The Dykes and Meier review (1975)

by Harri Hemilä


This text is based on pages 42-45 of Hemilä (2006).
This document has up to date links to documents that are available via the net.
Harri Hemilä
Department of Public Health
University of Helsinki,  Helsinki, Finland
harri.hemila@helsinki.fi
Home:  http://www.mv.helsinki.fi/home/hemila

This file:  http://www.mv.helsinki.fi/home/hemila/reviews/dykes

Version May 29, 2012


In 1975 JAMA published a review of vitamin C and the common cold by Michael Dykes and Paul Meier in the same issue as the Karlowski et al. trial (1975; see Karlowski). Meier is an eminent statistician, the second author of the widely used Kaplan-Meier survival analysis method (1958), and he was elected statistician of the year in the USA in 1985 (ASA 2007). This review in a wide circulation medical journal has thus been highly influential.

The Dykes and Meier review (1975   CH) has been extensively cited as evidence that vitamin C is worthless for colds; for example, in the US nutritional recommendations (FNB 1980 p 77, 1989a p 120), Cecil Textbook of Medicine (Kapikian 1985), and reviews and commentaries of common cold therapy (Editorial 1976; Hirsch & Swartz 1980; Anderson et al. 1983; Lowenstein & Parrino 1987; Spiers 2002). Dykes and Meier’s review (1975) was also referred to in the popular textbook Modern Epidemiology as an example of important topics that have been examined by controlled trials (Rothman 1986 p 54; Rothman & Greenland 1998 p 70).

The Dykes and Meier review (1975) contains several shortcomings (Hemilä 1996a). The authors state that "Material will only be considered that was published in the scientific literature and was, therefore, subjected to both the careful editorial peer review and the critical scrutiny of the general scientific community that are inherent in that process." It is puzzling that a statistician uses ‘publication in a journal’ as a surrogate for good quality, instead of critically reading the papers for himself. For example, Dykes and Meier (1975 p 1976) uncritically accepted the ‘placebo effect’ explanation of the Karlowski et al. trial (1975): "When the analysis is restricted to those subjects who did not know their treatment, no appreciable differences were found in any of the reported indices." However, critical reading of the Karlowski paper would have revealed to Dykes and Meier that 42% of common cold episodes were missing from the subgroup analysis without any justification, and there are many further problems with the ‘placebo effect’ explanation  (Hemilä 1996 and  Criticism). Apparently because of their blind faith in this explanation, Dykes and Meier (1975) did not present the actual results of the Karlowski et al. (1975) trial.  Compared with the placebo group, 3 g/day of vitamin C decreased the duration of colds by 6% to 9%, whereas 6 g/day decreased it by 17% (P = 0.025), suggesting dose dependency up to 6 g/day (Hemilä 1999a).

Dykes and Meier justified their review by referring to Pauling’s papers (1971a, 1971b). Furthermore, they stated that "Pauling gave great weight to the 1961 study of schoolchildren in a skiing camp in Swiss Alps by Ritzel." With such an explicit background it is odd that they did not present any of Ritzel’s data (Table 3 in Background). Dykes and Meier (1975) claimed that Ritzel’s "presentation of the data is confused and unclear in a number of respects." In fact, Ritzel (1961) listed the duration of colds in his text, and in table form showed the total days of colds per group, and the number of cases and total days of ‘constitutional symptoms’ per group, which are extracted in Table 3 of Background. Thus, Dykes and Meier did not mention that there was a 31% decrease in the mean duration of episodes, a 45% decrease in the incidence of colds, and a 61% decrease in the total number of days of colds in the group administered 1 g/day of vitamin C. Nor did they mention that there was a 62% decrease in the number of children with ‘constitutional symptoms,’ and an 81% decrease in the total number of days with constitutional symptoms in vitamin C group. Dykes and Meier merely commented that the difference in common cold incidence in the two groups was only marginally significant (P[2-t] = 0.04), which seems to be intentional camouflaging of the actual results. It is quite possible that some readers of JAMA might have considered that a 61% reduction in total number of sickness days, or an 81% reduction in days with constitutional symptoms might be worth serious considerations when the results arise from a randomized double-blind placebo-controlled trial with children, even though a one-week trial in a skiing camp does not allow any direct extrapolation to ordinary schoolchildren. Thus, concealing the actual results prevented a critically minded reader from drawing his or her own conclusions.

Dykes and Meier (1975) presented the results of the Anderson et al. trial (1972), but did not calculate any estimates of effect, such as percentage difference in favor of the vitamin C group, which hampers the reader in considering the practical significance of the findings. Dykes and Meier wrote of the Anderson 1972 trial: "The estimated effect is considerably less than that predicted by Pauling for the dose level." Anderson had reported a 30% decrease (P = 0.0005) in the ‘total number of days confined to house’ per participant. Dykes and Meier’s comment seems irrelevant as many readers might consider that with an inexpensive nutrient which costs pennies per gram and is safe in large doses, even such moderate benefits might be worthy of exploitation irrespective of how they compare with Pauling’s predictions. Moreover, on biological grounds one would expect the benefit of supplementation to be greater for subjects with a low dietary vitamin C intake, and in this respect Anderson’s subgroup finding that supplementation was substantially more beneficial for those participants that had low intake of fruit juices is highly interesting (Hemilä 2006 Table 13, p 35). Thus Dykes and Meier were not consistent in their review when they uncritically accepted Karlowski’s subgroup analysis which was based on the exclusion of 42% of cold episodes without any rationalization, and rejecting Anderson’s subgroup analysis, which is biologically well rationalized.

Dykes and Meier comment on the Anderson et al. trial (1974) that "A question is raised whether one of two control groups may have been biased, but the data given are not sufficient to evaluate this point." This is quite a puzzling comment by a statistician, since the discrepancy between the 2 placebo groups in the Anderson 1974 trial can be seen in several simple calculations (Hemilä 2006 Table 16, p 40), and accordingly the data are sufficient to evaluate this point, in contrast to Dykes and Meier’s statement.

Dykes and Meier (1975) explicitly commented on the Coulehan et al. trial (1974) with Navajo schoolchildren that "Because the data required for an appropriate analysis are not presented, the statistical significance of the differences reported cannot be considered to have been established." In fact, Coulehan et al. (1974) reported in their table 4 that 44% (143/321) of school children administered vitamin C were ‘never ill on active surveillance,’ while only 29% (93/320) of those administered placebo were ‘never ill’ (Table 20). Since it is highly unlikely that such a difference between the vitamin C and placebo groups would arise purely by chance (P = 0.00003; Table 20), in contrast to Dykes and Meier’s claim in their review, important elements of Coulehan’s results were presented and can be re-analyzed by a statistically oriented reader. When such highly significant differences between vitamin C and placebo groups are reported, it is grossly misleading to state in a review that significant differences "cannot be considered to have been established."

Furthermore, Dykes and Meier are not consistent in their text. They accuse Coulehan et al. (1974) of failure to present relevant data (partly true), whereas they themselves fail to present the data Coulehan did report (e.g., Table 20) in their own review. Coulehan et al. (1974) also found that the duration of colds was 12% and 29% shorter in children administered 1 and 2 g/day of vitamin C respectively, suggesting dose dependency up to 2 g/day, but this data was also missing from the Dykes and Meier review. Finally, Dykes and Meier failed to present any results of the Ritzel (1961) and Karlowski et al. (1975) trials, thereby preventing readers from drawing any independent conclusions about them.

Pauling (1976b) considered that "some significant studies in this field were not mentioned by Dykes and Meier, and some important aspects of the studies discussed by them were also not mentioned by them." Pauling thus wrote a manuscript in which he presented the results of the relevant studies, his own interpretations, and some criticism of Dykes and Meier’s argument. For example, commenting on the Ritzel (1961) trial, Dykes and Meier (1975 p 1075) stated that "Pauling infers the number of subjects [with colds] by dividing ‘illness days’ by ‘mean illness days’ and concludes that there is a significant difference in proportions of subjects experiencing colds. If his interpretation is correct, the difference is indeed significant." To which Pauling (1976b) replied that "It is hard for me to understand why Dykes and Meier should suggest that my interpretation might be incorrect. It involves a very simple calculation. Ritzel states (in his table 1 [see Table 3 in Background]), that the total number of days of illness for the ascorbic-acid subjects was 31. He also states (page 66) that the average number of days per episode of illness was 1.8. The ratio 31/1.8 is 17.2; that is, there were 17 episodes of illness in this group. A similar calculation gives 31 colds for the placebo subjects (80 total days of illness, 2.6 average number of days per episode). It is safe to assume that no subjects had two colds in the same week. With this assumption, the null hypothesis of equal probability of colds for the two groups is rejected at the level P (one-tailed) < 0.015. Dykes and Meier mention that I give great weight to the Ritzel study. I do give great weight to it, and I find it strange that they should reject it on the basis of trivial complaints, such as their apparent failure to understand the simple calculation described above" (Pauling 1976b).

Pauling submitted his manuscript to JAMA, but his paper was rejected even after Pauling twice made revisions to meet the suggestions of the referees, and the manuscript was finally published in a minor journal (1976b, 1976c; see also: 1976a pp 133-5; 1986a pp 231-2). The rejection of Pauling’s review was quite a strange policy by JAMA, since the readers were thereby prevented from seeing the other side of an important scientific controversy.


Table 20. Prophylactic benefit of vitamin C in large trials

Trial Free of illness during the trial / Persons in group Proportion of participants
getting benefit (95% CI)
P (1-tail) 1

Vitamin C Placebo

Coulehan-1974



641 Navajo schoolchildren
Duration 3 months
Alternative allocation, double-blind




All children 2 143/321 (44%) 93/320 (29%) 16% (8%, 23%) 0.000,03
Low grades (1 g/d vitamin C)



Boys 31/81 (38%) 18/87 (21%) 18% /4%, 31%) 0.007
Girls 30/109 (28%) 12/105 (11%) 16% (5%, 26%) 0.002
High grades (2 g/d vitamin C)



Boys 40/69 (64%) 34/61 (56%) 9% (-8%, 26%) 0.2
Girls 42/69 (61%) 29/67 (43%) 18% (1%, 34%) 0.021





Anderson-1972



818 adults in Canada with colds usually in winter
Duration 3 months
Randomized, double-blind, 1 g/d vitamin C




Free of any illness during the trial 105/407 (26%) 76/411 (18%) 7.3% (1.6%, 13%) 0.006
No 'days confined to house' 232/407 (57%) 195/411 (47%) 9.5% (2.7%, 17%) 0.003
No 'days off work' 275/407 (67%) 243/411 (59%) 8.4% (1.8%, 15%) 0.006

1 One-tailed mid-P-value. The CI was calculated using the CIA (confidence interval analysis, BMJ) program.
2 In the Coulehan trial, the number of children ‘free of illness’ in the placebo group does not match in the original paper: 18 + 12 + 34 + 29 = 93,
and not the 92 published by Coulehan; in this table N = 93 is used as the sum of children ‘free of illness’ in the placebo group since it matches
with the subgroup data.
The outcome in Coulehan’s table 4 is erroneously called as ‘No. of days without sickness observed per total in group’ whereas the actual data is
‘Number of children without sickness observed per total in group’ (Coulehan 1995 personal communication).

References


Anderson LJ, Patriarca PA, Hierholzer JC, Noble GR (1983) Viral respiratory illnesses. Med Clin North Am 67:1009-30

Anderson TW, Reid DBW, Beaton GH (1972) Vitamin C and the common cold: a double-blind trial. Can Med Assoc J 107:503-8

Anderson TW, Suranyi G, Beaton GH (1974) The effect on winter illness of large doses of vitamin C. Can Med Assoc J 111:31-6

ASA [American Statistical Association] (2007) Statistician of the Year Award Winners

Coulehan JL, Reisinger KS, Rogers KD, et al. (1974) Vitamin C prophylaxis in a boarding school. N Engl J Med 290:6-10

Dykes MHM, Meier P (1975) Ascorbic acid and the common cold: evaluation of its efficacy and toxicity. JAMA 231:1073-9    PubMed  
 CH   BM     *  comments in: JAMA (1976);235:1108  JAMA (1976);235:1108; Pauling (1976b, 1976c

Editorial (1976) Vitamin C and the common cold. BMJ 1:606-7

FNB [Food and Nutrition Board, National Research Council] (1980) Recommended Dietary Allowances, 9th edn. Washington DC: National Academy of Sciences *  Vitamin C pp 72-82

FNB [Food and Nutrition Board, National Research Council] (1989a) Recommended Dietary Allowances, 10th edn. Washington DC: National Academy Press *   Vitamin C pp 115-25  p 120

Hemilä H (1996) Vitamin C, the placebo effect, and the common cold: a case study of how preconceptions influence the analysis of results. J Clin Epidemiol 49:1079-84   Manu with links to Refs    CH *  Discussion in  J Clin Epidemiol 1996;49:1085 and  Hemilä (1996c)

Hemilä H (1996c) To the dissent by Thomas Chalmers. J Clin Epidemiol 49:1087      Manu with links to Refs   CH

Hemilä H (1996a) Vitamin C supplementation and common cold symptoms: problems with inaccurate reviews. Nutrition 12:804-9  Manu with links to Refs  CH

Hemilä H (1999a) Vitamin C supplementation and common cold symptoms: factors affecting the magnitude of the benefit. Med Hypotheses 52:171-8   CH

Hemilä H (2006) Do vitamins C and E affect respiratory infections? [Dissertation]. University of Helsinki, Finland   Hemilä 2006

Hirsch MS, Swartz MN (1980) Antiviral agents. Part II. N Engl J Med 302:949-53

Kapikian AZ (1985) The common cold. In: Cecil Textbook of Medicine, 17th edn [Wyngaarden JB, Smith LH, eds]. Philadelphia, PA: Saunders. pp 1691-5

Kaplan EL, Meier P (1958) Nonparametric estimation from incomplete observations. J Am Statist Assoc 53:457-81  Available   *    see also: garfield.library/classics   wiki Kaplan-Meier

Karlowski TR, Chalmers TC, Frenkel LD, Kapikian AZ, Lewis TL, Lynch JM (1975) Ascorbic acid for the common cold: a prophylactic and therapeutic trial. JAMA 231:1038-42  * see Problems

Lowenstein SR, Parrino TA (1987) Management of the common cold. Adv Intern Med 32:207-34


Pauling L (1971a) The significance of the evidence about ascorbic acid and the common cold. Proc Natl Acad Sci USA 68:2678-81  PMC 

Pauling L (1971b) Ascorbic acid and the common cold. Am J Clin Nutr 24:1294-9    NLM
Pauling L (1976a) Vitamin C, the Common Cold, and the Flu. San Francisco: Freeman

Pauling L (1976b) Ascorbic acid and the common cold: evaluation of its efficacy and toxicity. Part I. Medical Tribune 17(12):18-9

Pauling L (1976c) Ascorbic acid and the common cold. Part II. Medical Tribune 17(13):37-8

Pauling L (1986a) How to Live Longer and Feel Better. NY: Freeman

Ritzel G (1961) Kritische Beurteilung des Vitamins C als Prophylacticum und Therapeuticum der Erkältungskrankheiten [in German; Critical analysis of the role of vitamin C in the treatment of the common cold]. Helv Med Acta 28:63-8 TRANSLATION    Ritzel 1961 in German ch     Pauling note 30    Pauling note 33   Ritzel 1976 JAMA   Ritzel 1976 JAMA

Rothman KJ (1986) Modern Epidemiology. Lippincott Boston, MA: Williams Wilkins

Rothman KJ, Greenland S (1998) Modern Epidemiology, 2nd edn. Boston, MA: Lippincott Williams Wilkins, p 70

Spiers PS (2002) On the prevention of the common cold: no help from vitamin C. Epidemiology 13:4-5


Copyright:
© 2006-2009 Harri Hemilä. This text is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.  

Creative Commons License
The Dykes and Meier review (1975) by Harri Hemilä is licensed under a Creative Commons Attribution 1.0 Finland License.
Based on a work at www.mv.helsinki.fi/home/hemila/reviews/dykes.



Valid HTML 4.01 Transitional