This study protocol was approved by the Committee on Human Research Publications and Ethics of the Kwame Nkrumah University of Science and Technology (CHRPE/AP/298/14) and the Ghana Health Services Ethical Review Committee (GHS-ERC: 12/07/14). It was declared exempt by the Institutional Review Board at the Johns Hopkins Bloomberg School of Medicine. Written informed consent was obtained from all study participants before enrollment into the study. All relevant data are included in the manuscript and as Underlying and Extended data.

The Ghana Access and Affordability Program (GAAP) study is a quasi-experimental study with a pragmatic trial design to examine the effect of DP on improving access to innovator medicines for control of hypertension and diabetes in a multi-center, prospective Ghanaian cohort.

The GAAP study was conducted at five hypertension and diabetes specialty and general clinics in urban, semi-urban and rural locations in Ghana. The study was conducted at two tertiary institutions - Komfo Anokye Teaching Hospital (KATH) and Tamale Teaching Hospital (TTH) - two secondary level health institutions - Agogo Presbyterian Hospital (APH) and Atua Government Hospital (AGH) - and one primary level health institution - Kings Medical Center (KMC).

Participants were eligible if they were ≥18 years with known diagnosis of hypertension and/or type II diabetes presenting for routine care at either a general polyclinic (AGH, KMC, TTH) or a dedicated diabetes/hypertension clinic (KATH, APH). During the period of the study, consecutive potential participants were invited to participate after study nurses had explained the objectives of study. Participants meeting eligibility criteria were enrolled after written informed consent had been obtained by trained research assistants.

At enrollment participants were assigned to DP by research assistants if their monthly household income was <210 Ghana Cedis (i.e. minimum wage) or had a multidimensional poverty index score ≥ 6/18 ²⁰ . Study participants whose physicians prescribed study medications were to purchase them at either DP or MP from the hospital pharmacy. However, when participants allocated to MP could not afford prescribed medicines at MP, they were offered pricing at the lower tiered DP at the pharmacy. Participants who could not purchase study medicines at either MP or DP were prescribed generic alternatives available on the National Health Insurance Scheme ²¹ . The decision tree is shown in Figure 1.

Trained Research Assistants interviewed participants and collected demographic and household information such as age, gender, educational attainment, employment status, monthly income and health expenditures. Study participants were also interviewed on their lifestyle behaviors such as alcohol use, diet, cigarette smoking and physical activity. We categorized alcohol intake and cigarette smoking status as never, former or current users. Physical activity was assessed by asking if participants frequently performed physical activities that caused a small increase in breathing or made their heart rate go up, such as (fast/brisk) walking, jogging, bicycling, and how much time they spent doing physical activity.

A detailed medical history including duration of hypertension or diabetes and medication used for treatment were obtained. Compliance to treatment was assessed using the 14-item Hill-Bone compliance to high blood pressure therapy scale ²² and the 4-item version of Levine-Morisky Medication Adherence Scale ²³ among diabetes patients. A past medical history of stroke was elicited by asking if participant had ever experienced sudden onset of weakness or sensory loss on one side of the body, sudden loss of vision, or sudden loss of speech. Similarly, a history of heart failure was assessed by asking if participant had ever experienced shortness of breath on exertion, on lying down as well as swelling of both feet. Blood pressure and pulse rates were measured by study nurses following a standardized study protocol using an automated BP measurement device (Omron HEM-907XL). Two consecutive BP readings from the same arm taken 2 minutes apart was recorded and averaged for the present analysis. Anthropometric evaluations included measurements of weight and height for body mass index derivation and waist circumference.

To ensure standardization across all study sites, an ISO-certified, quality-assured laboratory in Ghana was contracted to run all biochemical panels including creatinine, lipid profile and hemoglobin A1 _C for participants with diabetes. Samples were transported to the laboratory on the day of collection often within 4 hours or where not feasible (KMC and AGH sites), samples were frozen and delivered to the laboratory the next day.

Following enrollment, each participant was seen at two monthly intervals by their physician for 18 months to assess disease control. At enrollment and during follow-up physicians used their clinical judgment to make any changes in medicines they felt, including adding or discontinuing study medications (either at MP or DP).

We developed a broad range of health systems strengthening activities, such as training of doctors on management of hypertension and diabetes using locally developed treatment guidelines (see supplementary material in 19) aligned with international guidelines, strengthening supply chain systems within health facilities to minimize important access barriers such as stock outs and medication expiry, and developing patient education leaflets on management of diabetes and hypertension ¹⁹ .

The following outcome variables were recorded:

Means and medians were compared using either the Student’s t-test or Mann-Whitney’s U-test for paired comparisons or ANOVA or Kruskal Wallis tests for more than 2 group comparisons. Proportions were compared using the Chi-squared tests or Fisher’s exact test for proportions with subgroupings <5. A multivariate logistic regression analysis was performed to identify factors associated with ability to purchase study medications. Predictors selected in this analysis include age, gender, location of residence, educational status, employment status, monthly household income, price allocation (DP or MP), willingness to purchase study medications should they be prescribed and level of health facility. These factors were selected based on their known or predicted impact on ability to make out-of-pocket payments for medications. A parsimonious logistic regression model was constructed to determine the predictors of BP and glycemic control among participants who purchased study medicines compared with those who could not purchase prescribed study medicines. Variables included in these models included demographic variables, socio-economic indicators, level of healthcare institution, adherence to therapy and number of times purchases of innovator branded study medicines were made. In bivariate analysis, factors that attained a p-value of <0.05 were included in the multivariable model with two-tailed p-values <0.05 considered statistically significant. Statistical analysis was performed using SPSS version 19.

There were 3,296 participants enrolled into the study with a mean ± SD age of 57.5 ± 12.7 years and a preponderance of females comprising 76.6% of the study population. There were more urban dwellers (43.6%), followed by rural (33.7%) and semi-urban dwellers (22.6%). There were 1,869 (56.7%) participants with hypertension only, 422 (12.8%) with diabetes only and 1,005 (30.5%) with both hypertension and diabetes. Overall, mean ± SD duration of hypertension was 7.8 ± 7.3 years and DM was 9.4 ± 6.9 years, with 42.0% of participants with BP controlled and 29.8% with optimal glycemic control at enrollment into the study ( Figure 2). Comparisons of baseline characteristics according to study site and pricing arm allocation are shown in Extended data: Tables S1 and S2.

At enrollment, all participants were on generic antihypertensive and anti-diabetic medicines available on National Health Insurance. A total of 24,632 patient clinic visits occurred over the course of the study with 2,469 (10.0%) decisions to modify treatment ( Table 1). Average follow-up per participant was 14 months. There was a higher tendency for physicians to prescribe generic medicines for treatment modifications than with innovator medicines 76.2% vs 23.8% respectively (p<0.0001). Overall, 238 (45.2%) participants were able to afford prescribed study medicines while 288 (54.8%) were unable.

Table 2 shows a comparison of baseline demographic, socio-economic and clinical characteristics of participants who purchased prescribed study medicines (n=238), those who could not purchase prescribed study medicines (n=288) and those not prescribed any study medicines (n=2,770). These participants were able to access medicines on the National Health Insurance Scheme list ²¹ . Study medicines were considered mainly for participants whose BP and/or glycemic indicators were not optimally controlled at baseline. Several other differences observed between these three groups with regards to demographic and socio-economic characteristics, healthcare expenditures, lifestyle behaviors and disease control are shown in Table 2. Factors associated with ability to afford study medicines with their adjusted OR (95% CI) include monthly household income >1,000 GHS, 2.90 (1.13 – 7.43) [1USD = 4.5 GHS]; allocation to MP arm at enrollment, 1.62 (1.06 – 2.47); willingness to purchase additional medicines for disease management, 2.59 (1.00-6.77); and tertiary level healthcare seekers, 0.29 (0.14-0.59), as shown in Table 3.

Of the 238 participants, 97 (40.8%) assigned to DP arm purchased study medicines at differential price while 141 (59.2%) assigned to MP arm purchased at the market price (n=91), but the remainder could only purchase at differential price (n=50) although they were assigned to MP arm. In total, 177 (74.4%) participants purchased innovator antihypertensive medicines and 61 (25.6%) bought anti-diabetic medicines. Overall, of the 458 study medicines purchases made by the 238 participants, 39.1% were at MP and 60.9% at the DP. However, the ability to afford study medicines were not sustained as 66% of participants could procure them just once ( Extended data: Figure S1). The mean ± SD number of purchases/refills of innovator medicines among those assigned to MP arm who purchased at MP was 1.7 ± 1.4, those assigned MP arm but purchased at DP was 2.4 ± 2.2 and finally those assigned DP who purchased at DP (n=96) was 1.8 ± 1.4, p=0.03 (by ANOVA). Each medication purchase was made for two months of supply.

Among participants who purchased innovator antihypertensive medicines, systolic BP declined from a baseline 147.8 ± 21.7mmHg through to 138.3 ± 22.5mmHg at month 18, (p=0.0007 by ANOVA). Among those who could not purchase prescribed study medicines, systolic BP decreased from 149.5 ± 22.8 to 142.0 ± 26.6mmHg at month 18 (p=0.02, by ANOVA) ( Figure 2). Similar trends were also observed in diastolic BP over time. Proportion of diabetics with HbA1 _C <7% increased from 23% at baseline to 39% at month 18 among participants who purchased study medicines for diabetes control, (p=0.10) and from 30% at baseline to 40% at month 18 (p=0.25) among those prescribed but could not purchase ( Figure 3). In a parsimonious multivariate logistic regression model where age, gender, location of residence, duration of disease, income level and adherence were accounted for, purchase of antihypertensive medications for 6 or more times was associated with an adjusted OR of 4.09 (1.02-16.29) of achieving averaged BP<140/90mmHg during follow-up compared with those unable to purchase prescribed study medicines ( Table 4). Similarly, purchase of study anti-diabetic medications for 5 or more times was associated with an adjusted OR of 6.73 (1.11-40.84) of achieving averaged HbA1c <7% ( Table 5).

It is estimated that approximately 90% of individuals in LMICs use their own funds to purchase medicines ²⁵ . The out-of-pocket expenditures for medicines are second to food, making government subsidies for medicines not realistic ²⁵ . Differential pricing may potentially contribute to attaining the sustainable development goal of Universal Health Care by assuring access to safe, effective, quality and affordable essential medicines ²⁶ . However, as a strategy, differential pricing of essential medicines places the burden of medication purchases on the patient. This burden is dependent on the cost of medicines, the purchasing power of patients and the extent of price reduction on the product by the pharmaceutical company and the size of government taxes, levies, and mark-ups by distributors and health facilities. In LMICs purchasing power of the majority of patients living with NCDs requiring life-long treatment is low. Our findings would support the need to find additional ways to reduce prices of cardiovascular medicines for LMICs which currently bears the greatest burden of CVD on the globe ^{27, 28} . Furthermore, in LMICs where national insurance policies are in existence, co-payment mechanisms for differentially priced innovator medicines to be shared between patients and national health insurance schemes would contribute to mitigating the financial burden on patients. Alternative price reduction mechanisms such as high-volume purchasing, reliable and adequate financing, public advocacy, negotiation, and market competition could contribute to further price reductions ^{15– 18} . Further advantages could come through an integrated approach that addresses supply chain and weak health systems to improve access and affordability of quality assured innovator medications in LMICs. Governmental reduction or elimination of tariffs and charges on medicines together with hospitals dropping pharmacy mark-ups would also be helpful.

A major strength of this pragmatic study is the enrollment of participants at primary, secondary and tertiary health institutions situated in rural, semi-urban and urban settings to enhance generalizability of our findings. Our study is among the few from sub-Saharan Africa to prospectively evaluate BP and glycemic control and we found evidence of modest improvements in disease control ( Figure 2 and Figure 3). Detailed analysis of determinants of disease control for the entire cohort is beyond the scope of the present report. We however exercise caution in over-interpretation of disease control rates in this prospective cohort in the light of high attrition rates during follow-up, which may have been influenced by survivorship bias. Also parsimonious logistic regression models were used to assess the impact of access to study medicines on disease control using a minimalistic set of covariates empirically known to be associated with disease control but not specified a priori.

Our findings have policy implications for pharmaceutical industry and governments. Strategic public-private partnerships and advocacy will be critical to the roll-out of differential pricing as a strategy to assure improved access and affordability of essential medicines for non-communicable diseases in LMICs. Health system strengthening activities such as regular and more frequent hospital visits as done in the present study, together with patient education and training of physicians at study sites may have contributed to different extents in the overall improvements in BP and glycemic control in this cohort. Undoubtedly, efforts at further improvements in hypertension and diabetes control are still urgently needed to avert the rising burden of CVDs and its accompanying unacceptably high morbidity and mortality from CVDs in this region ^{29– 40} . Furthermore enhancing adherence to CVD medicines through mobile health technology as has been shown in the context of stroke in Ghana may be worth pursing further to prevent major adverse cardiovascular events from uncontrolled hypertension and diabetes mellitus ^{41– 43} .

In conclusion, although purchases of study medicines were limited by cost even at differential price, sustained purchases of these quality assured medicines were associated with improved blood pressure and glycemic control. Further price reductions of study medicine below those in the present study are expected to lead to improved access and affordability of life saving cardiovascular preventive medications in LMICs.

Open Science Framework: Differential pricing of medicines to improve access to medicines for hypertension and diabetes control in Ghana: The Ghana Access and Affordability Program, a multi-center prospective trial, https://doi.org/10.17605/OSF.IO/C97HZ ⁴⁴ .

This project contains the following underlying data:

Open Science Framework: Differential pricing of medicines to improve access to medicines for hypertension and diabetes control in Ghana: The Ghana Access and Affordability Program, a multi-center prospective trial, https://doi.org/10.17605/OSF.IO/C97HZ ⁴⁴ .

This project contains the following extended data:

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).