Primary Care Prescriptions of Potentially Nephrotoxic Medications in Children with CKD

Visual Abstract

Abstract

Background and objectives

Pediatric CKD management focuses on limiting kidney injury, including avoiding nephrotoxic medications. Nephrotoxic medication prescription practices for children with CKD are unknown. Our objective was to determine the prevalence and rates of primary care prescriptions for potentially nephrotoxic medications in children with CKD versus without CKD.

Design, setting, participants, & measurements

We conducted a retrospective, matched population-based cohort study of patients aged <18 years, registered at a general practice participating in the UK Clinical Practice Research Datalink (CPRD) from 1997 to 2017. Children with a clinical code indicating an incident diagnosis of CKD were matched 1:4 to patients without CKD on CKD diagnosis date, sex, age, CPRD practice, and number of general practitioner visits in the year before cohort entry. We calculated the prevalence and the rate of potentially nephrotoxic medication prescriptions throughout the follow-up period in patients with versus without CKD. Primary analyses included the following medication classes: aminoglycosides, antivirals, nonsteroidal anti-inflammatory drugs, salicylates, proton pump inhibitors, and immunomodulators. Secondary analyses used an expanded nephrotoxicity definition that also included, among others, angiotensin-converting enzyme inhibitors and angiotensin receptor blockers. Adjusted prescription rates were calculated using multivariable binomial regression.

Results

From 1,535,816 eligible patients, we identified 1018 incident CKD and 4072 non-CKD matches (mean age, 9.8 years [range, 1.1–17.9 years]; 52% male; mean follow-up time, 3.3 years). Overall, 26% of patients with and 15% of patients without CKD were prescribed one or more potentially nephrotoxic medication during follow-up. The overall rate of nephrotoxic medication prescriptions was 71 (95% confidence interval [95% CI], 55 to 93) prescriptions per 100 person-years in patients with CKD and eight (95% CI, 7 to 9) prescriptions per 100 person-years in patients without CKD (adjusted rate ratio, 4.1; 95% CI, 2.7 to 6.1).

Conclusions

Potentially nephrotoxic medications are prescribed at high rates to children with CKD.

Background

CKD in adults is a strong risk factor for cardiovascular disease (1,2). Children with CKD have an elevated prevalence of hypertension and left ventricular hypertrophy (3). Furthermore, cardiovascular disease remains a leading cause of death in this population (4,5). Because children with CKD are at the beginning of their lifespan, reducing CKD progression may substantially alter their long-term risk for cardiovascular disease. One strategy for preventing CKD progression is limiting nephrotoxic medication exposure (6). Studies of outpatient nephrotoxic medication prescriptions in CKD are currently limited to adults, reporting prescription prevalences of 13%–70% (7−10). This raises concerns about whether children with CKD are also at risk for inappropriate nephrotoxic medication prescribing. Although many children with CKD are followed by nephrologists, general practitioners play an important role in their care (11). Population databases of general practitioners’ prescriptions will provide the information needed to understand the burden of nephrotoxic medication prescribing in children with CKD.

Our objective is to describe and compare the prevalence and rate of primary care prescriptions of potentially nephrotoxic medications in children with versus without an incident CKD diagnosis. We hypothesize that because of the under-recognition of the importance of kidney protection in children with kidney disease, children with CKD would be prescribed potentially nephrotoxic medications at a rate similar to matched controls without CKD.

Materials and Methods

Data Source

This is a retrospective, matched cohort study of children aged <18 years at the time of cohort entry, registered to the UK Clinical Practice Research Datalink (CPRD), with linkage to the Hospital Episode Statistics database. The CPRD contains anonymized primary care medical records providing information on patient demographics, diagnoses, and general practitioner prescriptions, capturing approximately 7% of the United Kingdom (UK) population (12). Health care information is recorded using Read Codes which represent, among others, diagnoses, signs, and symptoms of disease and test results (13). Drug data are obtained by automatic recording of general practitioner prescriptions coded according to the British National Formulary, the prescription reference for UK physicians (12). Mortality data are obtained via the Office of National Statistics (ONS) database, the UK’s national statistics institute (12). Data quality and consistency are subject to regular checks and validity has been demonstrated in adult populations (14−16). Hospital Episode Statistics is a database of hospital admissions linked to 75% of England-based CPRD practices (58% of all CPRD practices) (12,14).

We included patients registered in the CPRD between April 1997 and December 2017 to a practice that was linkable to the Hospital Episode Statistics database. At least 1 year of observation before cohort entry date was required for inclusion. Linkage with Hospital Episode Statistics allowed increased sensitivity of our CKD definition by including hospital diagnoses coded according to the International Classification of Diseases, Tenth Revision system. Approval for this study was granted by the CPRD Independent Scientific Advisory Committee (reference 17_190R) and the Research Ethics Board of the Jewish General Hospital.

Study Population

CKD was defined as first occurrence of either an outpatient or inpatient diagnostic code for kidney disease (including but not limited to CKD, cystic kidney disease, nephropathy, and renal dysplasia). Nephrectomy procedure codes were also included in the CKD definition. Outpatient dialysis codes were included if they appeared at least 3 months after a hospitalization where both an AKI and dialysis code had been recorded (to minimize exposure misclassification of children with resolving AKI but no future evidence of CKD codes). We excluded kidney transplant patients whose medication prescriptions were unlikely to be generalizable to the wider CKD population. Patients were classified as having CKD from date of first recorded code unless this code occurred during a hospitalization, in which case date of hospital discharge was used. Patients meeting CKD exposure criteria before having at least 1 year of CPRD history were excluded to restrict inclusion to incident CKD. The list of codes included in our operationalized definition of CKD was generated iteratively through a consensus process between two investigators, one with a high expertise in pediatric kidney disease (M.Z., C.L.) (Supplemental Table 1).

Cohort entry for patients with CKD was date of incident CKD diagnosis. Patients with CKD were matched 1:4 with those without CKD, on date of CKD diagnosis, sex, age (caliper: ±2 years), CPRD practice, and exact number of general practitioner visits in the 12 months before cohort entry (surrogate health care exposure measure). Patients without CKD inherited the cohort entry date from their CKD match. To create a comparison group free from kidney disease, we excluded patients from our unexposed group if they had any kidney-related diagnoses (Supplemental Table 2).

Follow-up was from cohort entry date to end of follow-up (maximum 5 years); end of the study period (December 2017); or censoring due to death (recorded in CPRD, Hospital Episode Statistics, or ONS), date of last data collection from the CPRD practice, or departure from the CPRD practice, whichever occurred first. We limited follow-up to 5 years to control for temporal trends of prescription patterns over time.

Outcome Definition

Our definition of a potentially nephrotoxic medication was mainly derived from three recent studies using robust methodologies to construct lists of medications recognized as nephrotoxic in adults or children (17−19). In addition to considering data from these studies, we assessed face validity of the included medications. We reached consensus about including additional drugs that were not on these lists but were deemed to have sufficient evidence supporting their potential nephrotoxicity. We refer to the medications included in our study as being “potentially” nephrotoxic because some may be indicated in CKD depending on the clinical context.

We developed two lists: category A and category B medications. These represent potentially nephrotoxic medications on the basis of the strength of consensus in the literature as to their nephrotoxic potential, with category A representing medications more consistently recognized as nephrotoxic (Supplemental Table 3). Because of their potential indication in CKD, we did not include angiotensin-converting enzyme (ACE) inhibitors in category A despite their recognized nephrotoxicity. Category B was broadened to provide as wide a capture of potential nephrotoxic medications as possible. This list included our category A medications plus medications with some evidence to support their nephrotoxic potential (see details in Supplemental Appendix 1).

Multiple drug prescriptions (including for the same drug or for a different drug within the same class) were considered as separate and cumulative. Medication refills were captured only if they were expressly represcribed by the physician.

Confounders

Confounders were established a priori after literature review, discussion, and investigator consensus on the basis of their association with CKD and their role as risk factors for receiving potentially nephrotoxic medications. Confounders included number of hospitalizations in the year before cohort entry and each of the following, measured at any time before cohort entry: premature birth, diagnosis of diabetes mellitus, hypertension, cancer, and either heart failure or heart surgery. Cancer diagnoses were limited to the most common pediatric cancer types: leukemia, lymphoma, central nervous system tumors, soft tissue sarcoma and neuroblastoma (20). Socioeconomic status was also included and estimated by Index of Multiple Deprivation quintile using the patient’s postal code (21).

Statistical Analyses

Ever-exposure to a potentially nephrotoxic medication was evaluated by comparing the proportion of patients with and without CKD who received one or more potentially nephrotoxic medication prescriptions during the follow-up period. Proportions of children receiving one or more potentially nephrotoxic medication prescriptions are also presented by class. We repeated analyses for category A and B medication definitions. These analyses did not account for differential follow-up time between exposure groups.

Prescription rates for category A and B medication prescriptions were calculated for both patients with and without CKD. These rates represent total number of medication prescriptions per person-time from cohort entry to end of follow-up. Multiple prescriptions, including those for the same drug or for different drugs within the same medication class, were counted individually for each patient to reflect nephrotoxic medication burden. The rates were estimated using binomial regression to account for data overdispersion (22). Rate ratios (RRs) are presented for the baseline matched analysis and for multivariable analyses adjusted for additional confounders not included as matching variables (described above). For increased robustness, adjusted analyses also include matching variables as covariates in the model (23,24). We calculated 95% confidence intervals (95% CIs) for RRs using robust SEMs. All RRs are reported for the overall follow up period.

Secondary analyses evaluated whether the proportion of patients receiving one or more potentially nephrotoxic medications varied by year since cohort entry (corresponding to year since CKD diagnosis for exposed patients). Secondary analyses also evaluated “potentially nephrotoxic medication load,” defined as the proportion of patients receiving one, two, three, ≥3, and five or more potentially nephrotoxic medications over the entire follow-up period, and by year since cohort entry. Sensitivity analyses were performed excluding ACE inhibitors and angiotensin receptor blockers from our category B list as these may be indicated in CKD. Data management and cohort matching was performed using SAS version 9.4. Analyses were performed using the COUNT package in R (version 1.1.423).

Results

Matching, CKD Diagnosis, and Population Characteristics

There were 1,535,816 patients registered to the CPRD between April 1997 and December 2017 who met our inclusion criteria (Figure 1). A total of 1019 were identified as having CKD (0.07% of total population). We identified four eligible matches for all but one patient with CKD who was excluded to create our final cohort of 5090 children (1018 with CKD, 4072 without CKD). There were six patients with missing socioeconomic data. Given this very low number, we used a complete-case approach in the regression analysis (see details in Supplemental Appendix 1).

Figure 1.

Flow diagram of patient inclusion from the UK Clinical Practice Research Datalink (CPRD) database.

Study population characteristics are described in Table 1. Mean age was 10 years (13 months to 17 years 11 months). A total of 52% of the cohort was male. Mean duration of follow up was 3.3 years. The CKD group had a higher proportion of patients with diabetes mellitus, hypertension, cancer, heart failure/surgery, and previous hospitalizations.

Table 1.

Characteristics of pediatric patients in the UK Clinical Practice Research Datalink with newly diagnosed CKD and matched controls without CKD

Characteristic	CKD	No CKD (Matched Controls)a
N	1018	4072
Male, n (%)	532 (52)	2128 (52)
Age in years, mean (SD)	10 (5)	10 (5)
Follow-up time
Median [IQR]	3.3 [1.5–5.0]	3.7 [1.8–5.0]
Mean (SD)	3.2 (1.8)	3.3 (1.7)
Categorization, n (%)
<1 yr of follow-up	170 (17)	552 (14)
1–2 yr of follow-up	149 (15)	563 (14)
2–3 yr of follow-up	152 (15)	590 (15)
3–4 yr of follow-up	104 (10)	455 (11)
4–5 yr of follow-up	76 (8)	371 (9)
Full 5 yr of follow-up	367 (36)	1541 (38)
General practitioner visits in year before cohort entry, n (%)
0	731 (72)	2952 (73)
1	137 (14)	654 (16)
>1	150 (15)	466 (11)
Hospitalization in year before cohort entry, n (%)	573 (56)	146 (4)
Premature birth, n (%)	45 (4)	115 (3)
Diabetes mellitus before cohort entry, n (%)	33 (3)	<5b
Hypertension before cohort entry, n (%)	33 (3)	<5
Cancer diagnosis before cohort entry, n (%)	12 (1)	<5
Heart failure or heart surgery before cohort entry, n (%)	20 (2)	9 (0.2)

IQR, interquartile range.

^aMatched on date of CKD diagnosis, sex, age±2 yr, general practice, and number of general practitioner visits in the 12 months before cohort entry.

^bTable values <5 are suppressed in accordance with UK Clinical Practice Research Datalink policy.

CKD Diagnosis Ascertainment

Of the 1018 participants with CKD, 351 were identified by outpatient and 551 were identified by inpatient diagnostic codes. A total of 106 patients were included after having undergone a nephrectomy, and 11 patients (1%) had an outpatient dialysis code recorded during follow-up. Further details regarding CKD inclusion codes are summarized in Supplemental Table 4.

Prescription of at Least One Potentially Nephrotoxic Medication during the Study Period

Overall, 26% of participants with CKD and 15% of participants without CKD were prescribed at least one category A medication during follow-up (Figure 2). Nonsteroidal anti-inflammatory drugs (NSAIDs) accounted for most category A medication prescriptions and were prescribed at least once to 17% of patients with CKD and 13% of patients without CKD.

Figure 2.

Proportion of patients with versus without CKD who received a category A medication, by medication class. Numbers next to figure bars represent number of participants receiving a potentially nephrotoxic medication. Values <5 are suppressed in accordance with CPRD policy. NSAIDs, nonsteroidal anti-inflammatory drugs; PPI, proton pump inhibitors.

When considering category B medications (which include category A medications), 71% of patients with CKD and 50% of patients without CKD received at least one potentially nephrotoxic medication during follow-up (Figure 3). When excluding ACE inhibitors and angiotensin receptor blockers, these proportions were 68% and 50%, respectively. No patient without CKD received an ACE inhibitor compared with 10% of patients with CKD. The two most commonly prescribed drug classes were penicillins and cephalosporins, prescribed to 57% and 18% of patients with CKD, respectively, compared with 44% and 4% of patients without CKD.

Figure 3.

Proportion of patients with versus without CKD who received a category A or category B medication, by medication class. Numbers next to figure bars represent number of participants receiving a potentially nephrotoxic medication. Values <5 are suppressed in accordance with CPRD policy. ACE-I, angiotensin-converting enzyme inhibitors; ARB, angiotensin receptor blockers.

CKD versus Non-CKD Potentially Nephrotoxic Medication Prescription Rate

The rate of potentially nephrotoxic medication prescriptions for category A medications was 71 (95% CI, 55 to 93) prescriptions per 100 person-years in patients with CKD and eight (95% CI, 7 to 9) prescriptions per 100 person-years in patients without CKD (adjusted RR, 4.1; 95% CI, 2.7 to 6.1) (Table 2). When considering category B medications, the rate of prescriptions was 278 (95% CI, 244 to 316) prescriptions per 100 person-years in patients with CKD and 44 (95% CI, 41 to 47) prescriptions per 100 person-years in patients without CKD (adjusted RR, 4.0; 95% CI, 3.3 to 4.9). When excluding ACE inhibitors and angiotensin receptor blockers, the adjusted RR for category B medication prescriptions in patients with versus without CKD was 2.7 (95% CI, 2.2 to 3.2) (Table 2).

Table 2.

Rate of prescription of potentially nephrotoxic medications over the entire study period in children with versus without CKD

Prescribed Medications	Patients with CKD		Matched Control Patients without CKDa
	No. of Prescriptions	Prescription Rate (per 100 Person-Years)	No. of Prescriptions	Prescription Rate (per 100 Person-Years)	Rate Ratio (95% CI)	Adjusted Rate Ratio (95% CI)b
Category A nephrotoxic medicationsc	2191	71 (55–93)	1002	8 (7–9)	9.2 (6.9 to 12.2)	4.1 (2.7 to 6.1)
Category B nephrotoxic medicationsd	8751	278 (244–316)	6010	44 (41–47)	6.4 (5.5 to 7.3)	4.0 (3.3 to 4.9)
Category B nephrotoxic medications excluding ACE-Is and ARBs	6234	198 (171–228)	6010	44 (41–47)	4.5(3.8 to 5.3)	2.7 (2.2 to 3.2)

95% CI, 95% confidence interval; ACE-Is, angiotensin-converting enzyme inhibitors; ARBs, angiotensin receptor blockers.

^aMatched on date of CKD diagnosis, sex, age±2 yr, general practice, and number of general practitioner visits in the 12 months before cohort entry.

^bAdjusted further for sex, age, Index of Multiple Deprivation quintile, general practice region, number of hospitalizations in the 12 months before cohort entry, prematurity, diabetes, hypertension, cancer, and either history of heart surgery or heart failure.

^cCategory A medications: aminoglycosides, nonsteroidal anti-inflammatory drugs, antivirals, salicylates, proton pump inhibitors, antibiotics, immunomodulators, lithium, and pamidronate.

^dCategory B medications: ACE-Is, ARBs, cephalosporins, other antibiotics, penicillins, immunomodulators, antiepileptics, antivirals, contrast agents, and other.

Secondary Analyses

When considering time since cohort entry, the proportion of patients with CKD receiving at least one category A medication remained relatively stable, ranging from 11% to 14% per year (Supplemental Figure 1). When considering the number of potentially nephrotoxic medications prescribed, the proportion of patients with CKD receiving one, two, three, ≥3, or five or more individual prescriptions did not vary with time since cohort entry (or time since diagnosis for patients with CKD) (Supplemental Figure 1).

Discussion

Our study compares the prevalence and rate of primary care prescriptions of potentially nephrotoxic medications to children with versus without an incident diagnosis of CKD. We found that in an outpatient, UK population-based data source, children with CKD were prescribed potentially nephrotoxic medications by general practitioners at four times the rate of matched children without CKD. The proportion of patients with CKD who were prescribed at least one category A and at least one category B medication during the study period was 26% and 71%, respectively. Our results have important implications because potentially nephrotoxic medications may contribute to kidney disease progression, which is associated with higher rates of mortality, ESKD, and morbidity from cardiovascular disease.

Previous studies on nephrotoxic medication prescriptions in patients with CKD have focused on adults. A comprehensive, population-based study revealed that 45% of patients with CKD received a nephrotoxic medication within 1 year of their CKD diagnosis and 56% received a prescription for an NSAID over their entire follow-up (19). Considering the lower medication burden in pediatrics, our corresponding proportions of 14% and 17% are similarly worrisome.

Children with CKD generally have more comorbidities than their healthy counterparts, occasionally justifying the use of potentially nephrotoxic medications. Although our models were adjusted for major comorbidities associated with CKD, residual confounding might explain part of the difference in prescription rates. It is also important to recognize that in the face of an opioid epidemic, the avoidance of medications such as NSAIDs may have important implications for individuals with pain in whom alternative medications, although safe from a kidney point of view, may carry other important risks. In the absence of data on kidney disease progression with the medications studied, NSAID avoidance should not be overstated and further studies on the clinical effect of NSAID versus opioid use in patients with CKD will be important.

Although we did not compare prescriptions before and after CKD diagnosis, we showed that yearly prescription rates did not differ with each passing year after CKD diagnosis, suggesting that the diagnosis may not have been considered by prescribing physicians (Supplemental Figure 1, Supplemental Table 5). Strikingly, this remained true for patients receiving three or more potentially nephrotoxic medications, which is important considering the association between multiple nephrotoxic prescriptions and AKI in pediatrics. Of note, many of our inclusion codes represented diseases strongly associated with CKD, although not explicitly coded as such (e.g., cystic kidney disease, small kidneys, obstructive and reflux nephropathy, etc.). These diagnoses may not be recognized by physicians as CKD but have been associated with a significantly increased risk of adult ESKD making nephrotoxic medication avoidance in these cases equally important (25).

Our study has several strengths. It is the first population-based pediatric study to evaluate potentially nephrotoxic medication prescriptions in children with CKD. The CPRD allowed us to follow a large pediatric CKD population. Inclusion of outpatient and inpatient procedure and diagnostic codes allowed for a greater sensitivity for our CKD definition and our matched design allowed for strong control of potentially confounding variables such as age, sex, and prior health care contact. Matching on previous number of general practitioner visits helped control for frequency of health care contact and the opportunity to receive a prescription which might otherwise be different between patients with and without CKD. Furthermore, matching on general practice allowed us to control for prescriber variability which can account for important differences when comparing patients with and without CKD.

Our study also has some limitations. CPRD contains limited information on medication indication. It is possible that some prescriptions for nephrotoxic medications were indicated. For example, ACE inhibitors may be indicated in CKD and immunomodulators may be indicated in autoimmune glomerular disease, which should be considered when interpreting the higher prescription rates for these medications in patients with versus without CKD. Furthermore, CPRD does not include information on hospital nor specialist prescriptions, which may have led to underestimating potentially nephrotoxic medication prescriptions in our study. Some underestimation of NSAID use is also likely as these medications are available over the counter in the UK. However, the main interest of this paper was to evaluate general practitioner prescribing behaviors, and CPRD provides excellent capture of this outcome. Also, the absence of staging information for >90% of the patients with CKD in our study did not allow us to compare prescription rates by CKD stage nor follow CKD stage progression over the study period.

Our CKD definition relied on clinical codes, which have modest sensitivity in adults when compared with eGFR (26,27). However, in children, sensitivity for CKD stages 3–5 has been reported to be as high as 75% in a UK primary care database similar to CPRD (28). Limited coding sensitivity could lead to misclassification of some patients with CKD as non-CKD, resulting in an underestimation of the relative difference in our prescription rate. Our results would therefore represent conservative estimates. We mitigated this misclassification by excluding children with kidney diagnoses not considered sufficient evidence of CKD (Supplemental Table 2). To offset modest sensitivity of CKD codes and given the paucity of code validity data in pediatric kidney disease, we derived a broad code list via expert opinion incorporating both inpatient and outpatient codes within a database recognized for its high coding validity overall (14,26,27). For example, we included codes representing kidney conditions that define stage 1 CKD independently of decreased eGFR. This was important as relying only on eGFR in children may underestimate CKD presence (and children at risk for CKD progression, in whom avoiding nephrotoxic medication may be desirable) by up to 50% (29).

Potentially nephrotoxic medications appear to be prescribed at high rates to pediatric patients with CKD. Although their use may sometimes be justified, there is an apparent need for increased awareness of their harmful potential in this high-risk patient group. Further research regarding nephrotoxic medication prescription practices in pediatric kidney disease patients could focus on determining the appropriateness of these prescriptions and identifying factors contributing to elevated prescription rates. This could eventually help direct clinical decision support systems and physician education programs to reduce inappropriate medication prescribing in pediatric CKD.

Disclosures

Dr. Filion reports receiving a grant from Canadian Institutes of Health Research (CIHR). He receives personal fees from Institut national d'excellence en santé et en services sociaux and from the Canadian Network for Observational Drug Effect Studies (CNODES), a collaborating centre of the Drug Safety and Effectiveness Network (DSEN) funded by the Canadian Institutes of Health Research (Grant Number DSE-146021), all outside of the submitted work. Dr. Platt reports receiving consulting fees from Amgen, Biogen, Eli Lilly, and Merck; an Advisory Board position at Biogen; and a position on a Study Steering Committee at Pfizer. Dr. Platt also holds the Albert Boehringer I Chair in Pharmacoepidemiology at McGill University. None of these disclosures are related to products relevant to the current research. Dr. Zappitelli reports receiving an honorarium from Baxter for giving an educational session to intensive care nurses on continuous kidney replacement therapies in 2018. Dr. Lefebvre and Dr. Reynier have nothing to disclose.

Funding

Dr. Filion is the recipient of a William Dawson Scholar award from McGill University and a Junior II salary support award from the Fonds de recherche du Québec - santé (Quebec Foundation for Health Research, FRQS).Dr. Lefebvre was supported by a Canadian Institutes of Health Research Frederick Banting and Charles Best Canada Master’s Scholarship (R246524C0G), funded through the McGill University Research Bursary Program, as well as a Master’s bursary from the Fonds de recherche du Québec - santé in partnership with Fondation des Étoiles.

Supplemental Material

This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.03550319/-/DCSupplemental.

Supplemental Table 1. Codes defining CKD for study inclusion.

Supplemental Table 2. Nonspecific renal codes defining exclusion from the unexposed cohort.

Supplemental Table 3. List of category A and category B potentially nephrotoxic medications included in the study.

Supplemental Table 4. CKD diagnosis and procedure codes of included study patients.

Supplemental Table 5. Detailed proportions for Figures 2 and 3.

Supplemental Figure 1. Trends in category A medication prescribing by year since cohort entry.

Supplemental Appendix 1. Supplemental methods and results corresponding to sections within the manuscript.