The Cost Utility of Nonpregnancy Laboratory Monitoring for Persons on Isotretinoin Acne Therapy

Abstract

We sought to project the 1-year cost utility of nonpregnancy laboratory monitoring cessation among patients initiating isotretinoin. We conducted a model-based cost utility analysis comparing (i) current practice (CP) and (ii) cessation of nonpregnancy laboratory monitoring. Simulated 20-year-old persons initiating isotretinoin were maintained on therapy for 6 months, unless taken off because of laboratory abnormalities in CP. Model inputs included probabilities of cell-line abnormalities (0.12%/wk), early cessation of isotretinoin therapy after detection of an abnormal laboratory result (2.2%/wk, CP only), quality-adjusted life-years (0.84–0.93), and laboratory monitory costs ($5/wk). We collected adverse events, deaths, and quality-adjusted life-years and costs (2020 USD) from a health care payer perspective. Over 1 year, and for 200,000 people on isotretinoin in the United States, the CP strategy resulted in 184,730 quality-adjusted life-years (0.9236 per person), and nonpregnancy laboratory monitoring resulted in 184,770 quality-adjusted life-years (0.9238 per person). The CP and nonpregnancy laboratory monitoring strategies resulted in 0.08 and 0.09 isotretinoin-related deaths, respectively. Nonpregnancy laboratory monitoring was the dominating strategy, realizing $24 million savings per year. No variation of a single parameter across its plausible range changed our cost utility findings. Cessation of laboratory monitoring could realize savings of $24 million per year for the US health care system and improve patient outcomes, with negligible effects on adverse events.

Introduction

Isotretinoin is a highly effective therapy for severe nodulocystic acne and an estimated 200,000 US adult patients per year currently receive isotretinoin (Barbieri et al., 2020; Lee et al., 2016; Leyden et al., 2014). As a known teratogen, patients on isotretinoin therapy receive frequent mandatory reporting on pregnancy and contraception practices (Tkachenko et al., 2019). The medication has also been associated with several laboratory abnormalities after initiation, and the label recommends baseline and regular laboratory monitoring of triglycerides and liver function testing, with the goal of detecting abnormalities portending more serious adverse effects and cessation of therapy to prevent such events.

However, while laboratory abnormalities are relatively common in patients initiating isotretinoin therapy, adverse events, primarily pancreatitis and hepatitis, are rare (Lee et al., 2016; Opel et al., 2017). Although the exact rate of pancreatitis and hepatitis due to isotretinoin are unknown, a recent systematic review identified 25 reported cases of pancreatitis associated with isotretinoin since its approval in 1982 (Opel et al., 2017). Furthermore, providers often do not make therapeutic changes in response to abnormal laboratory test results (Barbieri et al., 2020). As such, several guideline organizations and stakeholders advocate for reduced laboratory monitoring practices (although with continued pregnancy testing), especially for low-risk patients (Affleck et al., 2022; Altman et al., 2002; Xia et al., 2022). Indeed, laboratory monitoring for patients on isotretinoin has been associated with up to $17 million per year in unnecessary health care costs (Barbieri et al., 2020).

Although cessation of laboratory monitoring for patients on isotretinoin is known to reduce health expenditures, the expected trade-offs in terms of adverse events, quality-adjusted life-years (QALYs), deaths, and total costs of care remain unknown. The objective of this study was to project the 1-year clinical and economic outcomes, and uncertainties therein, of current practice (CP) of laboratory monitoring compared with cessation of nonpregnancy laboratory monitoring (NoLabMon) for patients with severe nodulocystic acne initiating isotretinoin therapy in the United States.

Results

Base case results

For a representative cohort of 200,000 young adults currently on isotretinoin therapy for acne, NoLabMon had higher person-weeks in isotretinoin treatment health states, with an estimated increase of 14,900 person-weeks in the high risk on isotretinoin state, and a reduction in 33,100 person-weeks in the off-isotretinoin health state, compared with the CP (Table 1). However, NoLabMon incurred small numbers of increased serious adverse events and deaths, with basecase inputs resulting in 0.3 increased adverse events and 0.02 increased deaths over 1 year. That said, the number of people who stopped isotretinoin therapy early per adverse event averted in the CP strategy was 5,000, and the number of early cessations per death averted was 100,300.

Table 1.

Total Person-Weeks in Each Health State at 1 Year (n = 200,000)

Strategy	Health State
	Low risk on isotretinoin	High risk on isotretinoin	Off isotretinoin	Serious adverse event	Death
CP	4,712,600	72,500	33,180	1.6	1.6
NoLabMon	4,712,600	87,400	40	1.9	1.9
Difference	-	14,900	(33,100)1	0.3	0.3

Abbreviations: CP, current practice laboratory monitoring; NoLabMon, no laboratory monitoring.

¹Parentheses indicate negative differences, that is, NoLabMon had less person-time in a health state or lower costs.

Over 1 year and for a cohort of 200,000 people, total QALYs amounted to 184,730 in CP and 184,770 in the NoLabMon strategies, an increase of 40 QALYs under NoLabMon (Table 2). The per-person QALYS were 0.9236 for CP and 0.9238 for NoLabMon. Total undiscounted costs over 1 year for CP were $819,129,000 and were $24,124,000 less for NoLabMon (total for NoLabMon = $795,005,000). NoLabMon was the dominating strategy, as it was both more effective and less expensive than CP. NoLabMon was estimated to save $100 per person on isotretinoin over the 1-year period.

Table 2.

Clinical, Costs, and Cost-Utility Results

Strategy	QALYs	QALYs/Person	Costs, $	Costs/Person	ICUR
CP	184,730	0.9236	819,129,000	4,100	-
NoLabMon	184,770	0.9238	795,005,000	4,000	Cost-saving
Difference	40 (95% CI: 39.5–40.5)2	0.0002 (95% CI: 0.0001–0.0003)2	(24,124,000)1 (95% CI: 24,081,000–24,167,000)2	(100)1 (95% CI: 95–105) 2

Abbreviations: CI, confidence interval; CP: current practice laboratory monitoring; ICUR, incremental cost-utility ratio; NoLabMon, no laboratory monitoring; QALY, quality-adjusted life-year.

¹Parentheses indicate negative differences, that is, NoLabMon had less person-time in a health state or lower costs.

²The presented 95% CIs are not analogous to “statistical significance” in an observational study or randomized trial. In the context of a decision model, 95% CIs provide an estimate of the uncertainty introduced into the model by imprecision in the estimates of the various parameters but cannot convey the uncertainty introduced by issues such as potential bias from observational study designs, applicability of data sources to the specific population being modeled, the inherent assumption within any model of independence of the parameters unless data are available to establish correlations, or model structural assumptions.

Deterministic sensitivity analysis

We performed univariate sensitivity analysis on key outcomes and report their effects on findings of economic efficiency of NoLabMon in Figure 1. Given the cost savings findings and negative incremental cost-utility ratio, we converted findings into net monetary benefit (NMBs) calculated as follows: (willingness to pay [WTP] × difference in efficacy) − difference in costs, assuming a WTP of $100,000. The basecase NMB for NoLabMon compared with CP was $142 per person. In every change across the model input parameters’ plausible ranges, the NMB of NoLabMon was >0, indicating it was the optimal use of resources under all conditions explored. When the difference in utility values between treated and untreated acne health states was changed to 0.02, the QALY benefit of NoLabMon compared with CP was reduced to 10 (from 40), and costs remained unchanged. We projected QALYs over 55 years, incorporating the difference in deaths between strategies, and NoLabMon remained more effective. Given the rarity of the event, our findings were not sensitive to changes in the disutility of acute pancreatitis, including increasing the disutility of pancreatitis from 0.5 to 1.0. We also included a sensitivity analysis where we assumed no complete blood cell count testing in the monitoring costs, which reduced the cost savings from $24 million to $20 million.

Figure 1.

Tornado diagram of univariate sensitivity analyses. This tornado diagram shows the influence of individual model input parameters, varied across their plausible ranges, on the net monetary benefit of the NoLabMon strategy. The net monetary benefit is calculated as the difference in the clinic effectiveness between the two strategies multiplied by the willingness to pay, and then less the difference in cost between the two strategies. Net monetary benefits greater than 0 indicate a cost-effective strategy at the assumed willingness to pay. Ranges assessed are presented as (base-case value: value that yields lowest ICUR − value that yields highest ICUR). ICUR, incremental cost-utility ratio; NMB, net monetary benefit.

We combined the three most influential parameters to create best- and worst-case scenarios for the NMB of NoLabMon. In the best-case scenario, we varied the cost of laboratory monitoring to $279/6 months, the utility of successful isotretinoin treatment to 0.97, and the probability of transitioning from low risk to high risk to 0.00225, and the NMB for NoLabMon rose to $319/person. In the worst-case scenario, we varied the cost of laboratory monitoring to $70/6 months, the utility of successful isotretinoin treatment to 0.89, and the probability of transitioning from low risk to high risk to 0.00005, and the NMB for NoLabMon decreased to $129 per person. Even in the worst-case scenario, NoLabMon remained the most economically efficient use of resources given a WTP of $100,000/QALY.

Probabilistic uncertainty analysis

From 10,000 simulations, Figure 2 shows that >99% were in cost saving, demonstrating that NoLabMon had lower costs and higher effectiveness than CP in nearly all simulations. There were two simulations in which NoLabMon yielded lower effectiveness (QALYs) than CP, but with lower costs. In no simulations did NoLabMon have a higher cost than CP.

Figure 2.

Probabilistic uncertainty analysis incremental cost-utility ratio plane. This figure shows the incremental costs (in USD) on the y-axis and incremental effectiveness (in QALYs) on the x-axis of NoLabMon compared with CP for 10,000 model simulations. Each simulation independently drew values for model inputs parameters from defined distributions. Negative incremental costs indicate that the no laboratory monitoring strategy is less expensive than CP. Dividing the incremental costs by incremental QALYs yields the ICUR. The results are presented for a hypothetical cohort of 200,000 patients over 1 year. CP, current practice laboratory monitoring; ICUR, incremental cost-utility ratio; NoLabMon, no laboratory monitoring; QALY, quality-adjusted life-year; USD, US dollars.

Discussion

In this simulation modeling study, cessation of laboratory monitoring for 200,000 US patients on isotretinoin resulted in increased QALYs and savings of ∼$24 million per year for the health care system. Although there were very small increases in adverse events and deaths for patients without laboratory monitoring, the large gain in QALYs from remaining on effective acne therapy far outweighed these probabilities. In addition, as only two deaths associated with isotretinoin therapy have been reported, and these were in patients with late-stage cancer, these probabilities may indeed be zero.

We explored the sensitivity of our results to plausible variations in input parameters through uncertainty analysis. In univariate uncertainty analysis, the cost of laboratory monitoring was shown to be the most influential parameter on the NMB for the NoLabMon strategy. In probabilistic uncertainty analysis, we showed that given our model assumptions and the uncertainty within the input data, NoLabMon is very likely to be a cost-saving strategy.

Our study provides a key contribution to the existing literature on laboratory monitoring for isotretinoin therapy by explicitly weighing the trade-offs between preventative early isotretinoin cessation and continuation of effective therapy. Understandably, dermatologists exercise caution when prescribing a quality-of-life–improving drug that has rare but serious potential side effects. Our simulation shows that these events are rare, and the population-level benefit to quality of life from receiving optimal acne therapy far outweigh the risks.

That said, in light of this evidence, some providers and patients may choose to err on the side of absolute caution when initiating isotretinoin therapy. Some recommend alternative laboratory monitoring strategies such as only implementing routine laboratory monitoring for patients with observed baseline abnormalities or reducing the frequency of routine laboratory monitoring. Indeed, recently published guidelines include checking triglycerides and alanine transaminase at baseline and peak dose, which would be a good strategy to consider. However, we were unable to consider this strategy principally owing to lack of data on the temporal relationship between laboratory abnormalities and adverse events. Current research has uncovered population-level trends in laboratory abnormalities in patients on isotretinoin, but to our knowledge, these have not yet extended to adverse events (Lee et al., 2016). Importantly, some reports show that the majority of adverse events on isotretinoin may not have laboratory abnormality precursors, which suggests adverse events may not be avoided by routine screening (Opel et al., 2017). In any case, our simulations show that any reduction in laboratory monitoring frequency from current varied practices is likely to save money, with negligible additional risk to patients.

We recognize that this study is not the first to call for reduced or stopped laboratory monitoring for patients on isotretinoin acne therapy (Shah and Kroshinsky, 2021). De-implementation strategies will prove critical in health care systems choosing to reduce, or stop, routine laboratory monitoring (Prasad and Ioannidis, 2014; Rietbergen et al., 2020). Future research might evaluate specific strategies, such as identification of clinical champions, trainings, and quality metrics, that successfully alter isotretinoin laboratory monitoring practices.

This analysis has several limitations. First, our model assumed that all adverse events were preceded by a laboratory abnormality, which may overestimate the value of laboratory monitoring. In this assumption, and all assumptions, we biased against cessation of laboratory monitoring whenever possible to ensure conservative estimates. Second, we simulated a CP strategy that aggregated across varied laboratory monitoring strategies (weekly, monthly, etc.) as opposed to individual laboratory monitoring strategies. This was due to our primary data source reporting dermatologic practices in aggregate, as opposed to by laboratory monitoring strategy (Barbieri et al., 2020). Third, we made several assumptions calculating the probability of experiencing a serious adverse event when on isotretinoin therapy, in particular using case studies to estimate event rates that may be an underestimate owing to underreporting. The true probability is currently unknown, and we tested our assumptions extensively in sensitivity analysis. However, we included all literature reports worldwide, despite only estimating for a US treatment cohort, again likely overestimating laboratory monitoring utility. Fourth, the model does not account for patients with mild laboratory abnormalities that might cease isotretinoin therapy and lose the quality-of-life benefits of acne therapy without any risk of an adverse event. This exclusion is conservative in that it makes no laboratory monitoring look worse. Fifth, it is difficult to accurately assess the frequency and QALY impact of rare adverse events, and the QALY estimates used in this study may not apply to the younger and healthier population typically prescribed isotretinoin. We assessed the effects of an extreme disutility effect of −1.0 for pancreatitis in a sensitivity analysis. Finally, we did not incorporate any societal costs, and decrements in quality of life for routine laboratory monitoring or any incidental findings, which may underestimate the societal benefits of laboratory monitoring cessation.

In conclusion, our results indicate that cessation of regular laboratory monitoring of young patients on isotretinoin acne therapy would likely be an efficient use of resources in the United States and yield significant cost savings for the health care system.

Materials and Methods

Analytic overview

We used a Markov model to compare (i) current practice laboratory monitoring and isotretinoin treatment cessation owing to laboratory abnormalities (CP) and (ii) cessation of complete blood count, lipid profile, and liver function laboratory monitoring (NoLabMon), where no preventative treatment discontinuation occurs. We simulated a cohort of 20-year-old persons initiating isotretinoin therapy, and maintaining therapy for 6 months, unless taken off early owing to observed laboratory abnormalities in the CP strategy. We informed model transition probabilities, health state utilities, and costs from the published literature. Over 1 year, we estimated the total person-weeks in each health state by strategy, total QALYs, undiscounted costs, NMB, and incremental cost-utility ratios. We conducted the cost-utility analysis from a health care payer perspective, and we assumed a US-specific WTP of $100,000 per QALYs (Neumann et al., 2014; Vanness et al., 2021). Throughout the analysis, and highlighted in Materials and Methods, we biased against NoLabMon in an attempt to maximize patient safety and welfare. We followed the US Second Panel on Cost-Effectiveness in Health and Medicine recommendations and followed the Consolidated Health Economic Evaluation Reporting (CHEERS) guidelines (Husereau et al., 2013; Sanders et al., 2016). All data used in this study were previously published and did not require institutional review board approval.

Model overview

We developed a Markov cohort simulation model with a weekly time step over a 1-year time horizon, implemented in Microsoft Excel (Redmond, WA). We chose a 1-week cycle to capture the probability of adverse events and potential for detection with laboratory monitoring. Model health states are dependent on risk level, isotretinoin therapy use, the occurrence of a serious adverse event, and death due to an adverse event (Figure 3). Simulated patients begin the model either high risk on isotretinoin, or low risk on isotretinoin, informed by the literature (Barbieri et al., 2020). Simulated patients who are low risk on isotretinoin experience monthly probabilities of remaining on therapy or becoming high risk owing to a cell-line abnormality. Patients who are high risk experience monthly probabilities of remaining high risk on therapy, transitioning off of isotretinoin therapy before course completion due to abnormal laboratory test identification, or experiencing a serious adverse event. The CP laboratory monitoring strategy has the potential to uncover high-risk status and then transition simulated patients to early cessation of isotretinoin therapy.

Figure 3.

Decision model health state diagram. This figure depicts the health states in the Markov cohort model of isotretinoin therapy. Circles represent health states, and the arrows between circles indicate either retention in the health state or transitions to other health states. Simulated patients begin in the low risk, on isotretinoin or high-risk, on isotretinoin health states. Throughout the simulation, patients experience weekly transition probabilities, informed by the literature, of becoming high risk, having early treatment cessation owing to identification of a laboratory abnormality, serious adverse event, death, or successful completion of a course of isotretinoin therapy. Each week, the model sums the number of patients in each health state and accrues utilities and costs specific to each state, as described in the Materials and Methods. We conduct the simulation twice, where in the NoLabMon strategy the transition probability between high-risk, on isotretinoin and off isotretinoin is 0, to represent no active laboratory monitoring and preventative treatment cessation. The difference between the effectiveness (measured in quality-adjusted life-years) and costs between the two strategies are used to calculate the base-case incremental cost-utility ratio.

Serious adverse events are assumed to last 1 cycle (1 week) and result in either death or cessation of isotretinoin therapy for the remainder of the simulation. Given the young cohort, we assume no background mortality occurs, and mortality in this 1-year time horizon occurs only after a serious adverse event. After 6 months, all high- and low-risk patients, in both strategies, still on isotretinoin are assumed to complete the treatment course and transition to the completed course health state where they remain for the remainder of the simulation.

Intervention

We incorporated current practices of isotretinoin laboratory monitoring cessation (CP) compared with a hypothetical scenario of no laboratory monitoring (NoLabMon). CP laboratory monitoring strategies are varied and include baseline testing, with weekly to 6 weekly testing for high-risk patients. The data that inform the modeled CP strategy aggregate across current practice patterns (Barbieri et al., 2020). NoLabMon consisted of cessation of laboratory monitoring for patients on isotretinoin, with identical input data (i.e., risks of adverse events and deaths) except without early cessation of isotretinoin therapy due to identification of laboratory abnormalities and no accumulation of laboratory monitoring costs.

Model input data

Cohort characteristics

We simulated a cohort of 20-year-old patients initiating isotretinoin. Results are scaled to the estimated current number of patients on isotretinoin (n = 200,000; this scaling factor does not impact relative findings, only absolute findings) (Barbieri et al., 2020; Lee et al., 2016; Leyden et al., 2014). At baseline, 99.6% of patients are without laboratory abnormalities and 0.4% have grade 3 or higher elevated triglycerides or hepatic function markers (i.e., high-risk; Table 3) (Barbieri et al., 2020).

Table 3.

Model Input Data

Variable	Value	Reference
Age, y	20	Assumed
Proportion high-risk at baseline, %	0.4	(Barbieri et al., 2020)
Health state transitions
Probability of laboratory abnormality, weekly (95% CI)1	0.001 (0.00005–0.002)	(Barbieri et al., 2020)
Probability of therapy cessation after observed abnormality, weekly (95% CI)1	0.022 (0.006–0.049)
Probability of a serious adverse event if high risk, weekly per 100 people	0.002	Literature Review, see Materials and Methods
Probability of death after a serious adverse event	0.05	(Banks et al., 2006)
Health state utility values
Utility on isotretinoin after 2 months of therapy, yearly	0.93 (0.89–0.97)	(Klassen et al., 2000)
Utility for early treatment cessation, yearly	0.89
Utility for completed therapy, yearly	0.93 (0.89–0.97)
Utility for adverse event, yearly	0.5	(Matza et al., 2020)
Costs, 2020 USD	Value	Reference
Isotretinoin, monthly	$425
Current practice laboratory monitoring, weekly	$5	(Barbieri et al., 2020)
Specialist visit cost	$216
Adverse event cost	$14,000	(Fagenholz et al., 2007)

Abbreviations: CI, confidence interval; USD, US dollars.

¹95% CIs were calculated from proportions of patients experiencing each event over 6 months, adjusted to attain weekly proportions and CIs.

Heath state transitions

We estimated the weekly probability of transitioning from low risk to high risk on isotretinoin therapy by averaging the number of grade 3 or higher abnormalities in triglycerides, total cholesterol, and aspartate aminotransferase/alanine aminotransferase reported in a study of laboratory abnormalities over 6 months in a nationally representative database (Table 3) (Barbieri et al., 2020).

We conducted a literature search on 13 May 2021 to estimate the weekly probability of a serious adverse event, defined as acute pancreatitis or fulminant hepatitis (see Table 4 for the search strategy). Our PubMed search yielded 46 articles, of which 20 were excluded at the title/abstract level. We retrieved 26 full-text articles, of which 15 reported isotretinoin-related pancreatitis or hepatitis. After removing duplicate studies included in the Opel (2017) systematic review of isotretinoin-related pancreatitis, three unique studies reporting pancreatitis and one unique study reporting hepatitis remained (Ashraf, 2020; Atiq et al., 2019; Lin et al., 1999; Opel et al., 2017; Tejedor Tejada et al., 2019).

Table 4.

Literature Search for Isotretinoin-Associated Pancreatitis and Hepatitis

Search Number	Query	Search Details	Results
4	#1 AND (#2 OR #3)	("isotretinoin"[MeSH Terms] OR "isotretinoin"[All Fields]) AND ("pancreas"[MeSH Terms] OR "pancreas"[All Fields] OR "pancreatic"[All Fields] OR "pancreatitides"[All Fields] OR "pancreatitis"[MeSH Terms] OR "pancreatitis"[All Fields] OR ("liver failure"[MeSH Terms] OR ("liver"[All Fields] AND "failure"[All Fields]) OR "liver failure"[All Fields]))	46
3	liver failure	"liver failure"[MeSH Terms] OR ("liver"[All Fields] AND "failure"[All Fields]) OR "liver failure"[All Fields]	75,149
2	pancreatitis	"pancreas"[MeSH Terms] OR "pancreas"[All Fields] OR "pancreatic"[All Fields] OR "pancreatitides"[All Fields] OR "pancreatitis"[MeSH Terms] OR "pancreatitis"[All Fields]	369,122
1	isotretinoin	"isotretinoin"[MeSH Terms] OR "isotretinoin"[All Fields]	4,772

For pancreatitis, we updated a systematic review that reported 25 isotretinoin-related cases of pancreatitis with 3 additional cases of pancreatitis reported since its publication (Ashraf, 2020; Atiq et al., 2019; Opel et al., 2017; Tejedor Tejada et al., 2019). We additionally included 1 case report of toxic hepatitis to yield 29 total adverse events reported in the literature (Lin et al., 1999). We then calculated the weekly event rate of adverse events (pancreatitis and hepatitis) for high-risk patients by estimating person-time of high-risk isotretinoin use in the United States since 1982 (Barbieri et al., 2020; Leyden et al., 2014; Wysowski et al., 2002). In our event rate, we included cases reported outside of the United States (even though the person-time of exposure was limited to the United States) to estimate higher adverse events rates and bias the analysis against the NoLabMon strategy.

There were two reports of mortalities attributed to isotretinoin identified in a recent systematic review, both in patients on chemotherapy for glioblastoma and small cell lung cancer (Opel et al., 2017). To remain conservative and bias against NoLabMon, we incorporated a probability of mortality after an episode of pancreatitis of 5% (range: 2–9%) (Banks et al., 2006).

Health state utilities

Utilities for untreated acne vulgaris (0.84) and for postisotretinoin therapy (0.93) health states were from the published literature (Klassen et al., 2000). We assumed that isotretinoin therapy imparted no benefit for the first 2 months of therapy, and full benefit after 2 months if the full 6-month course is completed. We assumed simulated patients removed from treatment early at any time owing to identified laboratory abnormalities received 50% of the full benefit of isotretinoin therapy (Klassen et al., 2000). In a sensitivity analysis, we distributed the isotretinoin treatment utility benefits over the 6-month treatment course linearly and found no significant differences in our main findings. Our basecase health state utility values are lower than other published values and to account for this difference, we conducted a sensitivity analysis that incorporated a utility difference between untreated and fully treated acne health states of 0.02 (Chen et al., 2008; 2004). For adverse events, we assumed a disutility of 0.5, which is in line with published estimates of the disutility of acute pancreatitis in patients with familial hypercholesterolemia (Matza et al., 2020). In a sensitivity analysis, we assumed a disutility of 1.0 (a health state equivalent to death) given the lack of estimate of the impact of acute hepatitis and pancreatitis in patients experiencing an adverse even of isotretinoin therapy.

Costs

We included costs of isotretinoin therapy ($106/wk), laboratory monitoring for the CP strategy ($5.4/wk), and specialist visits ($216/visit) from a health care payer perspective (Barbieri et al., 2020; Centers for Medicare and Medicaid Services, 2019; Isotretinoin. GoodRx, 2021). Simulated patients on isotretinoin were assumed to have 1 specialist visit per month, those who did not complete a full course had visits every 3 months, and those who completed a full course of isotretinoin once per year. Drug and specialist visit costs differed between laboratory monitoring strategies because patients discontinued from isotretinoin therapy before 6 months stopped paying for treatment and required specialist visits every 3 months to treat their ongoing acne. We incorporated a serious adverse event cost of $14,000 (Fagenholz et al., 2007). We updated all costs to 2020 USD using the consumer price index (Organization for Economic Co-operation and Development, 2023).

Sensitivity analysis

We performed univariate sensitivity analysis on key outcomes across their plausible ranges. To summarize univariate sensitivity analysis results, we converted cost-utility findings into NMBs (calculated as: [WTP × difference in efficacy] − difference in costs). We combined the three most influential parameters uncovered in one-way sensitivity analyses to create best- and worst-case scenarios for the NMB of NoLabMon. Finally, we assigned distributions to all uncertain parameters to conduct probabilistic uncertainty analysis (Table 5). We then ran the simulation 10,000 times, drawing from the defined distributions independently each simulation, and collected incremental QALY, cost, and incremental cost-utility ratio results.

Table 5.

Distributions Assigned for Probabilistic Uncertainty Analysis

Variable	Distribution	Reference
Weekly transition probabilities
Low risk on isotretinoin to high risk on isotretinoin	Beta (alpha = 17, beta = 7,326)	(Barbieri et al., 2020)
High risk on isotretinoin to off isotretinoin	Normalized gamma (scale = 1, shape = 2,221.8)	(Barbieri et al., 2020)
High risk on isotretinoin to serious adverse event	Normalized gamma (scale = 1, shape = 2.3)	Literature review, see Materials and Methods
Remain high risk on isotretinoin	Normalized gamma (scale = 1, shape = 97,776.9)	(Barbieri et al., 2020)
Serious adverse event to death	Beta (alpha = 5, beta = 95)	(Banks et al., 2006)
Utilities
On isotretinoin after 2 mo, and full course completion	Beta (alpha = 144.4, beta = 10.9)	(Klassen et al., 2000)
Costs
Laboratory monitoring cost	Gamma (scale = 1, shape = 139.5)	(Barbieri et al., 2020)
Specialist visit cost	Gamma (scale = 1, shape = 216.3)

Data availability statement

No large datasets were generated or analyzed during this study. Minimal datasets necessary to interpret and or replicate data in this paper are available upon request to the corresponding author.

ORCIDs

Ethan D. Borre: 0000-0002-8666-5382

Suephy C. Chen: 0000-0002-0678-7380

Matilda W. Nicholas: 0000-0002-2179-0529

Disclaimer

The funding source had no role in the design, analysis, or interpretation of the study or in the decision to submit the manuscript for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Conflict of Interest

The authors state no conflict of interest.

accepted manuscript published online 28 January 2023; corrected proof published online 2 May 2023