The estimated glucose threshold (eTmG/GFR) - a helpful tool in SGLT2 Inhibition?

Assessing tubular function by measuring urinary excretions of different substances is an important albeit somewhat dissipating part of clinical nephrology.

 Gradient-limited versus transport-limited tubular reabsorption

Fig. 1. Formulas for calculating the fractional excretion (FE) and the fractional tubular reabsorption (TR) of a substance Z.

Fig. 1. Formulas for calculating the fractional excretion (FE) and the fractional tubular reabsorption (TR) of a substance Z.

Absolute excretion rates and indices like the fractional excretion (FE) or its counterpart the tubular reabsorption (TR) are commonly used to assess tubular function (Fig. 1).

These are quite informative for substances that are mainly “gradient-limited” (e.g. sodium), where the reabsorptive capacity is confined by the maximal concentration gradients, that tubular cell membrane channels and tight junctions can sustain. Here absolute and relative excretion rates are mainly dependent on and as such representative of tubular function itself.

Fig. 2. The filtered load (FL) of a sustance Z, the amount filtered by the glomeruli, is given by the above formula. 

Fig. 2. The filtered load (FL) of a sustance Z, the amount filtered by the glomeruli, is given by the above formula. 

This is different with “transport-limited” tubular reabsorption (e.g. glucose, phosphate). These processes are enzyme-based, often follow Michaelis-Menten kinetics and exhibit an absolute upper limit to the amount that can be reclaimed from the tubular lumen. Therefore the urinary excretion of these substances is not only dependent on tubular function, but also strongly influenced by the filtered load (FL), i.e. by the GFR and the respective plasma concentration of the particular solute (Fig 2).

Fig 3.

Fig 3.

The theoretical relationship between filtered load and urinary excretion of transport-limited substances is shown in Fig. 3.

The glucose threshold

“In normal individuals as the concentration of glucose in the plasma rises the amount [of glucose] excreted [in the urine] remains extremely small till a limiting plasma concentration, the threshold, is reached. As the concentration in the blood rises above the threshold, progressively more glucose is excreted” (1).

The threshold concentration of glucose (and any other substance) can be reckoned by dividing the tubular transport maximum of glucose by the GFR (Fig 4). It represents the fraction of the plasma glucose concentration whose filtration can be maximally reabsorbed by the tubule. It has therefore also been called the aglucosuric glucose concentration (2).

Fig. 4. How to derive the glucose threshold concentration (TmG/GFR) from the filtered load (FL). The rectangular brackets [ ] signify concentrations.

Fig. 4. How to derive the glucose threshold concentration (TmG/GFR) from the filtered load (FL). The rectangular brackets [ ] signify concentrations.


By accepting the creatinine clearance as a surrogate of the GFR it can be easily estimated by measuring the concentrations of glucose and creatinine in a plasma and a spot urine sample respectively (Fig. 5).

Using this estimated tubular threshold concentration is a great way to characterize tubular handling of glucose (and other transport-limited substances) because it eliminates the filtered load from the equation and expresses only tubular function itself.

Fig. 5. Estimating the tubular threshold for glucose (eTmG/GFR). FE fractional excretion, TR tubular reabsorption.

Fig. 5. Estimating the tubular threshold for glucose (eTmG/GFR). FE fractional excretion, TR tubular reabsorption.


How to deal with splay

Due to heterogeneity in the number and affinity of glucose transport sites, the maximal tubular transport capacity of singular nephrons extends over a certain range. This means that some nephrons become glucosuric at lower concentrations than others. Only after the transport maximum of all individual nephrons is exceeded does the urinary excretion get proportional to the filtered load (Fig. 6a). The usually small amounts excreted before that have been called splay (Fig. 6b).

Fig. 6a.

Fig. 6a.

Fig. 6b.

Fig. 6b.

Fig. 6c.

Fig. 6c.

“Extrapolation of [the linear part of the urinary excretion line] to its point of intersection with the abcissa yields [the so called ‘line’ threshold (Fig. 6c)], … [the] value of blood glucose at which excretion would begin were there no splay” (2).

The estimated glucose threshold represents a good approximation of this theoretical (‘line’) threshold, as long as the splay comprises only a small part of the whole quantity of glucose excreted.  Predicting wether this is the case in an individual patient might not be trivial. I am not aware of any empirically validated approaches and adaptations to specific clinical situations might be needed. I would suggest the following aids:

  • In analogy to work done with phosphate (3), I would consider it safe to ignore splay when the fractional excretion of glucose is above 15%.*
  • The higher the absolute amount of glucose excreted, the more dependable the estimated threshold becomes.
  • In cases where splay is quantitatively significant the estimated threshold always underestimates the theoretical one.

 Using the threshold in SGLT2 inhibition

SGLT2 inhibitors interfere with glucose reabsorption in the proximal tubule. Their clinical effectiveness probably correlates with the absolute amount of glucose excreted. As explained above their clinical impact is therefore not only dependent on SGLT2 inhibition itself, but also strongly influenced by the glycemic control of the individual patient. The estimated glucose threshold, which eliminates the impact of glycemia, could thus be an important tool to better analyze the pharmacodynamic effects of the SGLT2 inhibitors.

I would expect the eTmG/GFR to be a helpful tool

  • to improve adherence monitoring
  • to allow dose titration in the individual patient
  • to improve safety (eg by reducing the risk if ketogenesis)


Clearly, estimating the tubular threshold of glucose is not rocket science and empirical validation of the concept is needed. Nonetheless I think it has the potential to be clinically helpful and I would love to see it evaluated in clinical practice.


  1. Woolf LI, Goodwin BL, Phelps CE. Tm-limited renal tubular reabsorption and the genetics of renal glucosuria. Journal of Theoretical Biology. 1966;11(1):10–21.
  2. Smith HW. The kidney: Structure and function in health and disease. New York: Oxford University Press; Dezember 1, 1951: 81-96.
  3. Payne RB. Renal tubular reabsorption of phosphate /TmP/GFR): indications and interpretation. Ann Clin Biochem. 1998;35:201-206.
  4. DeFronzo RA, Norton L, Abdul-Ghani M. Renal, metabolic and cardiovascular considerations of SGLT2 inhibition. Nat Rev Nephrology. 2017; 13:11-26.

* This should work in patients treated with dapaglifozin and empaglifozin where splay is reduced (4), also in glucosuric diabetic and nondiabetic patients not treated with SGLT2 inhibitors, and patients with familial renal glucosuria type A. More caution is probably needed in patients with familial renal glucosuria type B and possibly in patients treated with other SGLT2 inhibitors where splay might be increased. Nonetheless with high enough glucose excretion the estimation should work in these cases as well.

Identifying Crystals in the Urinary Sediment

Crystalluria is a common finding in the urinary sediment. Although its clinical significance is oftentimes limited, most people have a natural desire to identify these frequently nifty structures. Unfortunately this is a rather difficult affair. Osler's most famous dictum about medicine in general (" of of probability") perfectly sums up microscopic crystal identification in the urine as well.

Fig. 1. Different forms of uric acid crystals (all by bright field microscopy).

The basic problem relates to the nonspecific nature of many crystal morphologies: As a general rule the same substance can form many different crystals (Fig. 1), and very similar crystals can consist of completely different compounds (Fig. 2).

Fig. 2. Dumbbell-shaped crystals. Left composed of uric acid, right calcium oxalate monohydrate. (upper part bright field, lower left phase contrast, lower right polarized light).

Without any means of physicochemical verification, the accurate assignment of crystals really is a challenge. It depends on a thorough integration of diverse properties:

  • Morphology (form, structure, color, size) by phase contrast and bright field microscopy
  • Birefringence (intensity, color) in polarized light +/- red compensator
  • Surroundings (accompanying crystals, bacteriuria/leucocyturia, amorphous material...)
  • Urinary pH

With careful consideration off all these factors, many crystals -especially when abundantly present- can be identified with clinical certainty. In contrast more sporadic and/or very nonspecific crystalline structures frequently have to go unidentified. The ease with which crystals are sometimes ascribed to certain substances or drugs does not measure up to the scope of the problem and amazes me time and again.

I highly recommend allowing oneself enough room for error and uncertainty, when trying to identify crystals in the urinary sediment.

Liver Doppler Ultrasound - Pulsatility in Portal Vein Flow

The flow pattern in the portal vein is usually monophasic with a near constant velocity throughout the cardiac cycle (Fig. 1 and 2).

Fig. 1. A normal monophasic portal vein flow pattern.

Fig. 2. Another example of normal flow in the portal vein.

Decreased compliance of the liver vascular bed and/or increases in blood volume flow to the liver lead to enhanced pulsatility in portal vein flow. Common reasons include cirrhosis of the liver, right-sided CHF and tricuspid regurgitation (1).

Increased pulsatility usually takes the form of an undulating waveform with one dip and one peak per cardiac cycle (Fig. 3 and 4). 

Fig. 3. Pulsatile portal vein flow.

Fig. 4. Increased pulsatility in a patient with tricuspid regurgitation.

Occasionally the whole central venous phasicity is transmitted across the sinusoids resulting in a tetrainflectional, triphasic flow pattern in the portal vein (Fig. 5).


Fig. 5. Triphasic, tetrainflectional flow in the portal vein caused by severe right-heart failure.


Mirroring this waveform around the baseline (Fig. 6) results in a perfect depiction of a central venous flow pattern, in this case with a relatively blunted systolic (X) and a prominent diastolic (Y) descent typical of right-sided heart failure and/or tricuspid regurgitation (2).

Fig. 6. Portal vein flow shown in figure 5 (upper part) and its mirror image (lower part), which perfectly matches a central venous flow pattern typical of right-sided heart failure +/- tricuspid regurgitation. The flow waves are labeled in accordance to central venous pressure wave nomenclature.


  1. McNaughton DA, Abu-Yousef MM. Doppler US of the liver made simple. RadioGraphics. 2011;31(1):161–188. doi:10.1148/rg.311105093.
  2.  Scheinfeld MH, Bilali A, Koenigsberg M. Understanding the spectral Doppler Waveform of the Hepatic veins in health and disease. RadioGraphics. 2009;29(7):2081–2098. doi:10.1148/rg.297095715.

Digiscoping Urine - Amorphous Phosphates

Amorphous phosphates are a common finding in the urinary sediment (Fig. 1 and 2).

Figure 1. Amorphous phosphates (shiny elements) by phase-contrast microscopy (x400).

Figure 2. Same section by bright-field microscopy (x400).

Distinguishing phosphates from urates

Figure 3. White sediment typical of amorphous phosphates.

Morphologically, amorphous phosphates look pretty much identical to amorphous urates by bright-field and phase-contrast microscopy.
Contrary to amorphous urates, which usually show a clear glow in polarized light, amorphous phosphates exhibit no birefringence (1). Whereas urates form in acidic urine, phosphates are observed in alkaline urine (2). Importantly, macroscopic inspection of the sediment oftentimes allows separation of the two (2): Amorphous urates tend to give the sediment a brick-red hue. In contrast amorphous phosphates are characterized by a chalk-like white sediment (Fig. 3).


Elements in the urinary sediment that resemble casts but are not formed in the kidney tubules are referred to as pseudocasts (1). 
It is not unusual for amorphous phosphates to aggregate in castlike structures, which can resemble granular casts (Fig. 4-7). The absence of a cast matrix typically enables the identification of these aggregates as pseudocasts.

Figure 4. Phosphate pseudocast (phase contrast, x400).

Figure 6. Another phosphate pseudocast (phase contrast, x400).

Figure 5. Bright field (x400). Note the missing cast matrix.

Figure 7. Bright field (x400).

(all images taken with iPhone 5, microscopic pictures through Leica DMLB microscope)


  1. Fogazzi GB. The urinary sediment: An integrated view. 3rd ed. Italy: Elsevier Srl; 2010. 
  2. Mundt L, Shanahan K. Graff’s textbook of Urinalysis and body fluids. United States: Lippincott Williams and Wilkins; September 28, 2015.

Renal Ultrasound - Cortical Nephrocalcinosis

Figure 1. Kidney model depicting renal medulla ("pyramids") and cortex with interpyramidal renal columns ("columns of Bertin").

Macroscopic nephrocalcinosis - the detection of calcium deposits on radiological imaging - can be divided into cortical and medullary forms.

Cortical nephrocalcinosis is quite rare, accounting for less than 3% of all cases of nephrocalcinosis (1). 

Commonly cited causes include (1-3):

  • Renal cortical necrosis
  • Chronic glomerulonephritis
  • Alport syndrome (Fig. 2-4)
  • Oxalosis

Figure 2. Cortical nephrocalcinosis in a patient with Alport syndrome. The renal cortex displays an intensely hyperechogenic outer rim, which extends into the renal columns. (see captions in Figure 4)

Figure 3. Diffuse "twinkling" of the hyperechogenic rim.

Figure 4. Detail of Figure 2. Surface of the kidney marked by red line, pyramids shaded blue. Red arrows highlight hyperechogenic renal columns.

The mechanisms of calcium deposition in cortical nephrocalcinosis are not well studied, if at all. Most cases are probably the result of dystrophic calcifications (i.e. due to local tissue abnormalities). Occasionally, disorders of calcium metabolism may also play a role (1).


  1. Wrong O. Nephrocalcinosis. In: Oxford textbook of clinical nephrology. 2nd ed. Cameron S, Davison AM, eds. Oxford: Oxford University Press; October 9, 1997. 
  2. Schepens D, Verswijvel G, Kuypers D, Vanrenterghem Y. Renal cortical nephrocalcinosis. Nephrology Dialysis Transplantation. 2000;15(7):1080–1082. doi:10.1093/ndt/15.7.1080. 
  3. Zagoria RJ. Genitourinary Radiology: The requisites. 2nd ed. United States: Elsevier Health Sciences; June 4, 2004.



Digiscoping Urine - Hemoglobin Casts

Hemoglobin casts are thought to commonly arise from decomposing erythrocytes trapped within the usual cast matrix. They are a sign of renal parenchymal bleeding (1).

(all pictures phase contrast, x400)

When present, hemoglobin casts are easily detected due to their showy orange-brown coloration. At times it might be difficult to tell them apart from muddy-brown casts seen with acute tubular injury. Detailed examination, looking for remnants of red-cell membranes and excluding the presence of fine black granules, helps resolve this issue. Furthermore hemoglobin casts usually occur on a background of significant glomerular hematuria.

More rarely hemoglobin containing casts may be found in patients with severe hemolysis. (Sadly, I'm still waiting for it.) Obviously you would not expect to see erythrocyte remnants in these cases.


1. Fogazzi G, Ponticelli C, Ritz E. The urinary sediment: An integrated view. United Kingdom: Oxford University Press; February 3, 2000.

Dysnatremia Tables

swissnephrokalk Dysnatremia Tables (.xlsx)

In the management of hyper- and hyponatremia a quantitatively sound approach is of utmost importance to ensure adequate speed and magnitude of tonicity correction.

The Edelman Equation

The relationship between P-[Na] and the amount of exchangeable sodium and potassium and the TBW.

The relationship between P-[Na] and the amount of exchangeable sodium and potassium and the TBW.

Thanks to Edelman and coworkers (Ref. 1) we have an empirical validation of the theoretically plausible relationship between the plasma sodium concentration (P-[Na]) as a surrogate of tonicity on one hand, and the exchangeable amount of sodium and potassium in the body in addition to the total body water (TBW) on the other. 

The Edelman equation as published by Nguyen and Kurtz (Ref. 2).

The Edelman equation as published by Nguyen and Kurtz (Ref. 2).

Over the years the original equation has been used in different, often simplified formulations. For clinical practice the differences are probably of negligible magnitude. Here I use the formula published by Nguyen and Kurtz (Ref. 2).

By using the Edelman equation the effects of changes in sodium-, potassium- and water balance on the P-[Na] can be calculated accurately.

Myths and Reality

Unfortunately many clinicians use the equation only to estimate the impact on the P-[Na] of a certain amount of infusion. Even worse the misconception that this simple calculation will predict the future P-[Na] is widespread. No wonder that clinicians are often less than impressed with the formulas perfomance.

In reality the equation can of course only estimate the expected difference in P-[Na] solely due to the addition or excretion of the stated solution in the stated amount. All other changes that influence the P-[Na] are obviously not accounted for by one simple calculation. To predict the P-[Na] during the course of treatment the complete balance of water, sodium and potassium has to be considered.

Dysnatremia Tables

To use these rather laborious calculations in a clinically reasonable time frame, I'm using an excel sheet, the swissnephrokalk Dysnatremia Tables.

swissnephrokalk Dysnatremia Tables

swissnephrokalk Dysnatremia Tables

With the only input being the current P-[Na] and an estimate of the TBW, the Dysnatremia Tables allow rapid and easy

  • Assessment of the impact of complex infusion and supplementation regimens on the P-[Na].
  • Provision of corrective steps with hypertonic solutions or free water.
  • Development of safe follow-up strategies using output/urine flow monitoring.


The Importance of Urine Volume and Potassium

Two important points, that become very obvious when using the Dysnatremia Tables regularly, merit special emphasis, as they are often neglected in the care of patients with dysnatremia:

  1. It is virtually impossible to overcorrect hyponatremia by the infusion of isotonic fluids alone. Excessive loss of free water, usually by a water diuresis, is by far the most common reason of to fast an increase in the P-[Na]. Urine output monitoring is therefore an essential part of the management of hyponatremic patients. 
  2. Do not forget the impact of potassium supplementation or loss on the P-[Na]. It might not have the same clinical consequences, as the potassium will preferentially end up intracellularly, but the P-[Na] depends as much on the potassium balance as on that of sodium.

Final Remarks

  • Minor deviations in the estimation of the TBW usually do not have a significant impact on the results of the formula.
  • The calculations are based on equilibrated conditions, which obviously might not be present in every patient.
  • The volumes and infusion types used in the Dysnatremia Tables can easily be adjusted to your personal needs.
  • The calculations should not supersede regular measurement of the P-[Na]. They are a helpful tool to design the right therapy from the start and keep on track during the course of treatment.



  1. Edelman IS, Leibman J, O’Meara MP, Birkenfeld LW. Interrelations between serum sodium concentration, serum osmolarity and total exchangeable sodium, total exchangeable potassium and total body water. Journal of Clinical Investigation. 1958;37(9):1236–1256. doi:10.1172/jci103712.
  2. Kurtz I, Nguyen MK. Evolving concepts in the quantitative analysis of the determinants of the plasma water sodium concentration and the pathophysiology and treatment of the dysnatremias. Kidney International. 2005;68(5):1982–1993. doi:10.1111/j.1523-1755.2005.00652h.x


Appendix - Examples

Consider a 67yo woman with a P-[Na]=116mmol/l. She weighs 74kg and you estimate her TBW at 37l.

  • Because she is hypovolemic you want to give her 2000ml of NaCl 0.9% over the next 24h. What impact on the P-[Na] will this have? (Answer: +1mmol/l, (A)).
  • She is also profoundly hypokalemic. How would supplementing 80mmol of KCl affect the P-[Na]? (Answer: +2.2mmol/l, (B)).
  • You want the P-[Na] to increase no more than 8 mmol/l in the first 24h. So how much maximally dilute urine ([Na+K]=20mmol/l) can this patient pass without risking overcorrection? (Answer: 1500ml, (C), (8-1-2.2=4.8)).
  • Let's assume she develops a seizure due to cerebral edema. Now you want to increase her P-[Na] on the spot by 2mmol/l. How much NaCl 3% or NaBic 8.4% do you have to give? (Answer: 200ml NaCl 3%, 100ml NaBic 8.4%, (D))
  • Now a different scenario: 8 hours ago her P-[Na] was at 105mmol/l. You want to lower her P-[Na] by 4mmol/l. How much additional free water do you have to give? (Answer: ~1100ml, (E)).
Trust me, this only looks complicated!

Trust me, this only looks complicated!

Cautionary Note: Adding up the values from different columns is strictly speaking incorrect, as changes in the TBW that might be taking place are not correctly accounted for. The magnitude of the error is usually well below 1mmol/l. If that is not good enough for you, you can add up all sodium and potassium, calculate an average concentration of Na and K, and manually plot it against the water balance. (In the above example 2000ml of normal saline plus 80mmol of potassium gives an average [Na+K] of 194mmol/l, which translates into an increase in P-[Na] of 3.1mmol/l (F) versus 3.2mmol/l calculated above (A+B).)



Cystinuria calculator

swissnephrokalk Cystinuria (.xlsx)

Instructions for use (.pdf)

Influencing the amount and concentration of cystine in the urine plays a central part in the treatment of patients with cystinuria. These values are expressed in different ways by different labs and investigators.
The Cystinuria sheet allows easy conversion between some of the commonly found units. In addition the necessary urine volume to achieve a supersaturation ratio below 1 and the expected effects of cystine binding thiol drugs can be calculated (see instructions).
Feedback and corrections are welcomed in the comment section, by twitter or via email (

swissnephrokalk Cystinuria

CT-guidance for bone biopsies in renal osteodystrophy?

Last week, I went to a great symposium about osteoporosis and renal osteodystrophy, organized by Andreas Jehle at the University Hospital of Basel. There were several interesting talks. What really struck me as extraordinary, though, was the suggestion of doing bone biopsies with CT-guidance.

Getting enough trabecular bone is a critical and apparently common technical problem, when doing biopsies for metabolic bone diseases. C. Zech, an interventional radiologist, demonstrated conclusively how CT-guidance can assist in correct positioning and angulation of the bone biopsy needle, to get as much spongiosa as possible. Targeting the crista iliaca from a posterior approach, he uses only an 8 Gauge needle (outer diameter 4.2mm), which renders the whole procedure essentially identical to a bone marrow biopsy for hematological purposes (apart from the CT, of course). This approach appeals to me much more than commonly described techniques (e.g. Chappard, Hernandez), which seem to be more invasive and requiring special expertise.

As the team in Basel is just starting with their program, I would be very interested in hearing, whether someone else has experiences with CT-guided bone biopsies in the assessment of patients with renal osteodystrophy. Please leave a comment, send me an email (, or contact me via twitter!

Estimating the dynamic GFR when creatinine concentrations are changing rapidly

Have you ever wondered how to calculate GFR from plasma creatinine in the non-steady state? I did from time to time. So I was kind of shocked to learn from Chen (1) and Winter (2) how easy it actually is: 


You basically treat creatinine as a drug: 

  1. Your dosing rate is estimated by the creatinine generation rate, using for example the Cockcroft-Gault equation.
  2. The difference between two creatinine concentrations times the distribution volume of creatinine gives you the amount of creatinine that is retained in the body.
  3. Subtracting this from the generation rate you get the amount of creatinine excretion over the corresponding time period.
  4. Divide that by the mean creatinine concentration in the blood and -voila- you end up with the creatinine clearance.
Bildschirmfoto 2015-05-18 um 00.27.20.png

Crea-Clearance versus GFR

Most clinicians are using some form of GFR estimation (MDRD, CKD-EPI) in their daily clinical practice and not creatinine clearance estimates. The formulas used by Winter and Chen essentially calculate an absolute creatinine clearance, not a GFR. The term "kinetic GFR" coined by Chen is thus a misnomer.

To overcome this problem I have taken the result of the non-steady state Crea-Clearance (= Chen's kinetic GFR) and the creatinine generation rate to "back"calculate a corresponding -virtual- steady state plasma creatinine concentration. That value is then used to get a GFR estimate by CKD-EPI. I have termed this the "dynamic GFR".

Generation rate and distribution volume of creatinine

Contrary to Chen, I have chosen the Cockcroft-Gault formula as the default for estimating the creatinine generation rate, and not MDRD. First, because the calculation is based on creatinine clearance. Second because with MDRD your estimated creatinine generation rate changes with different levels of baseline kidney function. Also, I prefer using the distribution volume over the maxDeltaCrea suggested by Chen. (As a reference to Chen, I have included a calculator that allows conversion of maxDeltaCrea to Vd/kg BW, though). 


Using the formulas freehand in daily clinical practice is quite a challenge. So I put them together in an Excel sheet, which 

A) gives you the graphical trend of dynamic versus steady state GFR.

B) presents a tabular view of these values and the corresponding dynamic and steady state Crea-clearances.

C) works in SI and conventional units of creatinine concentration.

D) allows you to manually adjust the daily creatine generation rate, the distribution volume of creatinine and other parameters.

A simple instruction on how to use the calculator can be found here.

Use and limitations

Practical use and limitations of the method are nicely discussed by Chen. The limits obviously relate to problems with estimating generation rate, distribution volume and mean plasma concentration of creatinine. Although these problems can be profound, one should not forget, that they are not exclusive to this approach. In fact they are at work whenever we are using plasma creatinine concentrations to monitor rapidly changing renal function.

Any corrections, feedback, questions, or suggestions are highly welcomed!


  1. Chen S. Retooling the Creatinine clearance equation to estimate kinetic GFR when the plasma Creatinine is changing acutely. Journal of the American Society of Nephrology. 2013;24(6):877–888. doi:10.1681/asn.2012070653.
  2. Winter ME. Basic clinical pharmacokinetics. 5th ed. Philadelphia: Wolters Kluwer/Lippincott Williams & Wilkins Health; October 1, 2009.