The results demonstrate that Australian general practice patients with CKD are frequently prescribed PIP. Approximately 35% of the patients were prescribed at least one PIP. The findings have corroborated the high rate of potentially contraindicated medications and inappropriate dosing found in international studies, including a systematic review that reported rates of PIP between 13% to 80%.[2, 4] These results highlight the need for further research to determine the reasons, leading to interventions to optimise prescribing in patients with CKD in general practice.
The extent of PIP in our study and the previous literature highlights the complexity of prescribing in CKD and could be due to the contribution of several factors. Firstly, in Australia there is a lack of up-to-date accessible guidelines on drug dose adjustments in patients with impaired kidney function. Another important aspect to consider is the inconsistencies with dosage adjustment recommendations between different drug information sources, as well as the product information of different brands of the same drug.[14, 15] It is important to note that most drug information sources provide dosage adjustment based on the CG equation rather than eGFR formulae (MDRD/CKD-EPI), whilst eGFR (especially CKD-EPI) based formulae have been found to be the most accurate indicator of kidney function. Secondly, many of the recommendations based on the CG equation are questionable due to the variability in creatinine assays at the time and were prior to SCr being isotope dilution mass spectrometry (IDMS) standardised. CG-estimated GFR results are 5-10% higher using the standardised SCr measurements compared to the non-standardised SCr and relying on the CG equation could lead to unintended consequences, including insufficient dose adjustments for kidney function.[17, 18] However, the clinical significance of this theoretical issue is unknown and needs to be evaluated further. Hence, current recommendations include using either CG equation due to considerable experience with the formula even if it is based on creatinine assays not in use. It is also appropriate to use eGFR formula (CKD-EPI) for most drugs without body surface area (BSA) adjustment.[1, 16]
The nature of the electronic health record (eHR) may also be a potential factor for PIP. The current eHR may not alert prescribers to appropriately adjust dosages of medications cleared by the kidney unless the patient has CKD recorded as a diagnosis. Only 20% of the patients included had a formal diagnosis of CKD documented whilst all patients had laboratory evidence of CKD and this lack of coding of CKD as a condition may have had a role in the PIP findings.[7, 8] Whilst the overall prevalence of CKD in our study was comparable to the previous Australian Bureau of Statistics (ABS) results, the prevalence is lower than previous International literature. The potential reasons for the lower prevalence have been described in detail elsewhere. It is important to note that in many older people with stable eGFR values from 45-59 mL/min/1.73 m2 debate on the definition and staging of CKD exists as to whether CKD may be over diagnosed. However, these values are a sign of impaired kidney function that could affect the clearance of drugs. Hence, it is essential to consider renal function when prescribing renally-cleared medications as the PIP may lead to adverse outcomes.
The agreement between the two equations (CG and CKD-EPI) was excellent, with 97% of the medications rated as appropriate by eGFR being also rated as appropriate by the CG equation. Previous studies comparing drug doses derived from eGFR equations and CG have commonly reported discordance rates between 10% and 40%.[19, 20] These results highlight that for most patients with renal impairment the same recommendation between eGFR or CG estimate can be used, with any difference in the kidney function estimate unlikely to lead to an overdose  However, it is important for some drugs, and the potential clinical significance of disagreement between eGFR and CG-based dosing regimens should be minimised by using sound clinical judgment. For example, our study showed that rivaroxaban was one of the medications associated with disagreement between the equations. A recent study by Szummer et al comparing eGFR (CKD-EPI, MDRD) and the CG equation suggested that when prescribing one of the novel oral anticoagulants, the CG equation should be considered, as it provides a more conservative approach for avoiding drug exposure and reducing the risk of bleeding. A similar approach is recommended by the National Kidney Disease Education Program (NKDEP) in the USA and by KHA, where eGFR can be used for dosage adjustment for most drugs (without BSA adjustment) except for drugs with a narrow therapeutic index (e.g. anticoagulants), where conservative kidney function estimates and corresponding doses are recommended, particularly if therapeutic drug monitoring is not readily available.
Most of the medications with PIP were consistent with previous studies. Antidiabetic agents and cardiovascular drugs were commonly associated with PIP, which is not surprising given that both diabetes and CVD often co-exist with CKD. Almost half of our patients were being treated for diabetes and CVD. A large proportion of oral antidiabetic agents are excreted by the kidney. It is likely that concomitant chronic diseases complicate and confound the treatment of each other. These medications are adjusted frequently in response to cardiac, renal or electrolyte disturbances, and are at a higher risk of being potentially inappropriate.
It is important to note that for some medications commonly prescribed in the study (for example, atenolol) clinical markers, such as heart rate and blood pressure, are often more important than kidney function in guiding dosage adjustment. Additionally, some medications do not have adequate published data in patients with renal impairment and may have been conservatively recommended as to be avoided rather than having an increased potential for toxic effects.
Our study also showed that PIP was common across all stages of CKD. Although stages 1, 2 and 3a CKD seldom impact on the prescribing of renally cleared medications, several common medications such as metformin, gliptins and SGLT2 inhibitors do have specific recommendations in patients with early stages of CKD. Previous studies in patients with advanced CKD have shown contrasting results. Some studies have reported that in patients with advanced CKD the likelihood of PIP is less due to more vigilant monitoring and nephrology consultations, while other studies have shown an increased number of problems.[24-26] Further research to understand the phenomenon of PIP in general practice is needed, possibly leading to the design of interventions to optimise medication prescribing in patients with CKD. Optimal use of current eHR systems to flag renally-impaired patients so that PIP is identified is likely to be part of an intervention.
The major strength of the current study is that it assessed prescriptions in primary care in a large cohort of patients. Previous Australian studies of inappropriate drug prescriptions in CKD have focussed on older people in the community and aged care settings, or when patients were admitted to the hospital. However, the study does have some limitations. Prescribers may not always have had pathology results available at the time of prescribing. However, we evaluated the extent of PIP for medications prescribed within 90 days of the availability of a kidney function assessment. We did not examine any adverse clinical outcomes associated with PIP. Additionally, prescribing may be appropriate at times with benefits outweighing the additional risks or if no safer alternatives are available.