Ferrous Sulfate Supplementation Causes Significant Gastrointestinal Side-Effects in Adults: A Systematic Review and Meta-Analysis

Background The tolerability of oral iron supplementation for the treatment of iron deficiency anemia is disputed. Objective Our aim was to quantify the odds of GI side-effects in adults related to current gold standard oral iron therapy, namely ferrous sulfate. Methods Systematic review and meta-analysis of randomized controlled trials (RCTs) evaluating GI side-effects that included ferrous sulfate and a comparator that was either placebo or intravenous (IV) iron. Random effects meta-analysis modelling was undertaken and study heterogeneity was summarised using I2 statistics. Results Forty three trials comprising 6831 adult participants were included. Twenty trials (n = 3168) had a placebo arm and twenty three trials (n = 3663) had an active comparator arm of IV iron. Ferrous sulfate supplementation significantly increased risk of GI side-effects versus placebo with an odds ratio (OR) of 2.32 [95% CI 1.74–3.08, p<0.0001, I2 = 53.6%] and versus IV iron with an OR of 3.05 [95% CI 2.07-4.48, p<0.0001, I2 = 41.6%]. Subgroup analysis in IBD patients showed a similar effect versus IV iron (OR = 3.14, 95% CI 1.34-7.36, p = 0.008, I2 = 0%). Likewise, subgroup analysis of pooled data from 7 RCTs in pregnant women (n = 1028) showed a statistically significant increased risk of GI side-effects for ferrous sulfate although there was marked heterogeneity in the data (OR = 3.33, 95% CI 1.19-9.28, p = 0.02, I2 = 66.1%). Meta-regression did not provide significant evidence of an association between the study OR and the iron dose. Conclusions Our meta-analysis confirms that ferrous sulfate is associated with a significant increase in gastrointestinal-specific side-effects but does not find a relationship with dose.


Introduction
In the UK, iron deficiency anemia (IDA) affects around 4.7 million people every year [1,2]. Groups mostly at risk include those with increased iron demands, for example children and pregnant women [3,4,5], and especially those with increased iron loses, for example pre-menopausal women [6] and patients with inflammatory bowel disease (IBD) [7,8].
First-line treatment is oral therapy with ferrous iron (Fe(II)) salts [9]. For example, in 2012 more than 6.8 million prescriptions were filled for oral iron in England and 97.6% of them were for simple Fe(II) salts [10]. Gastrointestinal side-effects are the most commonly reported adverse effects associated with oral iron treatment and include nausea, flatulence, abdominal pain, diarrhoea, constipation, and black or tarry stools [11,12,13,14,15]. For very many years patient side-effects have been the main concern with oral iron therapy [16] but, recently, studies have consistently shown that soluble oral iron also negatively impacts the colonic microbiota, promoting the presence of potentially pathogenic bacteria at the expense of beneficial bacteria [17,18,19]. Most recently there have been concerns over 'available' iron in the colon as a risk factor for inflammatory signalling and colorectal carcinogenesis [18,20]. Nonetheless, irrespective of the mechanistic foundation of the GI side-effects that appear to be related to oral iron therapy, the impact results in non-adherence in up to 50% of patients. This leads to significant treatment failures and unnecessary follow-up investigations [11,21,22,23,24,25,26,27,28,29,30].
In trying to capture the true incidence of GI side-effects with oral iron therapy, discrepancies between studies and limitations of study design have made overall conclusions difficult to reach [11,14,15]. A recent review has tried to capture data for adverse events from studies with oral iron supplementation in patients without gastrointestinal disease and reported overall adverse event incidences of 32.3% for ferrous sulfate, 47% for ferrous fumarate and 30.9% for ferrous gluconate [11]. However, the meta-analysis reported herein has refined the study eligibility criteria in relation to the prior review [11]. Notably, we have only included randomized controlled studies with a ferrous sulfate intervention arm; we restricted adverse events to the common focus of concern with therapeutic oral iron, namely GI side-effects, and we have only included studies with a common comparator arm, which was either placebo or intravenous iron.
Since ferrous sulfate is the gold standard (most commonly prescribed) oral iron therapy in the UK and many other countries [10,31], our aim was to quantify the odds ratio for oral ferrous sulfate-associated gastrointestinal side-effects versus placebo or IV iron. To this end, we have carried out a systematic review and meta-analysis of all published randomized controlled trials (RCTs) reporting gastrointestinal-specific side-effects that have included ferrous sulfate against placebo or IV iron. We have performed sub-group analysis for pregnant women and patients with IBD since these are two population groups at higher risk of iron deficiency anemia and for whom sufficient robust data were likely to be available [14,32,33]. Finally, we explored whether the iron dose is associated with the odds of gastrointestinal side-effects using meta-regression analysis.

Search strategy
The following bibliographic online databases were searched ( Publications reporting randomized controlled studies with ferrous sulfate were identified by using the search terms ferrous, ferrous sulphate, ferrous sulfate, iron supplement, ferrous salt, iron, paired with the terms randomized controlled trial, randomized controlled trial, controlled trial, controlled clinical study. No limits on language or publication type were imposed. Please refer to S1 File in the online issue for the full search strategy. Additionally, the following Institutions were contacted (in January 2012) via email to request any additional studies not identified through the electronic search: WHO department of Nutrition for Health and Development (NHD) as well as 12 WHO

Eligibility criteria
Studies were selected for analysis if they met the following criteria: (1) the study was an RCT conducted in adult human participants with either a parallel or crossover design, (2) the study included an arm/group with ferrous sulfate, (3) the study included a comparator arm/group with either placebo or intravenous iron and (4) gastrointestinal side-effects were recorded separately for each arm (i.e. ferrous sulfate and comparator). Where co-interventions were administered with the ferrous sulfate arm (for example ascorbic acid or folic acid) the studies were only included if the same co-intervention was administered in the comparator arm and when in the authors' view the co-interventions should not influence side-effects in either arm.

Selection and data extraction
Titles of all records retrieved were initially screened by ZT. All potentially relevant studies were then further screened as abstracts by three independent assessors. Three authors (ZT, LS and DP) independently screened full texts of all potentially relevant papers to ensure they met the inclusion criteria. Data were extracted independently by two authors (ZT and LS) and discrepancies were resolved before data were compiled into a database that included the following fields: (1) patient population, (2) inclusion age criteria, (3) mean age, (4) participant numbers, (5) commercial name for ferrous sulfate (if any), (6) ferrous sulfate dose, (7) frequency of dose, (8) equivalent iron content, (9) duration of the intervention, (10) hemoglobin changes from baseline to end of the intervention, (11) gastrointestinal side-effects reported, (12) comparator group (placebo or intravenous iron), (13) co-interventions. In most studies, patient numbers used for the meta-analysis were those that were randomized, with the exception of studies that specified a subset safety population (i.e. if it was stated that there was only a subset of patients where adverse effects were assessed). If side-effects were reported multiple times during the trial, then only measures at the end of the intervention period were included in the meta-analysis. Data were extracted at the participant level, i.e the number and/or percentage of participants with at least one GI side-effect within a study arm. The 'total number of participants' was taken to be the number within the safety (i.e. side-effect assessment) population when specified, otherwise it was taken as the number of participants randomized. Some studies reported the number of participants experiencing each type of GI side-effect and not overall. In these cases, the commonest GI symptom was deemed (conservatively) to represent the number of participants that experienced GI side-effects. For example, in the study of Mirrezaie et al [34], 5 participants in the ferrous sulfate arm reported heartburn, 17 reported nausea, 2 reported abdominal cramps and 1 reported constipation, so the number of participants in the ferrous sulfate arm with a GI side-effect was taken to be 17.
The generated database for all of the above is available upon request.

Statistical analysis
Studies that met the eligibility criteria were included in the meta-analysis. The random-effects model proposed by DerSimonian and Laird [35] was used for the meta-analysis. The odds in each arm were calculated as p/(1-p) where p is the proportion with GI side-effects. The odds ratio (OR) was calculated as the odds in the ferrous sulfate arm divided by the odds in the comparator arm. The meta-analyses used study specific log odds ratios as the outcome and the resulting pooled estimates were converted to OR. Values for the OR above 1 indicate that ferrous sulfate is associated with more side-effects in comparison to either placebo or intravenous iron. Risk of bias in individual studies was assessed independently by two co-authors based upon the Cochrane Collaboration's Tool [36]. Furthermore, we assessed risk of bias connected to how the studies obtained information about gastrointestinal adverse-events (questionnaire, face to face interviews, spontaneous reporting by participants etc.). Publication bias was assessed by 'funnel plots' of the effect [log (OR)] against its standard error. The symmetry of the plots was assessed visually to look for asymmetry. Heterogeneity was assessed using the I 2 statistic and substantial heterogeneity exists when I 2 >50% [37]. Subgroup analyses were carried out for two pre-defined subgroups, namely pregnant women and patients with inflammatory bowel disease. When zero GI side-effects were reported for the placebo or IV comparator arm a standard correction of adding 0.5 of a side-effect to each arm was used to enable calculation of the study-specific log OR. However, since this was a relatively common occurrence in the studies with an IV iron comparator, the meta-analysis was repeated excluding those studies where less than one gastrointestinal side-effect was reported in the IV iron arm. We produced forest plots to visualise the ORs and 95% CIs of each study, each subgroup and all studies combined. For studies with a cross-over design both cross-over periods were treated independently in the analysis because there was insufficient information regarding individual tabulation of side-effects to apply the methodology proposed by Elbourne et al. [38].
Furthermore, meta-regression analyses were conducted to determine if the odds of gastrointestinal side-effects with ferrous sulfate was greater with higher doses. We produced bubble plots where individual studies are represented by circles, with the size of the circle being inversely proportional to the variance of the estimated odds ratio used in the meta-regression (i.e. the larger the circle, the more precise is the estimated OR). For the most commonly reported gastrointestinal side effects in the oral ferrous sulfate group, a random-effects metaanalysis (using log-transformed odds) was performed. This provided a pooled estimate of the incidence of each symptom in participants taking oral ferrous sulfate.

Study characteristics
A total of 44 studies with publication dates ranging from 1966 to 2013 met the inclusion criteria ( Fig. 1). One study was excluded from the final meta-analysis because it was conducted in critically ill surgical patients in intensive care so ferrous sulfate was mixed in the enteral feed rather than being delivered in pure supplemental form as in the other studies [39]. Tables 1 and 2 summarise the characteristics of the 43 studies (n = 6831 participants overall) included in the metaanalyses. Twenty studies (n = 3168 participants) evaluated side-effects associated with oral ferrous sulfate (20-222 mg Fe/day) against a placebo comparator (  (Table 2). We conducted 2 independent meta-analysis, one for the placebo-controlled trials and one for the IV iron-controlled trials, due to clear differences in trial design and study populations, namely: (i) the placebo-controlled trials were generally doubleblind with healthy non-anaemic participants whilst (ii) the IV iron-controlled trials were openlabel with moderate-severely anaemic patients. The result of the assessment of risk of bias for the individual studies is presented in Table A in S1 File. Included is also the assessment of variability in the methodology used to collate information for gastrointestinal-related adverse-effects. Publication bias was investigated with 'funnel plots', see Figure A in S1 File. The outliers in the funnel plots corresponded to studies reporting zero GI side-effects either in the IV ironarm (9 studies) or placebo-arm (2 studies) (refer to Tables 1 and 2) and to 1 small (n = 10) placebo-comparator study [40]. The pattern of the remaining studies, although not following the expected funnel shape, is symmetrical and therefore does not provide evidence of publication bias.

Ferrous sulfate versus placebo
For the 20 trials that included placebo as the comparator a significant increase in the incidence of gastrointestinal side-effects was observed with ferrous sulfate OR = 2.32 [95% CI 1.74-3.08, p<0.0001, I 2 = 53.6%] (Fig. 2). Nineteen of these 20 placebo-controlled trials were conducted in healthy non-anaemic individuals and therefore hemoglobin repletion was not a primary outcome. The only placebo-controlled trial in anaemic participants was carried out in patients who became anaemic following joint replacement surgery and reported an average increase in hemoglobin of 1.94 g/dl (range -0.3 to +4.2) following 6 weeks of ferrous sulfate therapy compared to an increase of 1.63 g/dl (range -1 to + 3.6) with placebo [41].

Ferrous sulfate versus IV iron
For the 23 studies that included intravenous iron as the comparator a significant increase in the incidence of gastrointestinal side-effects was observed with ferrous sulfate with an OR = 3.05 [95% CI 2.07-4.48, p<0.0001, I 2 = 41.6%] (Fig. 3A). Furthermore, mean hemoglobin increase was reported for 20 of the 23 eligible IV iron-controlled trials ( Fig. 3B and Table B in S1 File). Overall, for these 20 trials, the mean increase in hemoglobin for the ferrous sulfate arm was lower than for the IV iron arm although formal comparative analysis was not undertaken being outside the a priori objectives of our analysis (Fig. 3B).
Nine out of the 23 IV-iron comparator studies reported zero GI side-effects in the IV-iron arm so, as detailed in Methods, we added the standard correction of 0.5 of a side-effect to each arm to enable calculation of the study-specific log odds ratio. Nevertheless, when we excluded these 9 studies from the meta-analysis the same overall effect was observed with an OR = 2.41 (95% CI 1.66-3.50, p<0.0001, I 2 = 42.1%) for ferrous sulfate (seeFigure B in S1 File).

Ferrous sulfate in pregnancy: placebo-and IV iron-controlled trials combined
The subgroup analysis of RCTs in pregnancy generated contrasting results for the placebo-controlled and the IV iron-controlled trials although the number of trials were small (n = 2 and n = 5 respectively, Figs. 2 & 3). Combining data from the 7 RCTs (n = 1028 participants)

Dose-response
Meta

Odds of Gastrointestinal Side-Effects of FeSO4
Discussion WHO considers IDA as one of the most expensive diseases in the world due to lost productivity and the sheer numbers of the population affected (ca. 1 billion) [42]. First-line treatment is oral therapy with ferrous iron salts, but a substantial proportion of patients suffer from gastrointestinal side-effects, resulting in non-adherence and treatment failure [22,23,24,27,28,30,43]. Gastrointestinal symptoms most probably result from a combination of two factors: (i) free radical generation through iron-induced redox cycling in the gut lumen and at the mucosal surface which can promote inflammation [18,44,45,46] and (ii) changes to the microbiota composition or metabolism [17,18,19,47]. Ferrous sulfate remains the most commonly prescribed oral iron therapeutic [10]. We report the first meta-analysis of randomized controlled trials investigating the gastrointestinal side-effects associated with ferrous sulfate supplementation. In our analysis, IV iron was used purely as a comparator arm for gastrointestinal side-effects associated with oral iron for the trials that did not include a placebo arm. The discussion of the side-effects, other than gastrointestinal, associated with IV iron was, therefore, beyond the scope of this analysis.
Significant heterogeneity was observed in both the placebo-controlled studies analysis (I 2 = 53.6%, p = 0.002) and the IV iron-controlled studies analysis (I 2 = 41.6%, p = 0.02) mainly due to differences in patient population and methodology used to collate gastrointestinal adverseeffects. We have accounted for this heterogeneity by using random-effects modelling [35]. Even though we did not anticipate publication bias to impact the analyses presented here due to the fact that GI adverse-effects was not the primary outcome in most of the studies included in the meta-analysis, we have produced 'funnel plots' for visualisation of any asymmetry due to non-publication of studies with negative effects (Figure A in S1 File). The perceived outliers in Forest plot for the effect of daily ferrous sulfate supplementation on the incidence of gastrointestinal side-effects in pregnant women. Data for random-effects subgroup meta-analysis are shown (7RCTs, n = 1028). For each study the closed diamond represents the mean estimated effect and the horizontal lines the 95% CI. The grey shaded area surrounding each closed diamond represents the weight of each study in the analysis. Weight was assigned based on (inverse of) the sum of the within-study variance and between study variance. Open diamonds represent the subgroup mean difference and pooled overall mean differences as shown. Test for overall effect: z-score = 2.29; p = 0.02. OR, odds ratio; CI, confidence interval. doi:10.1371/journal.pone.0117383.g004 Odds of Gastrointestinal Side-Effects of FeSO4 Table 3. Individual side-effects reported in the FeSO 4 group/arm for the studies where this information was available.   these plots correspond to studies with zero events in the comparator arms and to one small study with only 10 subjects. The pattern of the remaining studies in the 'funnel plots' did not reveal asymmetry and, therefore, there is no indication of publication bias. Due to the heterogeneity of the study populations in both trial types we have kept the analyses separate overall but there was no change in our findings if the data were combined (see Figure C in S1 File). In order to minimize side-effect reporting bias, we have only included controlled trials with a comparator arm that included an intervention, either placebo or intravenous iron, since gastrointestinal symptoms are common even in the general population and are known to be influenced by psychological factors [48,49,50,51]. Both placebo-controlled and IV iron-controlled meta-analysis have shown that ferrous sulfate is associated with increased gastrointestinal side-effects.
Overall, daily ferrous sulfate supplementation was associated with 2.6 times the odds of GI side-effects compared to placebo or IV-iron in participants who were not pregnant and did not have IBD (Figs. 2 & 3). In pregnancy, adherence with ferrous sulfate has been reported as only 70-90% due to adverse-effects [27,28,43]. Two recent Cochrane reviews provide the most Odds of Gastrointestinal Side-Effects of FeSO4 comprehensive analysis of oral iron therapy in pregnancy [14,15] and draw on data from 11 trials (n = 4418 participants) reporting side-effects. Pena-Rosas et al. have shown that pregnant women receiving oral iron supplements were, overall, more likely than controls (albeit not quite statistically significantly so) to report side effects (25.3% versus 9.91%: RR 2.36; 95% CI 0.96-5.82), particularly at doses 60 mg of elemental iron [14]. Pregnant women receiving intermittent oral iron supplementation had less side effects (mean RR 0.56; 95% CI 0.37-0.84) than those receiving daily iron supplements [15]. Herein we only considered the 7 eligible trials in pregnant women (n = 1028 participants) that have used oral ferrous sulfate and our findings also show that ferrous sulfate is associated with a significant increase in the incidence of gastrointestinal side-effects (OR = 3.33, p = 0.02, Fig. 4), but there is marked heterogeneity in the data.
It is well recognized that inflammatory conditions of the GI tract significantly reduce adherence with oral iron in comparison to the general population, with reports of 52% IBD patients reducing dose or withdrawing from ferrous sulfate treatment due to poor tolerability [29]. Indeed, the four eligible RCTs in IBD patients showed that in this population ferrous sulfate is associated with a significant increase in the incidence of GI side-effects (OR = 3.14, p = 0.008, Fig. 3) in comparison to intravenous iron. However, in the IV iron-controlled studies all subjects were considerably ill with moderate-severe anemia and/or other underlying conditions and, therefore, the comparison with an 'otherwise healthy' population group was not possible.
Even following considerable efforts through local and national libraries we could not obtain the full text for 88 studies (Table C in S1 File). The vast majority of these papers were old and not available in English translation, and for many there was no abstract available. Nonetheless, based on a rate of inclusion of 2% from full-text (Fig. 1), we estimate that only 1-2 studies would have been eligible for inclusion from the 88 references that were not obtained. Even though this remains a minor weakness of the present meta-analysis, it would be surprising if data from these few anticipated, additional eligible trials would have changed our findings.
Even though we strived to include in this meta-analysis only robustly designed trials, one limitation remains in relation to blinding of the treatments. First, IV versus oral iron trials are difficult to blind due to the nature of the different interventions. None of the IV iron-controlled studies were blinded and, therefore, we have judged this to represent a high risk of bias in relation to reporting of the main outcome of the present analysis (i.e. GI adverse-effects) ( Table A in S1 File). However, we do acknowledge that for the vast majority of the studies the primary outcome measures were biochemical parameters (e.g. hemoglobin) and this is unlikely to have been impacted to the same extent by the lack of blinding.
Secondly, the placebo-controlled trials were not truly blind studies as blackened stools are commonly reported with oral iron [52,53] and in none of the trials included in our meta-analysis was use of a 'stool darkener' reported for the placebo arm. Therefore, patients would mostly have been aware when they were taking oral iron and perception of gastrointestinal symptoms could have been altered in a manner that even robust meta-analysis cannot account for. To support the double-blind design in future placebo-controlled clinical trials investigating gastrointestinal symptoms with oral iron it would be preferable to use a stool darkener It is generally considered that (i) doses 50-60 mg iron/day generate less side-effects than higher doses and that (ii) iron given in controlled release formulations is better tolerated [27,54,55]. However, our meta-regression analysis shows that there is no statistically significant dose-response effect or threshold whether considered as amount of iron per day (Fig. 5, Figure D-A in S1 File) or per dose ( Figure D-B in S1 File). A limitation of this analysis, however, is that data are not uniformly spread across iron dosage, particularly for the intravenous iron-controlled trials that mostly used~200mg Fe/day (see Fig. 5B). Nonetheless, if there is a benefit in terms of lower side-effects with lower doses of ferrous sulfate then the threshold appears very low (i.e. 20 mg iron per dose or per day, Figure D in S1 File).
A recent review of 111 studies (10695 participants) with different oral iron preparations has suggested that slow-release ferrous sulfate is better tolerated than regular gastric release ferrous iron salts [11]. This review has included all types of studies ranging from observational to RCT. In contrast, we restricted the work here to meta-analysis of RCTs with robust trial design which still included data from 6831 participants. In this case, subgroup analysis of the 6 included studies that used modified-release ferrous sulfate [56,57,58,59,60,61] in fact shows an OR = 3.60 (95% CI 1.32-9.87, p = 0.01, I 2 = 46.2%) compared to OR = 2.53 (95% CI 1.99-3.21, p<0.0001, I 2 = 49.6%) for the 37 studies that have used conventional delivery of oral ferrous sulfate (see Figure C in S1 File). Four of the studies with modified-release ferrous sulfate [56,57,58,61] reported zero GI side-effects in the IV iron arm so these data should be interpreted with caution. Nonetheless, our findings do not support the idea that modified-release oral ferrous sulfate markedly modifies its side-effects.

Conclusions
In summary, our analyses show that: (i) ferrous sulfate causes significant gastrointestinal sideeffects in adults in all the population groups investigated, with the caveat of potential biases associated with study blinding that are inherent to interventions with oral iron, as discussed above; (ii) the OR of side-effects in IBD is higher than in non-IBD and non-pregnant participants but overall numbers were small and significance not established; (iii) the pregnancy subgroup analysis revealed considerable heterogeneity; (iv) there is no evidence for dose effect; (v) there is no evidence that modified-release ferrous sulfate causes less side-effects than conventional gastric release ferrous sulfate.
Supporting Information S1 PRISMA Checklist. PRISMA checklist. (DOCX) S1 File. Figure A, Funnel plots of effect of daily ferrous sulfate supplementation on the incidence of gastrointestinal side-effects against standard error. Figure B, Forest plot for the effect of daily ferrous sulfate supplementation on the incidence of gastrointestinal side-effects in IV iron-controlled RCTs. Figure C, Forest plot for the effect of daily ferrous sulfate supplementation on the incidence of gastrointestinal side-effects. Figure D, Meta-regression analysis of the association between iron dosage and the odds ratio of gastrointestinal side-effects. Methods A. Table A, Assessment of 'risk of bias' according to the Cochrane Collaboration's tool. Table B, Mean hemoglobin increase (g/dl) reported in IV iron-controlled RCTs (n = 3267). Table C