This meta-analysis examined 71 randomized controlled trials that included NaP or PEG solution or both for bowel preparation prior to elective, outpatient colonoscopy. The findings indicate that NaP resulted in an excellent quality prep more often than PEG. Further, based on minimal or no overlap of the 95% CIs, NaP tablets resulted in a greater likelihood of achieving an excellent quality prep than did all PEG groups, while NaP solution was in between. There was no difference in prep quality among the various PEG subgroups. Among treatments arms where prep quality could be quantified as a composite of excellent or good, NaP tablets (87.8%) were numerically superior to all other forms of either prep. There was minimal overlap of the 95% CI of NaP tablets with those of both NaP solution and 4 L PEG arms, both of which included adjunctive medications. Despite an absolute difference of just over 21% between NaP tablets and split dose 4 L PEG that favored NaP tablets, the CIs showed a large degree of overlap, most likely due to the imprecision of the individual trial point estimate for split dose PEG. Finally, among the trials that included completion rates, NaP was more likely to be completed than PEG, with the exception of the split dose PEG regimen.
Previous meta-analyses of head-to-head trials of PEG vs. NaP have reported that sodium phosphate is more effective, better tolerated, and less costly than PEG [5, 18]. However, in 2007 a meta-analysis by Belsey et al reported that no single bowel preparation was consistently superior to others . Both meta-analyses excluded data from trials comparing either or both of these preps to other, less commonly used preps.
This analysis has strengths and limitations. One strength is the breadth of our search strategy and analysis, as we included studies that have been excluded in other systematic reviews. Another strength is the clinical homogeneity of the patient population studied: all groups were comprised of outpatients undergoing elective colonoscopy. Further, as best we could determine, the study populations, even after re-assembly by treatment arm, appear to be demographically and clinically comparable. Finally, the large sample size of this analysis provides reasonable precision for most of the point estimates of effect for both efficacy and tolerability.
With regard to limitations, the potential for clinical heterogeneity is always present when combining trials, particularly for factors that were not measured. The possibility of clinical heterogeneity appears to be low; to minimize its effects, we used a random effects model, which accounts for heterogeneity by both providing a point estimate that is less weighted by the studies with larger sample sizes and resulting in wider confidence intervals. Several candidate factors may contribute to clinical and/or methodological heterogeneity among trials. One factor is variation in timing of bowel prep. The time at which the bowel prep was started was not uniform among the trials ranging from 2:00 PM in the afternoon to 7:00 PM in the evening the day before the scheduled procedure. This may have affected those patients undergoing colonoscopy in the afternoon by affecting prep quality particularly in the right colon, where intestinal chyme can accumulate, obscuring the mucosa. Another factor potentially contributing to heterogeneity is the variation in dietary instructions prior to and during the prep, which also were not uniform among the trials, and which ranged from a regular diet to a clear liquid diet for lunch and clear liquid diet in the evening.
A second limitation is the uncertain acceptability of the "treatment-arm" method of doing meta-analysis. While this method has been used previously for comparing treatments for rheumatoid arthritis , for prevention of deep venous thrombosis following total hip replacement  and for treatment of premature labor , its validity is less well established than is head-to-head meta-analysis where comparators are the same in all studies. Limiting the analysis to a head-to-head comparison would not have allowed consideration of evidence from trials where either NaP or PEG was compared to another bowel preparation. An alternative to our "treatment-arm" approach is a mixed treatment comparison or "network" meta-analysis, which is another way of quantitatively aggregating data across studies containing disparate comparators [86, 87]. It allows comparison of multiple treatments, combining direct and indirect evidence in a single analysis. While head-to-head meta-analysis and network meta-analysis of the same data have been compared , there are no comparative analyses between network and treatment-arm meta-analyses. In the absence of such comparative data, it remains uncertain which method is most appropriate for synthesizing quantitative data, and under which circumstances the two methods differ in results.
A third potential factor is the variation in definitions of patient tolerance of the prep. While some trials defined patient tolerance by different parameters (e.g. completion rates, willingness to repeat the prep, palatability and adverse affects), others defined patient tolerance as a single parameter and reported it as a single cumulative estimate.
In recent years, three reports have described 22 patients who developed renal insufficiency due to nephrocalcinosis that was temporally associated with use of NaP for colonoscopy prep, 4 of whom progressed to end stage renal disease requiring dialysis [89–91]. The majority of these patients had co-morbid conditions such as diabetes mellitus, hypertension (treated with angiotensin-converting enzyme inhibitors [ACE-I] or angiotensin receptor blockers [ARBs] or diuretics), preexisting renal insufficiency, were elderly, or had small bowel disease that resulted in calcium and vitamin D malabsorption. Renal biopsies of many of the cases showed nephrocalcinosis with intra-tubular deposition of calcium-phosphate. The term for this pathologic condition is acute phosphate nephropathy (APN). The histopathology suggests that sodium phosphate ingestion leads to obstructive calcium-phosphate crystalluria followed by acute intra-tubular nephrocalcinosis. These reports have recently raised concerns that led Food and Drug Administration  to announce a safety alert in December 2008, stating that a boxed warning would be added to the labeling on prescription OSPs (Visicol and OsmoPrep) and recommending against use of over-the-counter OSPs for bowel preparation. Shortly after this announcement, all over-the-counter NaP products were voluntarily removed from the market, with a subsequent sharp decline in use of NaP solution.
Despite the FDA's action and resulting reaction, the published data suggest that absolute risk of APN is very low [93, 94]. Further, a recent systematic review and meta-analysis of seven controlled studies (patient N = 14,520) of the effects of NaP versus comparator on kidney function showed that there is significant clinical heterogeneity in the populations studied, study methods, definition of kidney injury, and results . Quantitatively, the pooled odds ratio for kidney injury among NaP-treated patients ranged from 1.08 (CI, 0.71-1.62) to 1.22 (CI, 0.77-1.92). The investigators concluded that it was not possible to discern whether there is a true association between NaP and kidney injury.
The results of this meta-analysis apply to patients undergoing elective colonoscopy who do not have a history of co-morbid conditions such as renal insufficiency, recent myocardial infarction, ascites due to cirrhosis, and congestive heart failure. Further, NaP should not be used by patients with suspected or established inflammatory bowel diseases because of aphthous ulcerations it may cause . While NaP solution is not currently available, tablet forms of NaP remain available by prescription only. Physicians should be aware of the risk of acute kidney injury with NaP preparations and should not use it in older patients, in those with preexisting renal insufficiency, and in patients on medications that can affect volume status or renal function (diuretics, ACE inhibitors, and ARBs). Further, all patients should be encouraged to adequately hydrate themselves prior to and while using NaP preparations.