Search strategy and identification of the studies

The study was carried out using a detailed protocol developed in advance, including predefined research questions and objectives, search strategy, study eligibility criteria, the methods of data extraction and statistical analysis. All the variables for subgroup analysis were predefined. We used the statement of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) for reporting the present study⁽ Reference Moher, Liberati and Tetzlaff ^{9 )}.

We searched the English-language medical literature published between January 1980 and November 2013 using the Medline and Google scholar database. We used structured search strategy using various combinations of keywords for vitamin D (online Supplementary Table S1). We also checked the references of recent systematic reviews that investigated the effects of oral intake or intramuscular injection of vitamin D supplements to find additional relevant studies.

RCT on vitamin D (with or without Ca) supplementation that involved apparently healthy human subjects or patients whose disease has no effect on vitamin D metabolism were included in the analysis. RCT were selected because the greatest validity and causal interference can be found in such studies⁽ Reference Richter and Berger ^{10 )}.

We included studies that fulfilled the following criteria: (1) vitamin D₃ ≥10 μg/d administered orally per se or with Ca on a daily basis (inclusion of vitamin D₃ and D₂ was chosen, although the Institute of Medicine dietary recommended intakes (IOM) DRI committee has defined DRI based on studies with vitamin D₃ ⁽ Reference Ross, Taylor and Yaktine ^{11 )} and there is evidence that vitamin D₃ increases serum 25(OH)D more efficiently than vitamin D₂ ⁽ Reference Trang, Cole and Rubin ^{12 ,} Reference Armas, Hollis and Heaney ^{13 )}); (2) separately reported serum or plasma 25(OH)D levels in intervention and control groups; (3) separately reported serum or plasma PTH levels in intervention and control groups; (4) a minimum duration of 6 weeks, because serum 25(OH)D concentrations reach equilibrium after at least 6–8 weeks in adults⁽ Reference Shab-Bidar, Bours and Geusens ^{5 ,} Reference Harris and Dawson-Hughes ^{14 ,} Reference Viljakainen, Palssa and Karkkainen ^{15 )}.

The exclusion criteria included the following: (1) use of compounds such as vitamin D metabolites (25(OH)D and 1,25(OH)2D) and analogues (e.g. α-calcidol) co-administered; (2) studies carried out in infants, children, adolescents and pregnant or lactating women; (3) studies in which vitamin D was administered as fortified food; (4) interventions that included patients with chronic renal disease, chronic heart disease, cirrhosis and hyperparathyroidism; (5) RCT that used cluster randomisations and cross-over studies; (6) trials without control or placebo groups; (7) studies published in languages other than English, because effect sizes did not differ significantly in language-restricted meta-analyses compared with language-inclusive ones⁽ Reference Moher, Klassen and Schulz ^{16 )}, as well as lower quality in the non-English medical literature⁽ Reference Jüni, Holenstein and Sterne ^{17 )}; (8) repeated studies, if the results of the trials had been published in more than one article, we used the reporting results on the largest sample of individuals, or the most recently published or the more detailed results; (9) abstracts, because of insufficient information; and (10) dissertations, because the full text was rarely available.

Variations between the extracted studies regarding supplement dosage, frequency of supplementation and use of either intramuscular or oral delivery methods were acceptable and were not excluded.

Data collection and synthesis

To identify and include eligible studies in the final analysis, two authors (S. S.-B. and N. M.), independently, reviewed the titles of the articles extracted by the search for relevance to our topic, and then we retrieved the full-text articles of those that were potentially relevant. Screening list was used to select eligible articles. Backward search was carried out through published reviews previously and those published after our search date.

Moreover, the quality control of the articles was carried out independently by two authors (S. S.-B. and N. M.). Discrepancies between authors were solved by consensus with the third author (F. H.). We included only data reported in the study, because recall bias in the information or data might be provided by authors⁽ Reference Steinberg, Smith and Stroup ^{18 )}.

All relevant information were abstracted on study characteristics including the following: first author, publication year, country of origin, study design, the number of participants in each arm of RCT, age, sex, the dose of supplement, frequency supplement use, duration of supplementation, type of supplement used in the RCT, mean values and standard deviations of the baseline and final values for 25(OH)D and PTH in the treatment and control arms at each time point and for each vitamin D dose. In studies with different doses, we included each dose as a separate study and used the dose subgroups v. controls separately. If a study had several intervals for follow-up measurements of 25(OH)D, we included each time interval as a separate study. If studies had subgroups such as sex, they were included in our study as a separate study.

For any other information pertinent to the review, such as potential confounders to the RCT (i.e. the season of implementing the intervention and BMI), the analysis technique chosen to assess serum 25(OH)D and the dropout rates were also noted when reported.

Quality of assessment

We assessed the quality of studies using Jadad scales⁽ Reference Jadad, Moore and Carroll ^{19 )}, which include the following four items: reporting of randomisation method, allocation concealment, blinding of outcome assessment and completeness of follow-up (online Supplementary Table S2).

Statistical analysis

The mean difference (MD) of achieved levels of 25(OH)D and PTH between the intervention and control groups for each individual study was calculated. If the standard error was reported for variation of mean, we calculated sd by dividing se/n ². For the calculation of the standard deviation from the range and confidence intervals, we divided the range by 5·88 and CI by 3·92.

Cochran’s Q statistic and the I ² statistic were used to assess statistical heterogeneity in the meta-analysis⁽ Reference Cochran ^{20 )}. Both the fixed-effects and random-effects models were used to calculate the pooled MD of PTH level in response to vitamin D. In this review, we present results from the random-effects model because significant heterogeneity was identified among studies⁽ Reference DerSimonian and Laird ^{21 )}.

Potential sources of heterogeneity were also investigated in predefined subgroups. We assessed treatment effects in preset subgroups: (1) dose (≤20 and >20 μg); (2) vitamin supplementation with or without Ca; (3) Ca dose ( no Ca, 400–600 and 600–1200); (4) duration (<3, 3–6, 6–12 and >12 months); (5) baseline 25(OH)D (<20 or ≥20 ng/ml); (6) baseline PTH (≤6·0, 6·1–38·0, 38·1–49·0 and ≥49·0); (7) BMI (>25, 25–30 and ≥30 kg/m²); (8) sex (men, women, both); (9) age (<50 and ≥50 years); and (10) study quality (low quality v. high quality).

The meta-regression was used to analyse factors within a trial that best explained the variance in MD of PTH. Using meta-regression, we analysed the effects of daily doses of vitamin D, duration of the trial, baseline 25(OH)D, baseline PTH, BMI and age on MD.

We performed ancillary analyses including curve estimation models for weighted mean difference (WMD) of serum levels of PTH according to dose and duration and baseline 25(OH)D and PTH as continuous variables.

A cumulative meta-analysis⁽ Reference Lau, Antman and Jimenez-Silva ^{22 )} was also performed to determine that the evidence was consistent over time. Influence analysis was carried out to show that no particular trial affected the pooled effect size.

A formal statistical test on publication bias was not meaningful because we excluded studies with sample sizes <30. However, publication bias was analysed by funnel plot analysis (online Supplementary Fig. S1) and Egger’s regression asymmetry test for the included studies⁽ Reference Egger, Davey Smith and Schneider ^{23 ,} Reference Duval and Tweedie ^{24 )}. In our analysis, the summary estimate for PTH was statistically significant when we included Suzuki et al.’s study⁽ Reference Suzuki, Yoshioka and Hashimoto ^{25 )} that reported a large WMD. We then considered this study to be a possible outlier, and thus excluded the study from our analysis. All tests were two-tailed, and a probability level <0·05 was considered to be statistically significant. Statistics were performed using Stata version 12.0 (Stata Corporation) and SPSS version 18.

Study characteristics

Of the 2360 studies identified, thirty-three studies⁽ Reference Chapuy, Chapuy and Meunier ^{26 –} Reference Suzuki, Yoshioka and Hashimoto ^{58 )} including fifty intervention groups with 7574 participants (n 3851 in intervention group and n 3663 in placebo group) were selected for the present meta-analysis (Fig. 1). All of them were RCT; however, sixteen studies did not clarify the method of randomisation⁽ Reference Chapuy, Chapuy and Meunier ^{26 –} Reference Krieg, Jacquet and Bremgartner ^{31 ,} Reference Pfeifer, Begerow and Minne ^{33 ,} Reference Grados, Brazier and Kamel ^{35 ,} Reference Bischoff, Stahelin and Dick ^{36 ,} Reference Chel, Wijnhoven and Smit ^{41 –} Reference Pfeifer, Begerow and Minne ^{44 ,} Reference Jorde, Sneve and Torjesen ^{47 ,} Reference Chung, Chin and Kang ^{50 ,} Reference Harris, Pittas and Palermo ^{53 )}. The mean age of the participants ranged from twenty-one to 85 years. The daily doses of vitamin D supplementation varied from 10 to 250 μg/d; only two studies supplemented vitamin D in the form of ergocaciferol⁽ Reference Chapuy, Chapuy and Meunier ^{26 ,} Reference Sokol, Srinivas and Crandall ^{51 )}, and two studies did not report the form of vitamin D supplement used⁽ Reference Dawson-Hughes, Dallal and Krall ^{27 ,} Reference Kenny, Biskup and Robbins ^{34 )}. The duration of supplementation ranged from 2 to 36 months. The majority of studies were conducted on women or on women and men together; only two studies were conducted only on men⁽ Reference Dawson-Hughes, Harris and Krall ^{30 ,} Reference Kenny, Biskup and Robbins ^{34 )}. Using the Jadad scale, 81·8 % of the studies were of high quality (scores≥3), with an average score of 3·6. Six studies were considered to be of low quality (scores≤2)⁽ Reference Chapuy, Chapuy and Meunier ^{26 ,} Reference Chapuy, Arlot and Duboeuf ^{28 ,} Reference Krieg, Jacquet and Bremgartner ^{31 ,} Reference Chel, Wijnhoven and Smit ^{41 ,} Reference Jorde, Sneve and Torjesen ^{47 ,} Reference Harris, Pittas and Palermo ^{53 )} (online Supplementary Table S2). Study characteristics are summarised in Table 1.

Fig. 1 Flow chart of study selection for inclusion in the systematic review. 25(OH)D, 25-hydroxyvitamin D; PTH, parathyroid hormone.

Table 1 Study and participant characteristics

1, male; 2, female; 3, male and female; 25(OH)D, 25-hydroxyvitamin D; MD, mean difference; PTH, parathyroid hormone; IFG, impaired fasting glucose.

* Final serum 25(OH)D in the intervention group minus final serum 25(OH)D in the placebo group.

† Final serum PTH in the intervention group minus final serum PTH in the placebo group.

Meta-analysis for serum vitamin D responses

The pooled mean difference (PMD) of 25(OH)D from the pre-trial was +15·5 ng/ml (−5 to +40 ng/ml) in the intervention group. The forest plot with MD in post-trial 25(OH)D concentrations between intervention and placebo groups and their 95 % confidence intervals are illustrated in Fig. 2. As there was significant heterogeneity between studies (test for heterogeneity: P<0·001 and I ²=97·3 %), we used the random-effects model to estimate the PMD in serum vitamin D concentration. Vitamin D supplementation resulted in a PMD of 15·52 ng/ml in serum 25(OH)D concentration (95 % CI 15·38, 15·67).

Fig. 2 Effect of vitamin D supplementation on serum 25-hydroxyvitamin D.

Meta-analysis for serum parathyroid hormone response

The PMD of serum PTH from the pre-trial was −10·17 pg/ml (−11·83, −8·50 to +7·5 pg/ml) in the intervention group. Individual and pooled MD in serum PTH concentration and 95 % CI after vitamin D supplementation that were derived from a random-effects model have been illustrated in Fig. 3. The meta-analysis demonstrated that the vitamin D supplementation decreased PTH levels significantly in the intervention group compared with the placebo (PMD: −10·17; 95 % CI −11·84, −8·50). There was significant heterogeneity between studies (test for heterogeneity: P<0·001), and the I ² value was 97·3 %, which can be interpreted as the amount of variation across the studies being attributed to heterogeneity rather than chance.

Fig. 3 Effect of vitamin D supplementation on serum parathyroid hormone (PTH).

Subgroup meta-analysis for serum parathyroid hormone response

Each subgroup analyses significantly affected the treatment effect except for the dose of vitamin D supplementation (Table 2). There was a very small non-significant difference in PMD of serum PTH between vitamin D dosages of ≤20 and >20 μg/d (−2·98 (95 % CI −3·24, −2·72) v. −3·05(95 % CI −3·28, −2·81)) (P=0·713). The addition of Ca to vitamin D supplementation increased the treatment effect of vitamin D supplementation (−4·08 (95 % CI −4·33, −3·82) v. −2·09 (95 % CI −2·33, −1·85); P<0·001). The treatment effect was also the best with Ca doses of 600–1200 mg/d. Duration of vitamin D supplementation changed the treatment effect significantly, the best effect being observed when the trial duration was >12 months. Participants with low baseline 25(OH)D concentration (25(OH)D <20 ng/ml) had higher PMD of serum PTH than those whose serum 25(OH)D was ≥20 ng/ml (−16·70 (95 % CI −17·75, −15·84) v. −2·44 (95 % CI −2·62, −2·26); P<0·001). Baseline serum PTH also affected responses to vitamin D supplementation; participants with highest baseline serum PTH had the highest PMD. The treatment effect was lower in people aged >50 years than in those who were under 50 (−2·98 (95 % CI −3·16, −2·81) v. −6·92 (95 % CI −8·74, −5·09)); P<0·001). The treatment effect was the greatest in people with BMI ranging from 25 to 30 kg/m² compared with those with BMI <25 kg/m² (−18·11 (95 % CI −19·07, −17·15) v. −2·01 (95 % CI −2·26, −1·77)) and those with BMI ≥30 kg/m² (−18·11 (95 % CI −19·07, −17·15) v. −5·86 (95 % CI −7·92, −3·80); P<0·001). The treatment effect appeared to be greater in men-only studies compared with those conducted only in women (−11·34 (95 % CI −18·63, −4·05) v. −2·95 (95 % CI −3·13, −2·77); P<0·001).

Table 2 Subgroup analysis for effectiveness of vitamin D supplementation on serum parathyroid hormone (PTH) (Mean differences (MD) and 95 % confidence intervals)

25(OH)D, 25-hydroxyvitamin D.

Meta-regression and source of heterogeneity for serum parathyroid hormone responses

We used univariate meta-regression analysis to examine the variation in treatment effect attributed to some pre-specified covariates. The univariate meta-regression analysis showed that none of the covariates including the dose of vitamin D supplementation, dose of Ca supplementation, baseline serum PTH, age, duration of trial and baseline serum 25(OH)D concentrations have significant effects on between-study heterogeneity (Fig. 4 and Table 3).

Fig. 4 Meta-regression analysis of baseline serum parathyroid hormone (PTH) (a), dose of vitamin D supplementation (b), dose of Ca supplementation (c) and trial duration (d).

Table 3 Summary of the meta-regression analysis (Slope and 95 % confidence intervals)

25(OH)D, 25-hydroxyvitamin D; PTH, parathyroid hormone.

Ancillary analysis

Using curve estimation regression models, we found non-linear associations between dose of vitamin D supplementation and WMD in the post-trial serum PTH concentrations. WMD in the post-trial PTH was negatively and quadratically correlated with the dose of vitamin D supplementation (R ² 0·03, P<0·001; Fig. 5(a)), duration of the trial (R ² 0·01, P<0·001; Fig. 5(b)) and 25(OH)D concentration (R ² 0·01, P<0·001; Fig. 5(c)), reaching a plateau following a dosage of 75 μg/d after 12 months and at baseline 25(OH)D of 30 ng/ml.

Fig. 5 Correlation of mean differences (MD) in parathyroid hormone (PTH) level with (a) dose of vitamin D, (b) duration of the trial and (c) baseline 25-hydroxyvitamin D (25(OH)D). , Observed; , Linear; , Logarithmic; , Quadratic.

Cumulative and influence analysis

No individual study was found to have excessive influence on the pooled effect when the influence analysis was carried out (Fig. 6). A cumulative random-effect meta-analysis showed consistency from the year 2000 (Fig. 7).

Fig. 6 Influence analysis.

Fig. 7 Cumulative analysis.

Publication bias

An asymmetric funnel plot suggested a possible publication bias (online Supplementary Fig. S1). Egger’s linear regression also confirmed publication bias among studies (P=0·003), which is not unexpected because publication bias testing does not work when the meta-analysis has only selected RCT with a minimum of thirty participants. Publication bias, including funnel plot, assumes that all published studies are included and what is missing are the unpublished studies. However, the trim and fill method did not reveal any missing study, and thus the PMD estimate in post-trial PTH concentration remained unchanged.

Dose of vitamin D

In the present study, meta-regression analysis did not show the dose of vitamin D as a source of heterogeneity among studies. Vitamin D supplementation significantly decreased PTH concentrations in forty trials⁽ Reference Krieg, Jacquet and Bremgartner ^{31 ,} Reference Talwar, Aloia and Pollack ^{38 –} Reference Chel, Wijnhoven and Smit ^{41 ,} Reference Pfeifer, Begerow and Minne ^{44 –} Reference Islam, Shamim and Viljakainen ^{46 ,} Reference Lips, Binkley and Pfeifer ^{48 ,} Reference Grimnes, Figenschau and Almas ^{49 ,} Reference Sokol, Srinivas and Crandall ^{51 –} Reference Kjaergaard, Waterloo and Wang ^{55 ,} Reference Goswami, Vatsa and Sreenivas ^{57 ,} Reference Ooms, Lips and Roos ^{59 –} Reference Grados, Brazier and Kamel ^{70 )}, thirty of which were in vitamin D-deficient populations⁽ Reference Krieg, Jacquet and Bremgartner ^{31 ,} Reference Talwar, Aloia and Pollack ^{38 ,} Reference Sneve, Figenschau and Jorde ^{40 –} Reference Bjorkman, Sorva and Risteli ^{42 ,} Reference Zittermann, Frisch and Berthold ^{45 ,} Reference Islam, Shamim and Viljakainen ^{46 ,} Reference Lips, Binkley and Pfeifer ^{48 ,} Reference Grimnes, Figenschau and Almas ^{49 ,} Reference Sokol, Srinivas and Crandall ^{51 –} Reference Harris, Pittas and Palermo ^{53 ,} Reference Kjaergaard, Waterloo and Wang ^{55 ,} Reference Goswami, Vatsa and Sreenivas ^{57 ,} Reference Salehpour, Hosseinpanah and Shidfar ^{60 ,} Reference Bischoff, Stähelin and Dick ^{64 –} Reference Chapuy, Arlot and Duboeuf ^{68 ,} Reference Grados, Brazier and Kamel ^{70 ,} Reference Ooms, Roos and Bezemer ^{71 )}. It, however, increased PTH levels in eight trials⁽ Reference Cashman, Hill and Lucey ^{43 ,} Reference Pfeifer, Begerow and Minne ^{44 ,} Reference Chung, Chin and Kang ^{50 ,} Reference Jorde, Sneve and Torjesen ^{63 ,} Reference Dawson-Hughes, Dallal and Krall ^{72 ,} Reference Hunter, Major and Arden ^{73 )} and caused no changes in one trial⁽ Reference Chapuy, Arlot and Duboeuf ^{68 )}. Thirty of those trials with decrease in serum PTH used a vitamin D dose ≥20 μg/d, and ten studies used vitamin D doses >75 μg/d. In those studies in which PTH responded to vitamin D, the mean vitamin D supplementation was 57 μg/d and mean baseline 25(OH)D was 25·2 ng/dl, whereas in those studies where PTH did not respond to vitamin D supplementation the mean dosage of vitamin D was 30·5 μg/d with baseline 25(OH)D level of 16·6 ng/dl. Cranny et al. ⁽ Reference Cranney, Horsley and O’Donnell ^{74 )} have found that vitamin D₃ doses ≥17·5 μg daily, significantly and consistently decreased serum concentrations of PTH in vitamin D-deficient populations. As Cranny et al. ⁽ Reference Cranney, Horsley and O’Donnell ^{74 )} mentioned in their systematic review, reasons for lack of achievement of reduction in serum PTH in some studies may be due to a very low amount of the vitamin D dose for a population with low baseline 25(OH)D concentrations. In addition, changes in PTH level may not occur with baseline serum 25(OH)D above the threshold of PTH suppression⁽ Reference Cranney, Horsley and O’Donnell ^{74 )}.

PTH level plateaued in a quadratic model at a dose of vitamin D >75 μg/d, a finding in contrast with the first dose–response RCT in older white women by Gallagher et al. ⁽ Reference Gallagher, Sai and Templin ^{75 )}, who found a linear relationship between vitamin D₃ dose and PTH level. The quadratic dose term and interaction between quadratic dose and time were NS in the PTH model. Heaney et al. ⁽ Reference Heaney, Davies and Chen ^{76 )} reported that the 25(OH)D level that PTH will suppress is 75 nmol/l. Interestingly, Vieth et al. reported that a dose above 82·5 μg of ergocalciferol and 20 μg of cholecalciferol was needed to ensure post-trial 25(OH)D levels of at least 50 nmol/l, whereas to ensure mean post-trial 25(OH)D levels of at least 75 nmol/l doses of 12·5 μg/d and 71·25 μg/d are needed⁽ Reference Björkman, Sorva and Tilvis ^{8 ,} Reference Vieth, Bischoff-Ferrari and Boucher ^{77 )}. It was also reported that very high doses of vitamin D can certainly increase 25(OH)D to levels high enough to suppress PTH, but there are sparse data available on this. It is also interesting to note that it was estimated that intoxication may not occur with 25(OH)D levels up to 375 nmol/l⁽ Reference Björkman, Sorva and Tilvis ^{8 ,} Reference Holick ^{78 )}.

An earlier meta-analysis by Shab-Bidar et al. concluded that the treatment effect of oral vitamin D₃ supplementation increases with increasing doses. Meta-regression results demonstrated a significant association between dose and serum 25(OH)D levels (P=0·04)⁽ Reference Shab-Bidar, Bours and Geusens ^{5 )}, and the results were confirmed by Cranny⁽ Reference Cranney, Horsley and O’Donnell ^{74 )} who suggested that 2·5 μg of vitamin D₃ increases serum 25(OH)D concentrations by 1–2 nmol/l, and therefore vitamin D supplements at doses of 10–20 μg daily may be inadequate to prevent vitamin D deficiency in at-risk individuals⁽ Reference Cranney, Horsley and O’Donnell ^{74 )}.

Duration of trial

Based on the findings of the present study, the best effect of treatment with vitamin D on PTH response was observed when the duration of the trial was >12 months. An increase in PMD was found, which plateaued after 12 months. Previous trials reported no significant change in PTH levels after 3 months of vitamin D supplementation⁽ Reference Shab-Bidar, Bours and Geusens ^{5 )}, an observation, however, supported by Gallagher et al.⁽ Reference Gallagher, Sai and Templin ^{75 )}, who also observed significant decreases in serum PTH levels with increasing vitamin D doses at 12 months.

Calcium intake

Serum PTH response may be partially modulated by the amount of Ca intake through diet or combined supplementation of vitamin D with Ca⁽ Reference Lips, Duong and Oleksik ^{79 )}. We noted a higher treatment effect in individuals with Ca–vitamin D supplementation than in those who were supplemented only with vitamin D (−4·08 v. −2·09; P<0·001). The treatment effect was also the best with Ca doses of 600–1200 mg/d, which is important because PTH suppression may not be ensured without sufficient Ca intakes, especially when there are several reports in which inadequate dietary Ca is prevalent throughout the world. In contrast, data from another study suggest that vitamin D sufficiency can ensure ideal serum PTH values even when the Ca intake level is <800 mg/d, whereas high Ca intake (>1200 mg/d) is not sufficient to maintain ideal serum PTH, as long as the vitamin D status is insufficient⁽ Reference Steingrimsdottir, Gunnarsson and Indridason ^{80 )}. This is further reflected in ionised Ca levels that were dependent on serum 25(OH)D levels but not on Ca intake.

Another study concluded that vitamin D supplementation had a reducing effect on serum PTH only when the vitamin D per se was given⁽ Reference Seamans and Cashman ^{81 )}. Although sufficient intakes of vitamin D and Ca are definitely important, Ca intake may not necessarily be a contributing factor in maintaining Ca homoeostasis as long as vitamin D status would be benefitted with vitamin D supplementation and sun exposure⁽ Reference Baron, Beach and Mandel ^{82 )}. Other investigators have suggested that the response of circulating 25(OH)D to supplemental vitamin D was similar whether Ca was co-administered or not⁽ Reference Weaver and Fleet ^{83 )}. Aloia et al. ⁽ Reference Aloia, Talwar and Pollack ^{84 )} reported that most of the studies examining optimal vitamin D status do not control for Ca intake, and they found that serum 25(OH)D and dietary Ca influence the PTH threshold independently and together account for about 67 % of the variance in reported thresholds among the studies. The contribution of dietary Ca to the prediction of the threshold remained significant even after controlling for serum 25(OH)D⁽ Reference Aloia, Talwar and Pollack ^{84 )}.

One study suggested that the response of serum PTH differs by Ca intake, only in those individuals with low vitamin D status, which can be explained by the less-active transport of Ca⁽ Reference Steingrimsdottir, Gunnarsson and Indridason ^{80 )}. It has been suggested that in the absence of sufficient active Ca transport in the gut, as in vitamin D insufficiency, one must meet the requirements of the body with higher Ca intakes⁽ Reference Heaney and Feldman ^{85 )}.

We explored the interaction between baseline 25(OH)D and Ca intake, and found that in studies with 25(OH)D >50 nmol/l Ca intake did not affect PTH response, whereas in those with a mean 25(OH)D <50 nmol/l dietary Ca was inversely related to PTH (data not shown). These results are in agreement with those of Aloia et al. ⁽ Reference Aloia, Talwar and Pollack ^{84 )} following a study of African-Americans.

Some studies have discussed the sparing effect of dietary Ca intake on serum 25(OH)D because PTH concentrations are suppressed, thus less serum 25(OH)D is converted to 1,25(OH)D⁽ Reference Lips ^{86 )}.

Baseline 25-hydroxyvitamin D concentration

In this study, participants with low baseline 25(OH)D concentration (25(OH)D<20 ng/ml) had more reduction in serum PTH than those in whom serum 25(OH)D was ≥20 ng/ml (−16·70 v.−2·44; P<0·001). According to our previous study, baseline 25(OH)D was one of the important determinants of response to vitamin D supplementation⁽ Reference Shab-Bidar, Bours and Geusens ^{5 )}. In the present study, we categorised both vitamin D deficiency and insufficiency together. Vitamin D deficiency is known to be associated with secondary hyperparathyroidism, increased bone turnover and bone loss⁽ Reference Lips ^{87 )}. A negative correlation between serum PTH and serum 25(OH)D levels has been reported by many investigators⁽ Reference Lips, Duong and Oleksik ^{79 )}. We expect that betterment of vitamin D deficiency would follow after significant improvements of PTH concentration.

Some have argued that serum PTH declines significantly after vitamin D and Ca intervention is initiated with low baseline serum 25(OH)D levels <20 ng/ml (<50 nmol/l)⁽ Reference Cranney, Horsley and O’Donnell ^{74 )}. Interestingly, Lips et al.⁽ Reference Lips, Duong and Oleksik ^{79 )} demonstrated that the mean serum PTH level was 30 % higher in those with low serum 25(OH)D (<25 nmol/l) than in women with higher serum 25(OH)D (<50 nmol/l). Based on the findings of the Gallagher et al. ⁽ Reference Sai, Walters and Fang ^{88 )} study, clinical importance was only observed in 25(OH)D-deficient status and elevated PTH level. However, the threshold of 25(OH)D to prevent a rise in PTH concentration varies widely, as many studies have found most estimates clustered between 40 and 50 nmol/l or between 70 and 80 nmol/l. The variability in the estimates may be due to different Ca intakes, different 25(OH)D assays, age of the participants and vitamin D insufficiency⁽ Reference Aloia, Talwar and Pollack ^{84 )}.

In the meta-regression analysis, we found a non-significant association between baseline 25(OH)D and response of PTH. Aloia et al. ⁽ Reference Aloia, Talwar and Pollack ^{84 )} in a review of twenty-five studies reported that the average correlation between PTH and vitamin D was −0·30 and serum 25(OH)D just contains 9 % of the variance in PTH.

Age

The treatment effect was lower in people aged >50 years than in those who were younger than 50 (−2·98 v. −6·92); P<0·001). Previous observations have demonstrated that older participants had a better response to vitamin D₃ intake, although the response was independent of baseline 25(OH)D⁽ Reference Shab-Bidar, Bours and Geusens ^{5 )}. Bjorkman et al. also showed that age of the patients can have major effects on the elevation of PTH levels independently. The higher effect could be attributed to the high prevalence of vitamin D deficiency in the elderly⁽ Reference Gloth, Tobin and Sherman ^{89 ,} Reference Dixon, Mitchell and Beringer ^{90 )}. We expected that following amelioration of vitamin D deficiency PTH level might be suppressed maximally. However, the better response to vitamin D intake was not enough to guarantee PTH suppression in the elderly, as the achieved 25(OH)D was not sufficient. Indeed, skin content of 7-dehydrocholesterol drops by 50 % between 20–80 years of age⁽ Reference MacLaughlin and Holick ^{91 )}, and the same dose of UV-B radiation in older individuals produces a smaller rise in serum 25(OH)D compared with young individuals⁽ Reference Holick, Matsuoka and Wortsman ^{92 )}. Ageing is associated with a decline in renal function, and higher concentrations of 25(OH)D are needed to prevent a rise in serum PTH in the elderly⁽ Reference Vieth, Ladak and Walfish ^{93 )}.

BMI

In the present study, the treatment effect was the highest in people who were overweight and obese. There is an altered vitamin D endocrine system in obese individuals⁽ Reference Bell, Epstein and Greene ^{94 )}. Studies have shown that obesity, and specifically body fat content, is inversely associated with 25(OH)D and is positively associated with PTH concentrations⁽ Reference Bolland, Grey and Ames ^{95 ,} Reference Snijder, van Dam and Visser ^{96 )}. In a recent study by Gallagher et al.⁽ Reference Gallagher, Sai and Templin ^{75 )}, underweight to normal weight and the overweight groups tended to have lower PTH levels than the obese group (P=0·065). It has been reported that, with a similar amount of 7-dehydrocholesterol in the epidermis, the increase in serum 25(OH)D after UV-B irradiation was 57 % less in obese compared with non-obese subjects⁽ Reference Wortsman, Matsuoka and Chen ^{97 )}. It is suggested that lower serum 25(OH)D may be a factor partially contributing to the relationship of higher serum PTH with greater adiposity⁽ Reference Bolland, Grey and Ames ^{95 ,} Reference Snijder, van Dam and Visser ^{96 ,} Reference Shapses, Lee and Sukumar ^{98 )}. In a recent study, Shapses et al. showed that PTH is suppressed at a lower 25(OH)D concentration in the obese compared with the entire population. Therefore, the lower average 25(OH)D concentrations in the obese may not have the same physiological significance as in the general population. Evidence also shows that, in spite of physiological changes associated with the higher BMI, including higher PTH levels and higher bone resorption, bone mineral density may not be reduced in overweight women⁽ Reference Macdonald, Mavroeidi and Barr ^{99 )}.

Limitations

This meta-analysis has some limitations. First, many analyses suffer from high levels of heterogeneity, but this is not unexpected because the included RCT had variable population groups, doses and supplementation forms (vitamin D₂ or D₃, with or without supplemental Ca). Second, our search was limited to the published studies. Third, not all studies reported data for seasonal influences, sun exposure, physical activity and dietary intake of vitamin D and Ca; therefore, we were unable to adjust for these variables in our analysis. Fourth, multiple comparisons in the subgroup analysis may increase the likelihood of type 1 error. Finally, the validity of the study results may be influenced by the use of different assay types.

Conclusion

In conclusion, although the present meta-analysis was hindered by some limitations, all of which contributed to the heterogeneity, it provides some interesting evidence, suggesting that suppression of PTH level needs higher vitamin D intake (75 μg/d) and longer duration (12 months) than those currently recommended, which should be taken into account for nutritional recommendations.