

deficiency

10

9718191

J Bone Miner Res. 1998 Aug;13(8):1238-42. 

deficiency

12

8585420

J Bone Miner Res. 1995 Aug;10(8):1177-84. 

deficiency

12

9425501

Osteoporos Int. 1997;7(5):439-43. 

insufficiency

12

3417845

J Clin Endocrinol Metab. 1988 Oct;67(4):644-50.     

borderline deficiency

12

9425501

Osteoporos Int. 1997;7(5):439-43. 

insufficiency

12

9539276

Acta Obstet Gynecol Scand. 1998 Mar;77(3):303-6.           

marginal vitamin D 

12

10757956

Obes Surg. 1993 Nov;3(4):421-424. 

deficiency

16

1997517

J Clin Endocrinol Metab. 1991 Mar;72(3):628-34. 

subclinical deficiency

16

1556361

J Am Geriatr Soc. 1992 Apr;40(4):343-7. 

mild deficiency

16

 

These thresholds were gradually raised in the literature between 1998 and 2007, largely driven by three key papers, all authored or co-authored by Michael Holick:

  • 1998: In a Lancet paper called Redefining vitamin D insufficiency, the authors proposed raising the threshold for defining vitamin D insufficiency to 20 ng/ml. Holick is the senior author.

  • 2005: In a paper in Osteoporosis International called Estimates of optimal vitamin D status, six co-authors (including Holick) each recommended thresholds for defining vitamin D insufficiency, ranging from 20 ng/ml to 32 ng/ml. 

  • 2007: In a single-author paper in the New England Journal of Medicine called Vitamin D deficiency, Holick declared that vitamin D deficiency is 20 ng/ml and insufficiency is 30 ng/ml.

 

Fact-checking/Statistical Sleuthing of Michael Holick’s work

Statistical cockroach 1

From Holick’s book The UV Advantage, first paragraph of chapter 1, “The Facts of Light”:

“Here it was the summer of 1997 and for months my staff and I had been studying the vitamin D status of a random group of people living in the Boston area. As Director of the General Clinical Research Center and Professor of Medicine, Dermatology, Physiology, and Biophysics at Boston University Medical Center, I had designed and was leading the study. I was sitting in my office when the results came in. Although I had a strong suspicion by that time of what the study would reveal, the actual numbers were staggering. Fully 42 percent of the people we studied were vitamin D deficient. My study, which was accepted for publication in Lancet (only one in 100 papers submitted to this renowned journal are published), confirmed what most scientists in the field believe. That is, there is an epidemic of vitamin D deficiency in the United States and much of the Western world.”

This appears to refer to Holick’s 1998 Lancet paper Redefining vitamin D insufficiency. The paper describes a study in which the authors gave vitamin D supplements to 35 patients from their bone clinic. The study is not about the prevalence of low vitamin D in a random sample of patients from Boston. The “42 percent” statistic appears to come from one sentence near the end of the paper that describes a chart review done on 169 patients from their bone health clinic: “After reviewing 169 patients seen in our bone health clinic (86% women, 14% men; average age is 62 [16] years), 41% were found at risk for vitamin D deficiency with 25(OH)D concentrations less than 20 ng/mL.” Obviously, this does not represent a random sample of people in Boston, and thus the book passage is misleading.

Note that in 1998, <20 ng/mL was not considered “vitamin D deficiency.” In fact, the paper advocates raising the threshold for vitamin D insufficiency, not deficiency, to 20 ng/ml. Thus, Holick hedges by saying that the patients are only “at risk” for vitamin D deficiency.

 

Ball of statistical cockroaches 2

From Holick’s 2011 book The Vitamin D Solution: A 3-Step Strategy to Cure Our Most Common Health Problems: “In fact, in one of my studies we looked at forty newborn babies whose mothers were seemingly doing everything right before giving birth. Seventy percent of them took prenatal vitamins, 90 percent drank fortified milk, and all ate fish—one of the best dietary sources for vitamin D—regularly during their pregnancy. Upon giving birth, 76 percent of the moms and a full 81 percent of the newborns were vitamin D deficient.” 

The book passage contains mathematical inconsistencies as well as discrepancies from the actual study, which was published in Clinical Pediatrics 2007, Vitamin D deficiency in a healthy group of mothers and newborn infants

  • It is not possible to have 76% of 40 moms or 81% of 40 infants. 75% and 80% of 40 are possible, but not 76% and 81%. 
  • The published study reports that 50% of 40 moms and 65% of 40 infants were vitamin D deficient, not 76% or 81%. This discrepancy likely arises because the study defined vitamin D deficient as <12 ng/ml. In the book passage, Holick must be using a threshold of <20 ng/ml to define deficiency. 
  • Table 1 of the study reports that the number and percent of moms who ate fish was 44 and 89.7%. The 44 is obviously impossible as there were only 40 moms in the study. Also, 44/40 is not 89.7%. Finally, the book passage says that “all” moms ate fish, which is discrepant with the 89.7%. 
  • The book passage also omits key context: the sample was mostly Black and Hispanic mothers measured in late fall and winter, all factors which would make it more likely to find low vitamin D.

Some of these errors appear to have arisen because at one point the study involved 50, not 40, mother-infant pairs. We know this because Holick references the study as unpublished data in a 2003 paper called Vitamin D deficiency: what a pain it is in Mayo Clinic Proceedings: “In Boston, 76% of 50 mother-infant pairs were found to be vitamin D-deficient, as were 69% of infants in the New York area (J. M. Lee, MD, B. L. Phillip, MD, D. S. Hirsch, MD, M. F. Holick, MD, unpublished data, 2003).” This would explain where he got the mathematically impossible 76% cited in The Vitamin D Solution (38/50=76%) and also why 44 mothers were reported to eat fish in Table 1 of the published study (it is possible to have 44 out of 50, and 89.7% is possible if one mom was missing data on fish consumption, since 44/49=89.7%). 

But the 2007 study in Clinical Pediatrics never mentions that the data for 10 out of 50 infant-mother pairs were dropped, so we don’t know why they were dropped. The lack of transparency raises suspicion of cherry-picking data. 

 

Statistical cockroach 3

In a 2006 paper, Holick claims that up to 100% of healthy adults in Europe are vitamin D deficient (Holick MF. High prevalence of vitamin D inadequacy and implications for health. Mayo Clin Proc. 2006;81:353-73.): “Typically, the prevalence of low 25(OH)D levels (<20 ng/mL [50 nmol/L]) is approximately 36% in otherwise healthy young adults aged 18 to 29 years…and even higher in Europe (28%-100% of healthy and 70%-100% of hospitalized adults).” 

To justify the 100% statistic, Holick cites a 2003 study in The Journal of Clinical Endocrinology & Metabolism called Low Vitamin D Status, High Bone Turnover, and Bone Fractures in CentenariansThis is a study of 104 Italian centenarians (aged 98 to 105 years old). All but one have undetectable or low vitamin D. But this is obviously not a group that is representative of healthy adults in Europe. 

Though the paper states that None of the centenarians had any acute disease at the time of the visit,” it is misleading to call them “healthy” as the paper also states that: “More than half presented levels of dementia” and “Fifty-four centenarians were unable to walk.”

 

Ball of statistical cockroaches 4

Holick published a paper in the American Journal of Medicine in 2002: Tangpricha V, Pearce EN, Chen TC, Holick MF. Vitamin D insufficiency among free-living healthy young adults. Am J Med. 2002 Jun;112:659-62. This study measured vitamin D levels in 311 medical students, residents, and doctors at Boston University Medical Center who attended a vitamin D awareness screening. The percentages with low vitamin D, defined as <20 ng/ml, were as follows (from text and Figure 3):

  • 21% of the total sample (65/311) 
  • 8% of those taking daily multivitamins (10/123)
  • 32% of 18-to 29-year-olds measured at the end of winter (22/69)
  • 4% of 18-to 29-year-olds measured at the end of summer (3/69)

He cites this study in at least four subsequent papers, but his citations are inaccurate and misleading. Note that he is the sole author on the first three papers and the first author on the final paper below, so we cannot attribute these mis-citations to errors introduced by coauthors:

“32% of healthy students, physicians, and residents were found to be vitamin D-deficient, despite drinking a glass of milk and taking a multivitamin daily and eating salmon at least once a week.” (Holick MF. Vitamin D deficiency. N Engl J Med. 2007;357:266-81.)

This is misleading because the 32% is only in 18-to 29-year-olds measured at the end of winter (n=69), not the whole sample of students and physicians (n=311), for which the percentage was 21%. Also, not everyone was drinking milk, eating salmon, and taking a daily multivitamin daily. Only 40% of the sample was taking a multivitamin daily and, among this group, only 8% had vitamin D <20 ng/ml. 

“Typically, the prevalence of low 25(OH)D levels (<20 ng/mL [50 nmol/L]) is approximately 36% in otherwise healthy young adults aged 18 to 29 years.” (Holick MF. High prevalence of vitamin D inadequacy and implications for health. Mayo Clin Proc. 2006;81:353-73.)

Here, Holick acknowledges the age group but does not clarify that they were measured at the end of winter. Among 18-to 29-year-olds measured at the end of summer, only 4% of this age group had vitamin D <20 ng/ml. The 36% also appears to be an error, as Figure 3 of the Am J Med 2002 paper shows a prevalence of 32% in this group at the end of winter. 

“Tangpricha et al reported that 32% of healthy young white men and women in Boston aged 18 to 29 years were vitamin D-deficient at the end of winter in 2003.” (Holick MF. Vitamin D deficiency: what a pain it is. Mayo Clin Proc. 2003;78:1457-9.)

Here he correctly notes that the 32% is only in 18-to 29-year-olds in Boston at the end of winter, but he gets other factual details wrong. The study took place in 1999, not 2003; and only 60% of the total sample was White. 

“32% of healthy students and physicians at a Boston hospital had 25(OH)D below 20 ng/ml. (Holick MF, Binkley NC, Bischoff-Ferrari HA, et al. Evaluation, Treatment, and Prevention of Vitamin D Deficiency: an Endocrine Society Clinical Practice GuidelineJ Clin Endocrinol Metab. 2011;96:1911-30.)

This is misleading because the 32% is only in 18-to 29-year-olds measured at the end of winter (n=69), not the whole sample of students and physicians (n=311), for which the percentage was 21%.

 

Statistical Cockroach 5

Holick’s 2007 paper in the New England Journal of Medicine, Vitamin D deficiency, was instrumental in establishing vitamin D deficiency as <20 ng/ml. To justify this cutoff, Holick writes: “Although there is no consensus on optimal levels of 25-hydroxyvitamin D as measured in serum, vitamin D deficiency is defined by most experts as a 25-hydroxyvitamin D level of less than 20 ng per milliliter (50 nmol per liter).” 

This statement is highly misleading. At the time, <20 ng/ml was not widely used as the cutoff for defining vitamin D deficiency. Holick cites four references to back up his statement, none of which suggests that “most experts” agree on this definition. None of the references define vitamin D deficiency as <20 ng/ml, and 2 of 4 references are to his own papers (self-referential):

  1. Reference 7: This is a reference to one of Holick’s own single-author review papers. The review paper summarizes data on the prevalence of low vitamin D using definitions of low vitamin D anywhere from <5 ng/ml to <30 ng/ml. It does not define a specific threshold for vitamin D deficiency. (Holick MF. High prevalence of vitamin D inadequacy and implications for health. Mayo Clin Proc. 2006;81:353-73.)
  2. Reference 8: This is a review paper that argues that >30 ng/ml is “optimal” for bone health but doesn’t define or discuss definitions for vitamin D deficiency. (Bischoff-Ferrari HA, Giovannucci E, Willett WC, et al. Estimation of optimal serum concentrations of 25-hydroxyvitamin D for multiple health outcomes. Am J Clin Nutr. 2006;84:18-28.)
  3. Reference 9: This is a reference to Holick’s 1998 Lancet paper. He and coauthors suggest that vitamin D insufficiency, not deficiency, should be “redefined” as <20 ng/ml. (Malabanan A, Veronikis IE, Holick MF. Redefining Vitamin D Insufficiency. Lancet. 1998;351:805-6.)
  4. Reference 10: This paper defines vitamin D deficiency as <15 ng/mL not <20 ng/ml. (Thomas KK, Lloyd-Jones DM, Thadhani RI, et al. Hypovitaminosis D in medical inpatients. N Engl J Med 1998;338:777-78.)