Tian T, Shao J, Shen Z, et al. Association of serum vitamin C with all-cause and cause-specific death: data from National Health and Nutrition Examination Survey (NHANES 2003-2006). Nutrition. 2022;101:111696.
To explore the nonlinear association between serum vitamin C and all-cause or cause-specific mortality
Vitamin C may not be as benign as generally assumed.
Observational study of the National Health and Nutrition Examination Survey (NHANES 2003–2006)
In this study, researchers analyzed data from participants in NHANES 2003–2006, specifically the 2 NHANES cycles (2003–2004 and 2005–2006) in which serum vitamin C concentration levels were tested.
Of the initial 20,470 participants, about half were excluded, mostly because of missing vitamin C data. In the end, 9,902 participants were included in the analysis.
The mean age of those included for analysis was 45.6 years; 51.6% were female. Self-identifying race/ethnicity was as follows: 72% Non-Hispanic White, 11.2% Black, 8% Mexican-American, 3.5% Other Hispanic, and 5.3% Other Race.
Serum vitamin C level (mg/dL) was the primary exposure variable of interest.
Researchers detected it using isocratic high-performance liquid chromatography at various laboratories.
Study Parameters Assessed
Outcome variables included all-cause and cause-specific death. Researchers obtained these data by matching the NHANES data with the national death index (NDI) records.
This study was designed to examine the possibility that the dose response to vitamin C might not be linear (ie, higher doses may not have the same benefits as low doses).
During a median follow-up of 10.6 years, there were 1,558 all-cause deaths, including 320 from cancer, 374 from cardiovascular disease (CVD), and 120 from respiratory diseases.
There was a significantly higher association between all-cause and CVD-associated mortality for those in the lowest and highest quintiles of circulating vitamin C (U-shaped relationship curve).
The authors concluded, “Interestingly, serum vitamin C levels lower than the threshold value (1.06 mg/dL) were negatively associated with all-cause (fully adjusted hazard ratio [HR], 0.71; 95% confidence interval [CI], 0.590.86) and CVD (fully adjusted HR, 0.70; 95% CI, 0.471.03) mortality. In contrast, serum vitamin C levels higher than the threshold value (1.06 mg/dL) were positively associated with all-cause (fully adjusted HR, 1.33; 95% CI, 1.151.54) and CVD (fully adjusted HR, 1.60, 95% CI, 1.232.10) mortality, respectively.”
While the association between vitamin C, cancer, and respiratory disease did not reach statistical significance, there were noticeable trends in the data suggestive of a possible positive association for these conditions as well.
Funding source disclosed, and no concerns raised regarding transparency.
Practice Implications & Limitations
If the findings reported in this study by Tian et al are valid, we should question our long-held assumptions about the safety of water-soluble vitamins. Prior to examining the implications of these data, let us first consider the arguments for these findings being valid.
These results are derived from epidemiological data rather than from randomized clinical trials, and these days the tendency is to place greater faith in blinded-trial data than from epidemiologic data. Still NHANES remains one of the most trusted cohorts from which to mine data, and it allowed the discounting of many possible confounding factors by the paper’s authors.
If one compares characteristics of the study participants based on vitamin C concentrations, the participants ranked into Quintile 5 (those participants with the highest vitamin C levels) were more likely to be White, better educated, more physically active, more affluent, and to consume more fruits and vegetables than those individuals with lower levels of vitamin C. They were also less likely to smoke, have diabetes, or be overweight (lowest mean body mass index [BMI] of all quintiles). Those in Quintile 5 also had the lowest mean homocysteine and C-reactive protein (CRP) of all participants. These characteristics would lead one to predict that these people would have the lowest risk of dying during the study period, but instead they were 77% more likely to die than those in Quintile 3, whose vitamin C levels were at or near the mean of the study participants.
In recent years, it has become acceptable in scientific publications to describe hormetic dose responses as U-shaped or J-shaped, omitting the term hormesis entirely from the discussion.
Second, although this association of high vitamin C with mortality or morbidity rates might seem to have emerged unexpectedly, this isn’t the first time such a nonlinear risk pattern has been reported. The authors note that this study was designed specifically to examine this phenomenon, as it has been reported in earlier research.
In 2016, Cadeau et al reported on vitamin C intake and breast cancer risk. They compared vitamin C intake among the 2,482 cases of invasive breast cancer that had occurred among the 57,403 postmenopausal women in a prospective cohort spanning 581,085 person-years. They used food frequency questionnaires to estimate vitamin C intake and reported that while use of vitamin C supplements was not associated with breast cancer risk across all quintiles, “vitamin C supplement use was associated with increased postmenopausal breast cancer risk in women with high vitamin C intake from foods. Our data suggest a potential U- or J-shaped relation between total vitamin C intake and postmenopausal breast cancer risk that deserves further investigation” (emphasis added).1
In 2018, a large review and meta-analysis by Jayedi et al, which examined “Dietary antioxidants, circulating antioxidant concentrations, total antioxidant capacity, and risk of all-cause mortality,” reported that while most antioxidants were associated with lower risk of dying, their composite data described a U-shaped relationship between vitamin C and mortality.2
Not all studies are finding this type of curve. The 2018 publication by Wang et al on a large Chinese cohort found that higher plasma vitamin C was associated with lower risk of heart disease and cancer in randomly selected elderly subjects; this association followed a simpler, linear association.3
These past reports led to this current study, which was designed to detect whether incrementally higher vitamin C levels might affect risk. Prior studies have focused on comparing inadequate versus adequate levels, never suspecting the possibility of a nonlinear relationship and that at higher circulating levels the relationship might shift.
In recent years U-shaped dose responses have been reported for several other vitamins that we have previously been unconcerned about. In early 2022, Xu et al reported that high serum folate might raise risk of CVD in some populations.4 A few months earlier, in September 2021, researchers reported that vitamin B12’s association with all-cause mortality was also U-shaped and at higher serum levels risk of dying increased.5
While each of these relationships will need to be studied individually in far greater depth before we can confirm or deny their validity, their collective publication in such a short period leaves me wondering if we are seeing this new pattern now only because researchers have only recently started looking for it or if they now simply feel safe to report on what they observe in their data.
The idea that varying doses of a substance might have strikingly different effects on biological systems hearkens back to the Arndt-Schulz Law that describes such biphasic dose-responses. This concept, however, was deeply “marginalized” for many years as it was closely associated with homeopathy. Using the proper term hormesis to describe biphasic dose responses was a hindrance to publication.6 In recent years, it has become acceptable in scientific publications to describe hormetic dose responses as U-shaped or J-shaped, omitting the term hormesis entirely from the discussion. Such recent “acceptability” seems to have allowed publication of a rapidly growing number of papers describing dose responses that clearly fit the definition of hormesis even if described simply as U-shaped curves.
This recent paper by Tian et al should certainly leave us wondering whether many of our patients are doing themselves long-term harm by taking daily doses of vitamin C and maintaining more than adequate serum levels of vitamin C. Past efforts investigating vitamin C first focused on the benefits gained from reducing damage associated with deficiency, and then they looked at long-term benefits of vitamin C against a variety of diseases. This study may be among the first to purposefully look for long-term harm associated with use. Would we even have noticed a problem if we didn’t know to look for it?
Perhaps we should use Tian et al’s data and say that without specific indications for need, our goal should be to keep serum vitamin C at close to 1.06 mg/dL or alternatively use the daily dosage of 125 mg/day as suggested by Jayedi et al. I write “perhaps” because such suggestions may seem strikingly low to both patients and practitioners alike and would require abandonment of long-held assumptions.