Association of serum biochemical metabolic panel with stone composition

Transcription

1 bs_bs_banner International Journal of Urology (2015) 22, doi: /iju Original Article: Clinical Investigation Association of serum biochemical metabolic panel with stone composition Daniel M Moreira, 1 Justin I Friedlander, 1,2 Akinwunmi Carons, 1 Christopher Hartman, 1 David A Leavitt, 1 Arthur D Smith 1 and Zeph Okeke 1 1 The Arthur Smith Institute for Urology, Hosftra North Shore-LIJ School of Medicine, New Hyde Park, New York, and 2 Department of Urology, Fox Chase/Einstein Urologic Institute and Einstein Healthcare Network, Philadelphia, Pennsylvania, USA Abbreviations & Acronyms BMI = body mass index BUN = blood urea nitrogen CO 2 = carbon dioxide DM = diabetes mellitus ESWL = extracorporeal shock wave lithotripsy HTN = hypertension SS = supersaturation Correspondence: Daniel M Moreira M.D., The Arthur Smith Institute for Urology, North Shore-Long Island Jewish Health System, 450 Lakeville Road, New Hyde Park, NY 11042, USA. dam@doctor.com Received 8 April 2014; accepted 25 August Online publication 25 September 2014 Introduction: To determine the association of the basic metabolic panel with stone type. Methods: The present study was a retrospective review of 492 stone formers with both stone composition analysis and basic metabolic panel available. Analysis of a basic metabolic panel across stone types was carried out using Fisher s exact test and analysis of variance. Multinomial logistic regression was used to predict stone type based on a basic metabolic panel. Results: A total of 272 (55%) patients had predominantly calcium oxalate stones, 100 (21%) had uric acid stones, 93 (19%) had calcium phosphate stones, 16 (3%) had mixed stones and 11 (2%) had other types of stones. Uric acid stone formers had the highest serum glucose, blood urea nitrogen and creatinine levels. Calcium oxalate stone formers had the highest serum sodium. No significant differences in mean serum calcium levels across different stone types were identified. The predicted risk of uric acid stone over the other stone types increased with an increase in serum glucose and decreased with an increase in carbon dioxide levels. The predicted risk of calcium oxalate stones increased with an increase in serum sodium and chloride levels. The predicted risk of calcium phosphate and oxalate stones over the other stone types increased with an increase in serum calcium levels. The overall accuracy of the basic metabolic panel alone to predict stone type was 59%. Conclusion: A basic metabolic panel alone or in combination with 24-h urinalysis and demographics does not accurately predict stone type. However, it can be used in combination with other variables to predict stone composition. Key words: calcium oxalate, calcium phosphates, kidney calculi, uric acid, urinalysis. Introduction The knowledge regarding stone composition is important in the management of urolithiasis, being useful in defining prognosis and directing medical and surgical management. For example, calcium oxalate and phosphate stones treated with ESWL have higher recurrence rates when compared with struvite and uric acid stones. 1 Furthermore, calcium phosphate stones, especially brushite, are particularly resistant to ESWL 2 and are possibly associated with lower stone-free rates after percutaneous nephrolithotomy. 3 Additionally, certain types of stones are associated with specific metabolic abnormalities. For instance, calcium phosphate stones are associated with primary hyperparathyroidism, whereas uric acid stones are correlated with hyperuricosuria. 4 In these patients, the treatment of the primary metabolic abnormality is essential to prevent further stone formation. Therefore, stone composition should be taken into consideration when deciding how to treat patients with urinary calculi. However, in some circumstances, the stone composition is not available, such as in cases where the stones were not sent for analysis or in patients with stones in the urinary system that have not yet been removed or eliminated. Previously, we have shown that 24-h urinalysis can be used to predict stone type in these cases. 5 However, the accuracy of 24-h urinalysis alone was just 64%. In addition to 24-h urinalysis, patients with urinary stones are routinely evaluated with serum metabolic panels. Some patients have metabolic abnormalities that are associated with the formation of specific stone types, such as hypercalcemia and calcium stones. Thus, it is plausible that metabolic abnormalities detected in the blood could be used to determine the type of stone a patient forms. Yet, it is unknown whether the serum metabolic profile, such as abnormal electrolytes, creatinine and glucose, contributes to an accurate prediction of the stone type. 195

2 DM MOREIRA ET AL. Therefore, we sought to identify the associations between basic metabolic panel and stone type. We also sought to determine the accuracy of combining the results of basic metabolic panel, 24-h urinalysis and patient demographics to predict stone type among patients from a large retrospective cohort. Methods Study population After obtaining institutional review board approval, data from adult patients (age 18 years or older) undergoing 24-h urine analysis between March 2002 and February 2012 at the Arthur Smith Institute for Urology in New Hyde Park, NY, USA, were combined into the study database. The database includes information on patient age, sex, height, weight, comorbidities (including DM, HTN and hypercholesterolemia), medication use (including potassium citrate, allopurinol and thiazides) and stone composition (including calcium oxalate monohydrate and dihydrate, uric acid, brushite, ammonium urate, struvite, calcium carbonate, apatite and cystine). The database also contains 24-h urine composition (including urinary ph, volume, calcium, oxalate, citrate, uric acid, sodium, potassium, magnesium, phosphorus, sulfate, creatinine, SS calcium oxalate, SS calcium phosphate and SS uric acid concentrations) and basic metabolic panel (including serum sodium, potassium, chloride, carbon dioxide, BUN, creatinine, glucose and calcium concentrations). Only patients with adequate 24-h urine collections (defined by 24-h creatinine excretion equal to or greater than 800 mg for men and 600 mg for women) were included. Patients with rare stone types, such as cystine, xanthine and medication-associated (including indinavir, sulfadiazine and triamterene) stones, were excluded. Of 1049 patients with complete 24-h urinalysis and basic metabolic panel, 497 (47%) had stone composition available and were included in the study. Five patients (1%) with a history of urinary diversion were excluded, resulting in a final study sample of 492 patients. Stone composition analyses were carried out at Mayo Medical Laboratories (Rochester, MN, USA) using infrared spectroscopy. Representative specimens were taken from all identifiable layers of the calculus. Each specimen was weighed and crushed into a fine powder. An infrared spectrum of each specimen was recorded, and the resulting spectrum compared against reference spectra of all known calculus components. 24-h urinalyses were carried out at Litholink Laboratories (Chicago, IL, USA). Patients followed in our stone clinic are routinely recommended to increase fluids, and decrease sodium and protein intake when appropriate, thus serum and urinary measurements might reflect this practice. Statistical analysis Patients were divided into five groups based on the predominant element in stone composition: calcium oxalate (including mono- and dihydrate), uric acid, calcium phosphate (including brushite and apatite), others (including ammonium urate, struvite and calcium carbonate) and mixed stones (when none of the elements in stone composition were equal or greater than ). Mixed and other stone type groups were arbitrarily created, so all stones could be assigned a group. Comparisons of baseline patient characteristics (including age, sex and BMI) across stone type groups were carried out using Fisher s exact test for categorical data and ANOVA for continuous variables. Univariable analysis of differences in a basic metabolic panel (including serum sodium, potassium, chloride, carbon dioxide, BUN, creatinine, glucose and calcium concentrations) across stone type groups was carried out using ANOVA. Multinomial logistic regression was used to model the association between basic metabolic panel, 24-h urinalysis and patient demographics (independent variables) with each stone type (dependent variable). We created three predictive models: (i) including the basic metabolic panel only as independent variables; (ii) with the basic metabolic panel and 24-h urinalysis variables as independent variables; and (iii) with patient demographics (age, sex, BMI and comorbidities), the basic metabolic panel and 24-h urinalysis as independent variables. Using these three models, we then calculated the predicted probability of each stone type for every single patient (one for each model). The stone type with the highest probability for each patient was considered the predicted stone type for that patient (one for each model). The predicted stone type was then compared with the actual stone type, and accuracy was then calculated for each individual model. Using predictions from the second model (the one with the basic metabolic panel and 24-h urinalysis data), we created local regression scatter plot smoothing plots to graphically represent the associations between basic metabolic panel variables with each of the predicted stone types adjusted for all other components of the basic metabolic panel and 24-h urinalysis. Only the basic metabolic panel variables were plotted against stone type. All statistical analyses were two-tailed and carried out using R (R Foundation for Statistical Computing, Vienna, Austria). A P < 0.05 was considered to show statistical significance. Results Of the 492 patients in the study, 263 (53%) were men. The mean (standard deviation) age and BMI were 56.4 years (14.1 years) and 30.2 kg/m 2 (7.6 kg/m 2 ), respectively. DM and HTN were identified in 83 (17%) and 202 (41%) patients, correspondingly. A total of 26 (5%), 45 (9%) and 17 (3%) patients were receiving thiazides, potassium citrate and allopurinol, respectively. The median time from stone composition analysis to basic metabolic panel was 6 days. The median time from 24-h urinalysis to basic metabolic panel was 20 days. A total of 272 (55%) patients had predominantly calcium oxalate stones, 100 (21%) had predominantly uric acid stones, 93 (19%) had predominantly calcium phosphate stones, 16 (3%) had mixed stones and 11 (2%) had other types of stones. Mixed stone types were composed on average by 37% calcium oxalate, 44% calcium phosphate and 19% of other elements (mainly uric acid and struvite). Other stone types were composed on average by struvite, 13% ammonium urate and 11% other elements. Significant differences in patient baseline characteristics across these stone type groups were identified (Table 1). For example, uric acid stone formers were older, heavier, and more likely to have DM than the other stone type groups. Calcium oxalate stone formers were predominantly men, whereas calcium phosphate stone formers were mostly women. 196

3 Metabolic panel predicting stone type Table 1 Univariable comparison of baseline patient characteristics across stone type groups Element Calcium oxalate Uric acid Calcium phosphate Mixed Other P-value* Patients, n (%) 272 (55%) 100 (21%) 93 (19%) 16 (3%) 11 (2%) n/a Sex, n (%) <0.001 Male 167 (61%) 59 (59%) 28 (3) 7 (44%) 2 (18%) Female 105 (39%) 41 (41%) 65 (7) 9 (56%) 9 (82%) Mean age, years (SD) 56.2 (13.5) 61.0 (12.4) 52.2 (15.6) 52.4 (16.8) 58.2 (17.1) Mean BMI, kg/m 2 (SD) 29.5 (7.6) 33.9 (8.1) 28.2 (6.2) 28.1 (6.1) 31.7 (7.0) <0.001 Diabetes mellitus, n (%) 34 (13%) 42 (42%) 5 (5%) 0 () 2 (18%) <0.001 Hypertension, n (%) 103 (38%) 59 (59%) 29 (31%) 4 () 7 (64%) <0.001 Thiazide, n (%) 14 (5%) 2 (2%) 5 (5%) 1 (6%) 4 (36%) Allopurinol, n (%) 6 (2%) 10 (1) 1 (1%) 0 () 0 () Potassium citrate, n (%) 19 (7%) 15 (15%) 8 (9%) 2 (13%) 1 (9%) *Fisher s exact test for categorical variables and analysis of variance for continuous variables. Table 2 Univariable comparison of basic metabolic panel across stone type groups Element Calcium oxalate Uric acid Calcium phosphate Mixed Other P-value* Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Sodium (2.4) (3.1) (2.7) (3.5) (2.8) Potassium 3.9 (0.5) 4.0 (0.4) 3.9 (0.5) 4.1 (0.4) 4.0 (0.7) Chloride (2.7) (3.2) (3.0) (4.7) (2.8) CO (2.7) 25.1 (2.6) 26.0 (2.5) 26.2 (4.0) 23.4 (3.3) BUN 16.5 (7.6) 20.4 (7.8) 15.6 (11.4) 16.5 (11.4) 15.1 (7.9) Creatinine 1.03 (0.35) 1.19 (0.48) 0.96 (0.49) 1.03 (0.58) 0.91 (0.2) Glucose (28.6) (47.5) 97.6 (25.5) 94.4 (15.9) (51.3) <0.001 Calcium 9.2 (0.6) 9.1 (0.6) 9.3 (0.5) 9.3 (0.5) 9.1 (1.0) *Analysis of variance. Several differences in the basic metabolic panel were observed across stone type groups (Table 2). For example, uric acid stone formers had the highest serum glucose, BUN and creatinine levels. Calcium oxalate stone formers had serum sodium slightly higher than the other stone groups. However, no significant differences in mean calcium levels across different stone types were identified. We used local regression scatter plot smoothing plots to show the association between basic metabolic panel with each predicted stone type adjusted for all the other urinary components (Fig. 1). The x-axis shows the values of the basic metabolic panel component. The y-axis shows the predicted probability of a specific stone type. For example, the predicted risk of uric acid stone over the other stone types increased with an increase in serum glucose and decreased with an increase in carbon dioxide levels. In contrast, the predicted risk of calcium oxalate stones over the other stone types increased with an increase in serum sodium and chloride levels. The predicted risk of calcium phosphate and oxalate stones over the other stone types increased with an increase in serum calcium levels. There was a non-linear trend towards more uric acid and fewer calcium oxalate stones with an increase in BUN and creatinine. Similarly, there was a trend towards more calcium phosphate and fewer uric acid stones with an increase with CO 2. The multinomial logistic model including the basic metabolic panel only correctly predicted stone type in 59% of the cases. The model with the basic metabolic panel and 24-h urinalysis correctly predicted stone type in 66% of the cases. Finally, the model with the basic metabolic panel, 24-h urinalysis variables and patient demographics correctly predicted stone type in 68% of the cases. Discussion Elements from the basic metabolic panel are frequently used in combination with other laboratory studies to screen stoneforming patients for specific metabolic abnormalities, such as hyperparathyroidism and renal tubular acidosis. 6 However, not all patients have well-defined metabolic syndromes. 4 In addition, it is not clear whether the abnormalities identified in a routine basic metabolic panel can predict stone type. Several studies have investigated the association of stone type with specific metabolic diseases including hyperparathyroidism and renal tubular acidosis. 4,7,8 However, only a few studies evaluated the association between specific changes in the basic metabolic panel with stone formation and stone types. For example, Xu et al. studied the serum electrolytes of 1164 participants of which 714 had a history of urolithiasis. 9 They did not find any significant difference in serum calcium, phosphorus, potassium, sodium, chlorine or carbon dioxide between the stone formers and non-stone formers. Furthermore, Babilash et al. evaluated the differences in metabolic panel between calcium oxalate monohydrate and dihydrate stone formers. 10 They found 197

4 DM MOREIRA ET AL. Probability Probability Probability Probability Sodium BUN Creatinine Chloride Glucose Potassium CO Calcium Fig. 1 Figure showing local regression plots of the basic metabolic panel and stone type predicted probabilities. Of note, the stone probabilities reflect the risk of a particular stone type over the other types of stones., Calcium oxalate;, uric acid;, calcium phosphate;, mixed;, other. no significant difference between the two groups. Thus, until now it was unclear whether basic metabolic profile can predict stone type. In the present study, we investigated the association between the basic metabolic panel and stone type. We found several significant associations between serum elements and specific stone types. For example, uric acid stone formers had significantly higher glucose levels. Similarly, previous studies found a higher risk of uric acid stone formation among patients with DM In addition, a recent study showed increased overall urolithiasis in patients with poorly-controlled DM. 14 In the present study, we found the predicted risk of uric acid stones over the other stone types increased with an increase in serum glucose. Thus, the present findings suggest that the increased risk of stones among poorly-controlled diabetics is likely a result of increased uric acid stone formation. Furthermore, we found the predicted risk of calcium oxalate stones over the other stone types increased with increases in serum sodium and chloride levels. Indeed, patients with calcium oxalate stones had the highest mean calcium levels. Several previous studies have shown an association between sodium chloride intake with hypercalciuria and calcium oxalate stone formation Although the mechanisms linking serum levels of sodium and chloride to a higher risk of calcium oxalate were not elucidated by the present study, our results reinforce the potential role of sodium chloride in calcium oxalate stone formation. Furthermore, we found the predicted risk of calcium phosphate increased with increases in serum CO 2. Also, patients with calcium phosphate stones had the highest mean CO 2 levels. This was an interesting finding that suggests, despite the association of calcium phosphate stones with renal tubular acidosis, on average calcium phosphate stone formers have higher levels of CO 2. Additionally, we found the predicted risk of calcium phosphate and oxalate stones over the other stone types increased with increases in serum calcium levels. Likewise, multiple previous studies have showed an association between hyperparathyroidism-associated hypercalcemia and calcium stones. 19,20 The present findings, however, indicate the risk of calcium stones over the other stone types increases with the severity of hypercalcemia, suggesting a potential dose response relationship. The knowledge regarding stone composition could change management. For example, patients with uric acid stones might be offered dissolution treatment. Thus, we evaluated the accuracy of the basic metabolic panel in predicting stone type. We found the multivariable model using the basic metabolic panel to predict stone type correctly predicted stone type in just 59% of the cases. Given that in a previous study a model with 24-h urinalysis data correctly predicted stone type in 64% of the cases, we combined patient demographics, 24-h urinalysis data and basic metabolic panel with a resulting accuracy increase to 68%; i.e. approximately two-thirds of the stone types were correctly predicted by the model. 5 These findings suggest the basic metabolic panel alone does not accurately predict stone type. There are several potential reasons as to why demographics, the basic metabolic panel and 24-h urinalysis do not accurately predict stone type. For example, lithogenic factors not routinely detected by the basic metabolic panel and 24-h urinalysis might play an important role in the formation of specific stone types. 5 In addition, given that the components of the basic metabolic panel and 24-h urinalysis vary over time; the use of one measurement does not capture this temporal variability. 21,22 Nevertheless, given the significant associations between the basic metabolic panel and specific stone types, the basic metabolic panel could be used in conjunction with other variables to predict stone composition. The current study is limited by its retrospective nature, where only patients selected for 24-h urinalysis, basic metabolic panel and stone composition analysis were included. Although this can introduce selection bias, it reflects the current clinical practice in a busy urolithiasis clinic. In addition, our sample comes from a tertiary-care center in a specific region of the USA (Long Island, NY, USA), which might not necessarily replicate 198

5 Metabolic panel predicting stone type the characteristics of the average USA population. Furthermore, detailed data on medications, comorbidities, stone burden and dietary factors were not systematically available for all patients. Given the basic metabolic panel could be influenced by the aforementioned factors; it is conceivable that differences in such variables across stone types could in part explain the present results. Additionally, we had to group several types of stones (e.g. ammonium urate, struvite and calcium carbonate) into one group (i.e. other types of stones) because of the limited number of such stone types. Finally, we analyzed only one basic metabolic panel per patient, and therefore we were unable to determine how changes in serum values over time affect how the basic metabolic panel can predict stone type, as multiple serum values were not available for all patients. In conclusion, among a cohort of stone formers with the basic metabolic panel, 24-h urinalysis and stone composition available, the basic metabolic panel alone or in combination with 24-h urinalysis and demographics did not accurately predict stone type, and should not be used alone to predict stone type. However, variations in the basic metabolic panel were associated with one or more specific types of stone. Specifically, higher glucose levels were associated with uric acid stones, and higher sodium and chloride levels were associated with calcium oxalate stones. Therefore, the basic metabolic panel could be used in combination with other variables (such as radiological findings) to predict stone composition. Conflict of interest None declared. References 1 Kamihira O, Ono Y, Katoh N, Yamada S, Mizutani K, Ohshima S. Long-term stone recurrence rate after extracorporeal shock wave lithotripsy. J. Urol. 1996; 156: Parks JH, Worcester EM, Coe FL, Evan AP, Lingeman JE. Clinical implications of abundant calcium phosphate in routinely analyzed kidney stones. Kidney Int. 2004; 66: Kacker R, Meeks JJ, Zhao L, Nadler RB. Decreased stone-free rates after percutaneous nephrolithotomy for high calcium phosphate composition kidney stones. J. Urol. 2008; 180: ; discussion Pak CY, Poindexter JR, Adams-Huet B, Pearle MS. Predictive value of kidney stone composition in the detection of metabolic abnormalities. Am. J. Med. 2003; 115: Moreira DM, Friedlander JI, Hartman C, Elsamra SE, Smith AD, Okeke Z. Using 24-hour urinalysis to predict stone type. J. Urol. 2013; 190: Goldfarb DS, Arowojolu O. Metabolic evaluation of first-time and recurrent stone formers. Urol. Clin. North Am. 2013; 40: Pak CY, Adams-Huet B. Elucidation of factors determining formation of calcium phosphate stones. J. Urol. 2004; 172 (6 Pt 1): Raj GV, Auge BK, Assimos D, Preminger GM. Metabolic abnormalities associated with renal calculi in patients with horseshoe kidneys. J. Endourol. 2004; 18: Xu QQ, Huang XB, Ma K et al. Analysis of serum electrolytes in urolithiasis patients. Beijing Da Xue Xue Bao 2010; 42: Bibilash BS, Vijay A, Fazil Marickar YM. Stone composition and metabolic status. Urol. Res. 2010; 38: Daudon M, Traxer O, Conort P, Lacour B, Jungers P. Type 2 diabetes increases the risk for uric acid stones. J. Am. Soc. Nephrol. 2006; 17: Chen HS, Su LT, Lin SZ, Sung FC, Ko MC, Li CY. Increased risk of urinary tract calculi among patients with diabetes mellitus-a population-based cohort study. Urology 2012; 79: Pak CY, Sakhaee K, Moe O et al. Biochemical profile of stone-forming patients with diabetes mellitus. Urology 2003; 61: Weinberg AE, Patel CJ, Chertow GM, Leppert JT. Diabetic severity and risk of kidney stone disease. Eur. Urol. 2014; 65: Lemann J Jr, Gray RW. Idiopathic hypercalciuria. J. Urol. 1989; 141 (3 Pt 2): Borghi L, Schianchi T, Meschi T et al. Comparison of two diets for the prevention of recurrent stones in idiopathic hypercalciuria. N. Engl. J. Med. 2002; 346: Massey LK, Whiting SJ. Dietary salt, urinary calcium, and kidney stone risk. Nutr. Rev. 1995; 53: Nouvenne A, Meschi T, Prati B et al. Effects of a low-salt diet on idiopathic hypercalciuria in calcium-oxalate stone formers: a 3-mo randomized controlled trial. Am. J. Clin. Nutr. 2010; 91: Parks JH, Coe FL, Evan AP, Worcester EM. Clinical and laboratory characteristics of calcium stone-formers with and without primary hyperparathyroidism. BJU Int. 2009; 103: Berger AD, Wu W, Eisner BH, Cooperberg MR, Duh QY, Stoller ML. Patients with primary hyperparathyroidism-why do some form stones? J. Urol. 2009; 181: Bek-Jensen H, Tiselius HG. Repeated urine analysis in patients with calcium stone disease. Eur. Urol. 1998; 33: Healy KA, Hubosky SG, Bagley D. 24-Hour urine collection in the metabolic evaluation of stone formers: is one study adequate? J. Endourol. 2013; 27: Editorial Comment Editorial Comment to Association of serum biochemical metabolic panel with stone composition Stone disease afflicts a large population. Many patients have recurrent disease. Stone retrieval renders only temporary relief, as the cause for stone formation continues to exist in the patient. Stone composition varies widely geographically. 1 Identification of stone type by assessing metabolism would help planning stone dissolution and prophylaxis. 2 Several studies have attempted to identify this with little success. In our study, we recognized that the correlation between metabolic abnormality and stone composition is positive in patients with uric acid stone and children with primary hyperoxaluria. 3 In a study of 137 patients, stones were classified as calcium oxalate monohydrate (COM), calcium oxalate dihydrate (COD), uric acid, COM and COD, oxalates and uric acid, and mixed stones. The 24-h urine and blood biochemistry showed that there are some differences in the different types of stone formers. COM stone patients had significantly higher urine oxalate compared with COD patients (P < 0.01). COD patients had significantly lower mean values for urine magnesium compared with COM patients (P < 0.05). COD patients had significantly higher mean values for the urine calcium-to- 199