The prognosis for individuals with HIV disease improved dramatically with the advent of highly active antiretroviral therapy (HAART).^{1} Benefits of HAART include an improved likelihood of viral suppression, enhanced immune functioning, and a lower probability of developing viral resistance compared with dual therapy or monotherapy.^{1-3} Although the field continues to strive toward simplification of treatment, many HAART regimens remain extremely complex, often involving a heavy pill burden, multiple doses per day, and complicated dosing instructions. Such complexity may lead to problems with adherence,^{4} particularly for patients with cognitive impairment.^{5}

Substantial variation exists across studies in the conceptualization and quantification of regimen complexity, limiting one's ability to make aggregate conclusions about the impact of regimen complexity on treatment adherence. For instance, regimen complexity has been characterized by the number of doses of medication per day,^{4-6} number of pills per day,^{7} perceived convenience of the regimen,^{8} and a combination of factors that include dosing frequency and specific dosing instructions.^{9} In addition, much of the existing research on this topic is focused on populations such as patients with diabetes and chronic obstructive pulmonary disease (COPD). Methods of quantifying regimen complexity in these studies do not account for all factors relevant to typical treatment regimens prescribed to individuals with HIV disease. For example, George et al^{10} put forth one of the most comprehensive methods of calculating regimen complexity in the COPD population, including instructions for weighting types of treatments (eg, oral intake, inhalation), dosing frequencies, and additional administration directions (eg, take with specific fluid). One drawback that surfaces when attempting to apply this system to regimen complexity in the HIV population is that a weight of “1” is added if multiple units are taken at a single time (eg, more than 1 pill per dose) but there is no distinction between different numbers of pills taken per dose. For example, 2 and 10 pills per dose would both add “1” to the score and result in an equivalent weight. Thus, the complexity of the high pill burden common in antiretroviral (ARV) regimens is not captured.

Among the investigations of factors associated with adherence among HIV-positive patients, few have included regimen complexity as a possible influential variable. Stone et al^{11} examined the complexity of ARV treatment regimens by assessing dosing frequency and special administration instructions, such as whether the medication should be taken on an empty stomach. These factors were considered separately rather than in combination as a total regimen complexity score. Results indicated that patients whose regimens included a medication that needed to be taken on an empty stomach or whose regimens required 3 or more doses per day were significantly more likely to have missed a dose in the past 3 days. Gao et al^{12} defined regimen complexity by adding the number of doses per week across all medications “taken regularly,” although these calculations presumably were not limited to ARV medications. Regimen complexity was not a significant predictor of adherence in this study. When defined in terms of pill burden, results of a longitudinal study indicated that high regimen complexity predicted discontinuation of HAART because of patient preference and/or noncompliance, among other reasons.^{7}

Although these findings point to the need for further investigation of the role of regimen complexity in adherence to ARV therapy, to our knowledge, no published reports describe specific methodology for quantifying the construct as a comprehensive score with appropriate attention to factors such as pill burden, dosing frequency, and special dosing instructions. The purpose of the current article is to present a reliable and valid method of quantifying the complexity of ARV treatment regimens in the HAART era for use with patients with HIV disease.

#### PART I: DEVELOPMENT OF THE ANTIRETROVIRAL REGIMEN COMPLEXITY INDEX SCORING SYSTEM

##### Methods

##### Procedures

During the initial development of the Antiretroviral Regimen Complexity (ARC) Index, a list of regimen characteristics was generated based on the system described by George et al.^{10} Two psychologists, 1 physician, 2 nurses, 2 HIV-positive patients (1 school-aged child and 1 adult), and 2 caregivers of children with HIV disease then provided input with respect to factors for inclusion and weights allotted to various regimen characteristics. The system then was modified slightly by the authors based on their clinical experience. Next, these factors were incorporated into a chart designed for ARC Index computation. To facilitate ease of use, changes to the structure and format of the chart were made based on comments of several coauthors after they had practiced using the chart on a random sample of regimens. The final version of the chart is presented in Figure 1 and may be copied and used by researchers or clinicians. An online version of the scoring system also is available^{13} and is described later in this article.

##### Scoring Method and Rationale

To determine the ARC Index for a regimen, all the required information needs to be completed in the chart (see Fig. 1). First, each medication in the ARV regimen is entered into the first row of the ARC Index chart. Because of limited space, abbreviations are used in place of the full names of the medications. The chart has columns for 6 ARV medications; if a regimen includes more than that amount, an additional chart can be attached that includes the extra drug information. Next, weights are added according to the dosing schedule, method of administration, special instructions, and required preparations. Finally, all points associated with the aforementioned factors are added together to yield a single number that represents the overall ARC Index. More detailed information pertaining to the method and rationale for each section of the ARC scoring system is provided here.

##### Number of Medications

As shown in row 1 of the ARC Index scoring form, the total number of ARV medications in the patient's regimen is multiplied by 0.5, because clinical judgment indicated that each medication did not add an entire unit of weight. For example, from a patient's viewpoint, taking 4 medications may not seem exactly 2 times as complex as taking 2 medications.

##### Dosing Schedule

With respect to dosing schedule, we considered the number of pills per dose and the number of doses per day for each medication as well as whether the regimen includes medications with schedules that are inconsistent with each other. According to patient reports and clinical experience, taking a once-a-day medication and a twice-a-day medication was not notably more complex than taking all medications on a twice-a-day schedule, but taking twice-daily and three-times-daily medications in the same regimen adds a substantial level of complexity. Because dosing times for the various medications in such a regimen may not coincide, patients may end up taking medication at 5 different times throughout a 24-hour period. To account for this situation, the ARC index scoring system first incorporates the highest number of doses taken for any medication in a patient's regimen. Two points are then added if dosing schedules for medications taken more than once a day are inconsistent with each other (eg, if 1 or more medications are administered twice daily and 1 or more medications are administered 3 or more times daily).

To represent pill burden, the number of doses per day is multiplied by the number of pills per dose for each medication, and the products are summed to provide the total number of pills per day. This number is multiplied by 0.25, because each pill did not add significant weight to the complexity of the regimen. For liquid medications, the amount per dose was not factored into the scoring system, because there is substantial variation across the type and size of dosing utensils (ie, some people use spoons of various sizes, whereas others use syringes).

##### Medication Administration

To account for the complexity of various methods of medication administration (pill, liquid, or injection), pill form was used as the “baseline” and not given a weight. Liquid medications were determined to be slightly more complex to administer than pills because of the patient or caregiver having to measure the dose; thus, a weight of “1” is given to a regimen containing 1 or more liquids. Injections were considered significantly more complex to administer than pills; thus, this method is weighted with a “2.” Finally, patient reports indicated that administering all liquid and/or pill medications through a g-tube removes a level of complexity, because the individual does not have to be concerned with having something to drink with the pills or trying to fill the spoon or syringe without spilling. Therefore, 1 point is removed for patients receiving all medications (except injections) by this method.

Although not common, more than 1 administration method may apply to a particular medication. For example, a pediatric patient may take lamivudine (3TC) in tablet and liquid formulations to obtain the right dose for his or her body weight. In such a case, both administration methods should be checked in the column corresponding to that medication and the associated points for each should be included in the computation of the ARC Index.

##### Medication Instructions

This category refers to any specifications or requirements associated with taking a particular drug. Table 1 presents all ARV medications currently approved for use by the US Food and Drug Administration (FDA) and their associated instructions and preparations according to American Hospital Formulary Service (AHFS)^{14} guidelines. Although many ARV medications do not include any special instructions, some require patients to plan their dosing schedule with respect to certain dietary guidelines, thus increasing the complexity of the regimen. The 3 primary instructions are (1) take with food, (2) take on an empty stomach, and (3) take in accordance with a particular dietary content guideline. Current ARV medications that involve a dietary content guideline include amprenavir (AMP) and fosamprenavir (avoid high-fat meals), liquid didanosine (ddI; take antacid with each dose), and indinavir (high fluid intake). Weights are added in this category with a baseline of “0” in mind (no dosing instructions), with scores increasing as the instructions become more complex. Medications may meet criteria for more than 1 instruction. For example, indinavir has 2 instructions (empty stomach and dietary content), and weights for each would be added to the score accordingly. Some regimens may contain medications involving more than 1 diet content guideline. Because each contributes cumulatively to the complexity of a regimen, the different types of diet content guidelines are noted as A (avoid high-fat meals), B (take with antacid), and C (high fluid intake) in the appropriate box in the chart; the total number of different diet content guidelines is then entered and multiplied by a weight of “2.”

##### Medication Preparations

Several ARV medications involve preparations that are necessary for the patient or caregiver to perform before administration of the drug. Presently, ritonavir (RTV) and previously prepared enfuvirtide (ENF) are the only ARV medications for which FDA guidelines recommend refrigeration. Although the recommendation for refrigeration applied to the original formulation of lopinavir/RTV, the newer version of this drug does not maintain this requirement. The regimens used in the development of the ARC Index were prescribed before the availability of the newer version of lopinavir/RTV (May 2005); thus, regimens analyzed for this article are scored in accordance with the previous version of the drug. The complexity of refrigerating medication was the only factor for which there was a lack of consensus among patients and providers. Some patients we asked were stricter about adhering to this rule than others, which influenced their estimation of complexity, and providers had slightly different opinions about the appropriate weight as well. After considering all input and comparing the complexity involved in refrigeration relative to other factors included in the ARC Index scoring system, it was decided that a weight of “1” best represents the complexity of the refrigeration requirement.

Reconstituting a medication involves mixing the liquid with sterile water until the powder is completely dissolved, a process that adds to the time and effort required to prepare the drug for administration. Stavudine and ddI liquids should be reconstituted once a month, and 0.5 is added to a patient's ARC Index for each medication requiring monthly reconstitution. All other current ARV liquids are premixed, eliminating the need for the patient or caregiver to perform this task. ENF, administered by means of injection, is the only ARV drug that requires daily reconstitution at the present time. This process is lengthy and complicated, and it involves adding sterile water to the solution, tapping and rolling the vial to facilitate dilution, and waiting up to 45 minutes for the powder to reconstitute fully. The morning and evening doses of ENF can be prepared at the same time. Two points are added to a patient's ARC Index to reflect the complexity of these requirements.

##### Example of Antiretroviral Regimen Complexity Index Scoring Procedures

Instructions for computing the ARC Index are shown in the Appendix. Use of the scoring system is outlined in the following example (Fig. 2), and multiple examples can be accessed from the on-line scoring system.

##### Sample Regimen

This regimen includes 150 mg of 3TC (1 pill twice a day), 300 mg of tenofovir (TDF; 1 pill once a day), 100 mg of RTV (1 pill twice a day), 1000 mg of AMP (8 pills twice a day), and 65 mg of ENF (1 injection twice a day).

##### Row 1

Abbreviations of the patients' ARV medications are entered in row 1, columns B through F. A “5,” representing the number of ARV medications, is placed in row 1, column I. This number is multiplied by 0.5, and the product (2.5) is placed in row 1, column K.

##### Rows 2 Through 5

TDF is administered once a day; thus, a “1” is placed in row 2, column C. 3TC, RTV, AMP, and ENF all are administered twice a day; thus, a “2” is placed in row 2, columns B, D, E, and F. Because “2” is the highest number of doses per day, a “2” is entered in row 2, column K. One or more medications are given 2 times a day, but none of the medications are administered 3 or more times a day; thus, row 3, column K is left blank.

One pill is administered per dose of 3TC, TDF, and RTV; thus, a “1” is entered in row 4, columns B through D. Eight pills per dose of AMP are administered; thus, an “8” is placed in row 4, column E. Row 4, column F is left blank, because ENF is administered as an injection. For each column (B-E) in row 5, the number in row 2 is multiplied by the number in row 4. The resulting products are entered in row 5 as a “2” in columns B and D, a “1” in column C, and a “16” in column E. The sum of these numbers is entered in row 5, column I. In this case, 2 + 1 + 2 + 16 = 21. The “21” is then multiplied by 0.25 as specified in row 5, column J, and the result is entered in row 5, column K (21 × 0.25 = 5.25).

##### Rows 6 Through 8

ENF is administered by means of injection; thus, a check is placed in row 7, column F and a “2” is entered in row 7, column K. No other medication is injected, and no medications are taken in liquid form or through a g-tube; thus, the corresponding spaces (row 7, columns B-E; row 6, columns B-F; and row 8, columns B-F) are left blank.

##### Rows 9 Through 12

No dietary instructions are associated with 3TC, TDF, or ENF; thus, rows 9 through 11 in columns B, C, and F are left blank. RTV should be taken with food; thus, a check is placed in row 9, column D, and a “1” is entered in row 9, column K as specified in column H. Instructions for AMP specify a dietary content restriction (avoid a high-fat diet); thus, a check and the letter “A” (to indicate which of the 3 restrictions applies) are placed in row 11, column E, and a “1” is entered in row 11, column I. This number is multiplied by 2 as specified in row 11, column J, and the result (“2”) is placed in row 11, column K.

Two different instructions are checked in rows 9 through 11 (with food and dietary content); thus, a “2” is entered in row 12, column K.

##### Rows 13 Through 15

RTV should be refrigerated, and if both daily doses of ENF are prepared with the first dose (which is usually the case), the second dose should be kept refrigerated until it is administered later. Thus, checks are placed in row 13, columns D and F, and a “1” is entered in row 13, column K. ENF is reconstituted daily; thus, a check is placed in row 14, column F, and a “1” is entered in row 14, column I. This number is multiplied by 2 as specified in row 14, column J, and the result (“2”) is entered in row 14, column K.

##### Row 16

The numbers in rows 1 through 15, column K are added together, and the sum represents the total ARC Index score: 2.5 + 2 + 5.25 + 2 + 1 + 2 + 2 + 1 + 2 = 19.75.

#### PART 2: VALIDATION OF THE ANTIRETROVIRAL REGIMEN COMPLEXITY INDEX SCORING SYSTEM

##### Methods

##### Antiretroviral Regimens

This study rated the complexity of the most current ARV regimens, collected through retrospective database review, of 67 HIV-infected children, adolescents, and adults (mean age = 15.3 years, range: 6 to 25 years) evaluated at the Pediatric HIV Clinic at the National Cancer Institute between 1999 and 2005. The 67 regimens from these patients contained a median of 3 (range: 2 to 6) ARV drugs. As shown in Table 2, regimens most commonly included 1 or more nucleoside reverse transcriptase inhibitors (NRTIs) and 1 or more protease inhibitors (PIs; n = 44, 66%). All but 1 regimen (99%) contained a combination of different classes of drugs.

##### Procedures

##### Reliability

Three raters, including 2 psychologists and 1 intermediate-level psychology graduate student, helped to establish reliability by scoring a list of regimens according to the proposed ARC Index system (using the paper chart rather than the on-line scoring system) without any accompanying patient information. First, 1 rater scored all 67 regimens. Of these regimens, 20 were chosen at random and scored by the 2 additional raters to assess interrater reliability. In addition, 2 of the raters rescored the same 20 regimens 2 weeks later as a measure of test-retest reliability.

##### Validity

Construct validity was assessed by ordering all 67 regimens according to their ARC Index from lowest to highest. The regimens then were divided into quintiles, and 1 regimen from within each quintile was chosen at random, with the provision that there was a difference of at least one half of 1 standard deviation (SD) between each regimen's score. Next, 4 additional raters, masked to the identity of the patients and unfamiliar with the ARC Index scoring system, independently ranked the 5 regimens in order from the least to the most complex, using their clinical judgment. These raters included 1 clinic nurse, 1 nurse practitioner, 1 physician, and 1 pediatric neurologist, all of whom had extensive experience in working with HIV-infected patients. The rankings then were contrasted with ARC Indexes to determine construct validity. Discriminant validity was assessed by determining the relations between the 67 ARC Indexes and factors that theoretically should not be related to regimen complexity, including caregiver education, patient gender, and caregiver-child relationship (biologic vs. nonbiologic).

##### Statistical Analyses

An interclass correlation coefficient (ICC) using 2-way analysis of variance (ANOVA) was calculated to determine interrater reliability. Test-retest reliability was calculated with a Pearson correlation coefficient. For testing discriminant validity, the relations between ARC Indexes of HAART regimens and theoretically unrelated variables were determined by conducting correlations between continuous variables or *t* tests for comparing continuous data between groups.

#### RESULTS

##### Regimen Characteristics

The mean ARC Index of the 67 regimens was 10.38 (SD = 3.2). The regimen receiving the lowest score (4.0) included 3TC, 1 pill twice a day, and stavudine, 1 pill twice a day. The regimen receiving the highest score (19.75) included TDF, 1 pill twice a day; AMP, 8 pills twice a day; RTV, 1 pill twice a day; 3TC, 1 pill twice a day; and ENF, 2 injections per day.

##### Reliability

Two-week test-retest reliability for the ARC Index was high (*r* = 0.98). Similarly, interrater reliability across 3 independent raters, as calculated by ICC, was excellent (*r* = 0.97). The few discrepancies between the raters were attributable mostly to mathematic errors or to a rater forgetting to include a weight for a drug that required multiple instructions (eg, take on an empty stomach and high fluid intake).

##### Validity

With respect to construct validity, the 4 raters who were asked to rank 5 regimens in order of complexity produced identical rankings. Moreover, the order of their rankings was consistent with the order of the regimen scores obtained with the ARC Index system. The characteristics of these 5 regimens and their ARC Indexes are shown in Table 3.

Results of analyses assessing the discriminant validity of the ARC Index were as predicted. Specifically, regimen complexity scores were unrelated to years of caregiver education (*r* = −0.12, *P* = 0.34), patient gender (*t* = 0.45, *P* = 0.66), and the caregiver-child relationship (biologic vs. nonbiologic; *t* = 0.84, *P* = 0.40).

#### DISCUSSION

Despite the complexity of ARV treatment regimens for patients with HIV disease, little is known about the impact that regimen complexity has on treatment adherence. This is partially attributable to the lack of an objective comprehensive system of quantifying ARV regimen complexity. Results of the current study show that the ARC Index scoring system is a valid and reliable method of quantifying the complexity of ARV treatment regimens. Measures of test-retest reliability, interrater reliability, construct validity, and discriminant validity all affirm the utility of this system. Moreover, the ARC Index can be used by professionals from various disciplines and with a wide range of experience; psychology graduate students, psychologists, nurses, and physicians were able to use the system successfully by following the instructions provided. The regimens used for analyses in the current article were prescribed to children, adolescents, and young adults ranging in age from 6 to 25 years. The ARC Index is generalizable to regimens taken by HIV-positive patients of all ages, however, because AHFS guidelines and input from pediatric and adult patients confirmed that the factors contributing to the complexity of ARV regimens are relevant when applied to pediatric and adult regimens. Furthermore, even broader applications are possible with other chronically ill populations requiring complex medical and pharmacologic management.

Although the ARC Index scoring system was designed to be relevant to any individuals taking ARV medications, users may need to modify the weights or include additional factors if there are special considerations that are not applicable to the HIV-infected population in general. For example, among patients with renal complications, the FDA recommends decreasing the dosing frequency of some medications based on the patient's creatinine clearance. This may involve adjusting a patient's dosing schedule from once a day to once every other day. Incorporating an every-other-day medication into one's routine can be considered more complex than taking a once-a-day medication, because in the latter case, it is unnecessary to keep track of whether or not one needs to take the medication on a given day. Thus, our recommendation is to prescribe a weight of 2 for any every-other-day dosing schedule.

The ARC Index scoring system does not consider side effects or other situational factors that may vary across patients (eg, pill swallowing ability). Instead, scores are determined based on fundamental aspects of the regimen itself and the time and difficulty involved in taking the regimen for a standard patient. In addition, the ARC Index represents the complexity of a regimen when taken in accordance with all instructions. A patient taking AMP who chooses not to adhere to the dietary content guideline (avoid high-fat meals) essentially is simplifying his or her regimen. Research suggests that such modifications are not uncommon, even among individuals who adhere to other regimen requirements (ie, some patients take the correct number of pills per day but not always at the right times or with appropriate attention to dietary guidelines).^{11} Clinicians and researchers should be aware of this possibility and its potential impact on the interpretation of the ARC Index.

Calculating a patient's ARC Index should not overshadow the clinical importance of asking patients how they are coping with the complexities involved in adhering to their ARV regimen, even when their regimens are less complicated. This is especially important because patients who are prescribed the same or similar regimens may express substantially different subjective experiences. Asking patients about their individual experiences in managing their regimen facilitates patient-clinician communication about the challenges involved in remaining adherent and helps providers to identify specific circumstances that may need to be considered in the interpretation of a patient's ARC Index.

An electronic version of the ARC Index is available online^{13} and provides researchers with the ability to utilize the system quickly and accurately. Users of this system enter the names of the patient's medications from drop-down menus as well as the doses per day and pills per dose. If any medications are given in liquid form or by means of a g-tube, this is indicated by the user as well. All other information (medication instructions and preparations) is added automatically but can be adapted if needed. The e-scoring system then calculates the ARC Index, making the process simple, accurate, and efficient.

##### Limitations

The ARC Index incorporates ARV medications only; yet, it is common for HIV-infected patients to be taking numerous other medications, such as antibacterial and antifungal drugs for prophylaxis and to counteract ARV-related side effects. Some of these non-ARV drugs have somewhat complex instructions, particularly when combined with the ARV drug schedules, and many need to be refrigerated. Thus, non-ARV medications may add significant complexity to an individual's regimen that is not accounted for by the ARC Index. Users who wish to incorporate additional non-ARV drugs in the scoring system can utilize an extra form to enter each concomitant medication and the associated dosing information for each.

The computation of ARC Index scores by hand is susceptible to human error. Users of the system must ensure that all instructions have been followed carefully and that all mathematic calculations have been performed correctly. The individuals who used the system during its development and validation reported that the process was somewhat slow initially but became much quicker and easier with practice. As mentioned previously, the e-scoring system decreases the time involved and the likelihood of computational errors in calculating ARC Indexes.

##### Future Directions

One important area for future research is the examination of how ARC Index scores relate to objective measures of treatment adherence, because patients who are prescribed more complex regimens may be less likely to be adherent, and to markers of immunologic and virologic function. For example, there may be a threshold value of complexity beyond which the probability of maintaining optimal adherence decreases. An exploration of patient variables that may mediate this relation, such as age, treatment history, psychologic stability, and neurocognitive functioning, may yield valuable findings as well. Results of these investigations should help to identify patients who are at risk for poor adherence and who could benefit from ancillary services. Additional support from the medical team and/or home nursing visits aimed at problem solving with patients to determine effective strategies for dealing with the specific challenges their regimen entails could be provided accordingly. Our group currently is collecting data to investigate the aforementioned relations.

#### CONCLUSIONS

Researchers are beginning to recognize the theoretical importance of regimen complexity and have examined this concept among HIV-positive individuals and other disease populations, most often in relation to treatment adherence. Without a comprehensive, consistent, and objective technique for measuring the construct, however, the capacity for research in this area and the generalizability of the results are substantially limited. The establishment of a valid and reliable method of quantifying regimen complexity that is universally applicable to children and adults with HIV disease is the first step toward obtaining a better understanding of how the complexity of ARV regimens affects patient adherence and, in turn, medical outcomes among this population. The ARC Index provides a method of examining this important construct consistently across investigations, thereby facilitating better communication among researchers and furthering progress toward the goal of maximizing treatment adherence and diminishing the detrimental effects of HIV disease.

#### ACKNOWLEDGMENTS

The authors thank the patients and caregivers who provided input on the ARC Index scoring system and Paul Jarosinski, PharmD, Robin Bent, RN, Lucy Civitello, MD, Alexandra Freeman, MD, and Julia Purdy, CRNP. They also thank Sally Hunsberger, PhD, for her assistance with interclass correlation coefficient analyses.