Investigation of the association between oral sodium phosphate use and kidney injury.

Sir,

Recently we published a report in which we found no apparent association between oral sodium phosphate (OSP) purgative use and incident chronic kidney disease (CKD): OR (95% CI) 0.70 (0.44–1.11) [1]. Published concurrently with our study was another study by Hurst [2] that reported a potent association between OSP use and acute kidney injury (AKI): OR (95% CI) 2.35 (1.51–3.66) [2]. We were intrigued at the dramatic differences in findings, and undertook additional sensitivity analyses of our data in order to explore possible explanations. Potential sources for discrepancy include differences in timing of post-colonoscopy creatinine considered, case definition and population under study.

In order to determine the impact of case definition on findings, we conducted analysis 1 in which case status was reassigned according to Hurst's criterion (50% rise in serum creatinine [2]). The adjusted OR (95% CI) was 0.75 (0.32–1.73) (Table 1).

Table 1

Results of analyses 1, 2 and 3

	Analysis 1 (n = 465)	Analysis 2 (n = 465)		Analysis 3c (n = 281)
Case definition (% rise in SCr baseline to follow-up)	50%	25%	50%	25%	50%
Baseline SCr	Most recent pre-colonoscopy
Follow-up SCr	Latest within 6 months post-colonoscopy	Earliest post-colonoscopy		Earliest post-colonoscopy
# Cases/# OSP exposed	26/14	90/56	19/11	50/32	10/6
# Controls/# OSP exposed	439/282	375/240	446/285	231/160	271/186
Unadjusted OR (95% CI)	0.65 (0.27–1.58)	0.93 (0.56– 1.54)	0.78 (0.28– 2.27)	0.79 (0.40– 1.60)	0.69 (0.16–3.40)
Adjusted OR (95% CI)a	ND	0.87 (0.52– 1.46)	ND	0.76 (0.37– 1.56)	ND
Adjusted OR (95% CI) using propensity score analysisb	0.75 (0.32–1.73)	0.89 (0.54– 1.46)	1.15 (0.43– 3.07)	0.78 (0.39– 1.54)	0.96 (0.25– 3.79)

aAdjustment was made for age, race (white, non-white and unknown), gender, clinical site, diabetes, congestive heart failure, ACEi/ARB use, diuretic use and colonoscopy indication (screening/surveillance versus symptomatic) using multivariable logistic regression model.

bAdjustment was made for above using propensity score analysis.

cAnalysis 3 was restricted to subjects over the age of 50, undergoing colonoscopy for routine screening or surveillance, and excluding those who received purgatives other than OSP or polyethylene glycol. (No adjustment was made for colonoscopy indication.)

SCr: serum creatinine; ND: not done.

In order to examine the association between OSP and AKI, Hurst defined outcome based on the most proximate post-colonoscopy serum creatinine [2]. Thus, we conducted analysis 2 in which case status was defined according to the change between baseline and earliest available post-colonoscopy creatinine. Adjusted ORs (95% CIs) were 0.87 (0.52–1.46) and 1.15 (0.43–3.07) using 25% and 50% serum creatinine rise as case definition, respectively (Table 1).

Hurst limited inclusion to subjects over the age of 50 undergoing colonoscopy for screening purposes, and considered as controls only subjects receiving polyethylene glycol [2]. Thus, we conducted analysis 3 in which we restricted observation to those subjects meeting Hurst's criteria (n = 281). The adjusted ORs (95% CIs) were 0.76 (0.37 to 1.56) and 0.96 (0.25–3.79) based on case definitions of 25% and 50% rise in serum creatinine, respectively.

Multiple case reports and series suggest an association between OSP use and ‘acute phosphate nephropathy.’ The issue here is whether these represent rare, idiosyncratic reactions or a more generalized risk to the population at large. The only prior controlled study of the renal consequences of OSP exposure suggests that there is not a predisposition toward systematic CKD [3], consistent with our present (and past) findings.

These analyses suggest that timing of post-colonoscopy creatinine considered, case definition and certain characteristics of the subjects studied do not account for the dramatic differences between Hurst's findings and our own. Our study included a much higher proportion of patients who were non-white, had diabetes or congestive heart failure; it is possible that differences result from as yet unrecognized effect modification on these bases.

Some limitations of the present analyses bear note. In reassigning case/control status, there was diminution of the number of available cases and statistical power; thus, we cannot exclude the possibility of type II error. As AKI was not our original outcome of interest, we lack data on certain exposures that potentially confound the association between OSP and AKI (e.g. iodinated contrast exposure and non-steroidal anti-inflammatory use), and, therefore, we could not adjust on these bases.