Trial design and optimal status: What can we learn from the DOIT vitamin D trial

NutraIngredients recently reported​ the results from a Canadian vitamin D trial - known as the vitamin D Outcomes and Interventions in Toddlers (DO IT) trial - which found daily doses of 2000IU were no more effective than 400IU in protecting against winter colds.

One simple explanation for this result is that baseline vitamin D blood levels in both high-dose and standard-dose groups were already optimal - at 91 nanomoles per litre (36.5 nanogrammes per millilitre).

The study leader, Dr Jonathon Maguire of St. Michael’s Hospital, Toronto  concluded: "We may have just busted a myth. More is not always better. Our findings do not support the routine use of high dose vitamin D supplementation for the prevention of wintertime upper respiratory tract infections among healthy children."​

Such statements risk unfairly discrediting the whole rationale for vitamin D supplementation. If the children’s baseline had been deficient or insufficient, there would have been full justification in supplementing 2000 IU/day.

Furthermore, anti-supplementation advocates may also seize on the ‘no benefit’ outcome to hinder the use of high-dose supplements in trials where such a dose is actually appropriate.

We must ask why researchers would not pay more attention to elements of trial design such as baseline status?

Randomised controlled trials (RCTs) are extremely expensive to conduct, so surely it would be more appropriate to select subjects who were not already replete – and therefore highly unlikely to show benefit from a high-dose supplementation.

On a similar vein, why give high-doses of supplements to a vulnerable age cohort (1-5 years) when previous research has shown rising vitamin status to the high levels reported in this trial - of 48.7 ng/ml (122 nmol/l) - might actually raise disease risk?!

Vitamin D thresholds​

National Institute for Clinical Excellence (NICE) guidelines define vitamin D deficiency as a blood level of below 12 ng/ml (30 nmol/l), while insufficiency is classified as between 12 - 20 ng/ml (30 – 50  nmol/l).

These levels have been set primarily for the purposes of prevention of rickets, and other aspects of bone disease. However, previous research has established vitamin D deficiency raises the risk of other conditions including some cancers, multiple sclerosis, diabetes, preeclampsia, depression and elements of heart disease.

Defining 'optimal' vitamin D level remains difficult and somewhat controversial.

This is because most research linking any disease risk with blood concentrations comes from observational studies. Furthermore, 'optimal status' may be seen as disease specific.

For example, for optimal bone health, a vitamin D level which minimises parathyroid hormone (PTH) secretion is widely considered optimal. Research suggests this occurs around 30-36 ng/ml (75-90 nmol/l).

In contrast, numerous studies suggest cancer risk may be minimised at between 36-48 ng/ml (90-120 nmol/l).

A ‘J’ or ‘U’-shaped curve?​

Different meta-analyses have tried to put together an all–cause mortality (ACM) hazard ratio curve related to serum vitamin D. But even with pooled groups of studies the findings have varied, with some reporting a reverse J-shaped curve​ and others suggesting a somewhat flattened U-shaped​ relationship. 

In the reverse J-shaped relationship, risk minimisation occurs around 50 ng/ml (125 nmol/l), with little additional benefit observed at above 40 ng/ml (100 nmol/l). According to the U-shaped curve, ACM risk is minimised at around 28 ng/ml (70 nmol/l).

But at what vitamin D level does further benefit cease, and an increased disease risk kick in?

The reverse J-shaped curve is flat from 50 ng/ml (125 nmol/l) to its cut-off point at 70 ng/ml (175nmol/l); -with no documented data above this level.

According to the U-shaped curved, ACM hazard ratio increases to 1.50 at 48 ng/ml (120 nmol/l) – in other words, the same disease risk as at the lower end of the insufficient concentration range – around 14 ng/ml (35 nmol/l).

The U-shaped curve study does not show data beyond the 48 ng/ml (120 nmol/l) level, however, even linear extrapolation of the graph (a likely underestimate of a steepening curve), implies that at a vitamin D level of 56 ng/ml (140 nmol/l), the ACM hazard ratio is around 1.78 – potentially as harmful as at the deficient level of 8 ng/ml (20 nmol/l).

Assuming the reverse J-shape relationship is applicable, no additional disease risk reduction is achieved by raising vitamin D level beyond 50 ng/ml (125nmol/l). However, advocates of a U-shaped curve would limit supplementation to restrict vitamin D blood concentration to an even lower level.

The threshold for vitamin D toxicity (hypervitaminosis D), resulting in hypercalcaemia, is well established as 150 ng/ml (375 nmol/l).

Implications​

No matter which hypothesised curve shape is correct, there are important implications for supplementation trial design. Firstly, should supplementation trials ever be conducted in subjects who are already at potentially optimal levels?

The children in the DO IT trial were already above optimal levels (according to the U-shape hypothesis), and even according to the J-shape curve would receive little benefit from high-dose supplementation.

Critics might argue that these curve shapes are aggregated from many different disease outcome measures and therefore suffer from ‘heterogeneity’. However, even a previous observational study on upper respiratory tract infections suggested no further reduction in incidence occurs above 38 ng/ml (95 nmol/l) – only fractionally above the DO IT trial baseline levels.

Secondly, given the possible validity of the U-shape curve, should very young children (1-5 years) have been exposed to a supplementation programme that might have increased their all-cause mortality risk?