The Impact of Healthcare Benefit Changes on High-Cost Utilizers

Transcription

1 2008 Milliman All Rights Reserved M I L L I M A N The Impact of Healthcare Benefit Changes on High-Cost Utilizers Prepared by: Frank Kopenski, Jr., ASA Milliman, Inc. and Grant Lawless, MD, RPh Amgen Inc. November, 2007 Commissioned by Amgen Inc Bluemound Road, Suite 400 Brookfield, WI Tel Fax

2 M I L L I M A N November 2007

3 The Impact of Healthcare Benefit Changes on High-Cost Utilizers PAGE FORWARD... 1 I. INTRODUCTION AND OTHER STUDY FINDINGS... 3 II. THE BIG PICTURE... 5 III. THE DILEMMA: MEMBER COST SHARING VERSUS MEMBER PREMIUM IV. HIGH-COST UTILIZERS HIGH-COST CONDITIONS V. WHY THE FOUR-TIER DRUG BENEFIT? VI. FINDINGS ACKNOWLEDGEMENTS REFERENCES November 2007 M I L L I M A N

4 M I L L I M A N November 2007

5 FORWARD The recent advent of increased cost-sharing programs and changes in benefit designs have led some to predict negative consequences resulting from limitations in prescription drug benefits, especially for the elderly and severely ill. Unfortunately, there has been limited research to date on the long-term impact of reducing access for the growing number of chronically and severely ill Americans. Several recent studies have begun to measure the real impact of limited access to drug therapy and the resulting clinical and total cost implications on the healthcare system. Consumer-advocacy groups and others have attracted much public interest in the ages 65 and over Medicare Part D Drug Program with its $3,850 annual out-of-pocket (OOP) provision, often called the coverage gap. In addition, some are now focusing on the ever-expanding number of chronically and severely ill patients below 65 years of age who possess private insurance; these ill patients are feeling the impact of higher cost sharing for drug therapy through new employer and managed care benefit designs. The current unilateral practice of assigning drugs to various copayment schedules, tiered formularies, and coinsurance based on cost effectiveness leaves many coordination-of-care concerns unresolved. Unresolved concerns include the unpredictable impact on patient adherence, the variability in clinical impact across patients for a given drug, and total cost burden based on an individual s mix of comorbidities and other therapies. In addition, this potential loss of drug compliance may result in even greater use of the healthcare system in ways otherwise unanticipated. November 2007 M I L L I M A N 1

6 2 M I L L I M A N November 2007

7 I. INTRODUCTION AND OTHER STUDY FINDINGS The following study is a result of Milliman research and cost analysis, using primarily a four-year study period ( ) comprised of commercial and retiree group healthcare experience. The purpose of this study is to take a closer look at the current healthcare benefit landscape and, with the help of actual employer active and retired claim data, try to quantify the cost impact of such change, particularly for high-cost utilizers of healthcare services. This study focuses on: The disproportionate use of health care services by insured individuals, The relationship between cost of living and healthcare cost trends, Benefit designs that merely shift costs from employer to member, and Certain characteristics of high cost utilizers of health care. Recently, a Medicare Advantage study by Kaiser Permanente - Northern California Region found that elderly patients with chronic disease who meet their $1,000 annual spending allowances were far more likely to use costly alternative hospital and emergency services. Kaiser s Dr. John Hsu stated, Their blood pressure, cholesterol levels, and blood sugar control were all worse when compared with seniors who did not hit their limits. Dr. Hsu went on to say, Although their drug costs were lower, the higher costs of hospitalizations and emergency department visits offset those savings, resulting in consistent unfavorable health effects and failed to save money. (1) To better understand the impact of patient risk stratification, a RAND study looked at the association between compliance and hospitalizations use over four years. The study, using health plan data for those ages 20 and above, found that when copays for cholesterol-lowering drugs were increased from $10 to $20, compliance fell by up to 10%. This resulted in greater use of hospitalizations and emergency department visits for the noncompliant group versus flat trends overall. RAND went on to simulate the potential impact of eliminating copays for highand medium-risk cholesterol patients while increasing copayments for the lowest-risk members. The net result was a substantial savings, over $1 billion annually. In their conclusion, RAND stated that Strategically reducing copayments for patients most at risk can improve overall compliance and reduce use of other expensive services. (2) Recent benefit design and drug use research by RAND had looked at privately insured patients covered by high out-of-pocket (OOP), employer-based benefits with four major medical conditions: cancer, kidney disease, rheumatoid arthritis, and multiple sclerosis. The RAND study concluded that care management should focus on making sure that patients who will most benefit (from specialty drugs) receive them. Once such patients are identified, it makes little economic sense to limit coverage. The study also revealed a generally linear relationship between patients severity of illness and their relative inelasticity to rising copays based on current OOP limits, and found that the most severely ill patients, those with few therapeutic November 2007 M I L L I M A N 3

8 alternatives, were willing to tolerate the highest copays. RAND concluded that increased cost sharing for specialty products would serve to primarily transfer a much larger financial burden from the health plan to the patient, thus reinforcing the necessity of ensuring reasonable OOP copay limits when designing both medical and pharmacy benefits. (3) The federal Agency for Healthcare Research and Quality (AHRQ) recently studied data from the Medical Expenditure Panel Survey, a nationally represented review of patients under 65 years old, to investigate total OOP burdens, including premium costs. AHRQ found that the proportion of patients where OOP was greater than 10% of household income rose from 15.8% in 1996 to 19.2% by Similarly, OOP representing greater than 20% of household income also rose from 5.5% in 1996 to 7.3% by The impact of rising OOP trends hit age groups between hardest, along with female subscribers. Additionally, patients with severe diseases including stroke, arthritis, cancer, and kidney disease were also disproportionately affected by high OOP burdens. Hospital and prescription copays represented the greatest disproportion of OOP burden when compared with the average member. The study concluded that high OOP burdens are associated with delaying or forgoing medical care for financial reasons, behavior that can have severe consequences for those in poor health. (4) In the January 2007 issue of Health Affairs, the new concept of Value-Based Insurance Design (VBID) is discussed. When everyone is required to pay the same OOP amount for health care services whose benefits depend on patient characteristics, there is enormous potential for both under and overuse. Unlike most current health plan designs, VBID explicitly acknowledges and responds to patient heterogeneity. It encourages the use of services when the clinical benefits exceed the cost and likewise discourages the use of services when the benefits do not justify the cost. (5) 4 M I L L I M A N November 2007

9 II. THE BIG PICTURE Group health insurance benefits have gone through a significant change over the past four years, with more emphasis on cost shifting to members to keep healthcare expenses down. This change will invariably have an impact on insured members and, in particular, individuals with high-cost conditions who utilize a disproportionate share of the healthcare system resources. Roughly 20% of commercially insured employees consume about 80% of healthcare resources. The OOP cost burden to these high-cost employees may make it difficult for them to afford healthcare if plan sponsors shift a greater portion of the healthcare cost burden to employees. Table 1 shows this disparity among commercially insured members. TABLE Commercial Population Healthcare Cost Distribution Percentage of Highest Utilizers Responsible for x% of all Healthcare Costs 1.0% 23.1% 5.0% 46.6% 20.0% 77.7% 50.0% 96.0% 75.0% 99.2% Source: Milliman 2005 proprietary data for a commercial population Although not quite as pronounced, this disparity also exists in the senior population (Medicare eligible), as shown in Table 2. TABLE Ages 65 and Older Population Healthcare Cost Distributions Percentage of Highest Utilizers Responsible for x% of all Healthcare Costs 1.0% 13.6% 5.0% 31.5% 20.0% 64.4% 50.0% 89.7% 75.0% 97.9% Source: Milliman 2005 proprietary data for an ages 65 and older population November 2007 M I L L I M A N 5

10 The Milliman data sources show that roughly 50% of the senior population produces 90% of the total healthcare costs. Compared with the commercial population (50% responsible for 96% of cost), there are more senior utilizers driving the overall program cost. This difference is partly due to average age and lack of children (who are lower utilizers of healthcare services). In 2005, the average commercially insured member in the Milliman data had an estimated annual OOP cost of $1,214; however, the average OOP cost of healthcare services for highcost utilizers (i.e., people in the top 5% of healthcare consumption) was in excess of $2,800 (see Table 3). According to the US Census Bureau, the median household income per household member for people under 65 years of age was $24,892 in Thus, the high-cost utilizer would consume more than 11.0% of median household income per household member for healthcare. TABLE Commercial OOP Expense as a Percentage of Median Household Income Assuming $24,892 Household Income per Household Member Healthcare Cost Percentile Member Out-of-Pocket Top 1% $3, % 5% $2, % 20% $2, % 50% $1, % 75% $1, % 100% $1, % Percentage of Estimated Income Source: Milliman 2005 proprietary data for a commercial population, and 2006 US Census Bureau Economic Supplement Household Income The retiree or Medicare-eligible member 65 and older has a slightly lower household income per household member, and high-cost utilizers could consume in excess of 22% of household income. This is as shown in Table 4, based on the Milliman data. 6 M I L L I M A N November 2007

11 TABLE Ages 65 and Older OOP Expense as a Percentage of Median Household Income Assuming $23,316 Household Income per Household Member Healthcare Cost Percentile Member Out-of-Pocket Top 1% $6, % 5% $5, % 20% $4, % 50% $4, % 75% $3, % 100% $3, % Percentage of Estimated Income Source: Milliman 2005 proprietary data for a commercial population, and 2006 US Census Bureau Economic Supplement Household Income It comes as no surprise that healthcare use is not equal across the insured population. Group insurance was created to alleviate this disparity by spreading costs and premiums more evenly across all members. One way of analyzing the equality or inequality of these costs and premiums is by applying the Gini coefficient, a statistical measure of equality / inequality developed in an academic setting. A Gini coefficient value of 0.0 indicates perfect equality, and a Gini coefficient value of 1.0 indicates perfect inequality. Thus, if one individual consumed 100% of all healthcare services, statistically that would represent perfect inequality, or a Gini coefficient of 1.0. Table 5 shows the Gini coefficients for the 2005 Milliman study data: TABLE Members: Gini Coefficients Commercial Medicare Total Healthcare Spending Member Out-of-Pocket Source: Milliman 2005 proprietary data for a commercial and an ages 65 and older population November 2007 M I L L I M A N 7

12 The following graphic depicts the 2005 commercial total healthcare spend Lorenz Curve (a graphical representation of the cumulative distribution function for healthcare expenditures in our case) and corresponding Gini coefficient area which represents 75% of the area below the 45 line. 100% Milliman 2005 Proprietary Commercial Population All Benefit Types 90% 80% Percent of Total Spend 70% 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Percent of Population Gini Coefficient Area Lorenz Curve of 2005 Total Spend It is important to note that there is less inequality in member OOP expense for healthcare costs than there is for total healthcare spending. This is because benefit designs include copayments, OOP limits, and services that require no cost sharing at all. If members were required to pay, for example, 20% coinsurance on all healthcare expenditures and no premium, the member OOP Gini coefficient would be the same as the total spending coefficients in Table 5. Benefit designs that do not include or remove the member OOPlimit will cause the Gini coefficient column values in Table 5 to converge, which may begin to test the principles of group insurance. 8 M I L L I M A N November 2007

13 Member OOP spending on healthcare services between 2002 and 2005 slightly outpaced the percentage increase in total healthcare spending. However, member OOP spending increased by more than four times the increase in cost of living during this period for the commercial population and seven times for the senior population as shown in Table 6. As a result, income is being eroded to pay for healthcare expenditures, and those who consume the highest percentage of healthcare services may need to make some difficult decisions regarding both nonessential and essential healthcare services. TABLE 6 Income versus Healthcare Cost Trends Three-Year Cost Per Person Change 2002 to 2005 Cost of Living 8.3% Total Commercial Healthcare Costs 37.4% Commercial Member Out-of-Pocket Costs 39.5% Total Ages 65 and Older Healthcare Cost 56.9% Ages 65 and Older Out-of-Pocket Cost 60.1% Source: Milliman proprietary data for a commercial and an ages 65 and older static population, and American Institute for Economic Research November 2007 M I L L I M A N 9

14 III. THE DILEMMA: MEMBER COST SHARING VERSUS MEMBER PREMIUM An insured member will typically pay for healthcare expenses through point-of-service cost sharing (i.e., deductibles, copayments, and coinsurance) and a monthly premium charge. In its simplest community-based form, the premium is the same for group-insured members regardless of their healthcare service use. The cost sharing, however, affects individuals differently based on the services utilized and the cost of those services. If the employer increases the employee premium contribution by 1%, therefore decreasing the employer contribution by 1%, all employees are affected. If the employer reduces benefits by 1%, it affects only the employees using healthcare services and generally results in a decrease in employee premium contribution. The end result is less than a 1% reduction in employer contribution and a potential loss of care-management efficiency. If the benefit design requires no member cost sharing and the premium is uniform or community based, then the corresponding Gini coefficient is 0.0, which represents perfect equality across all members. When introduced to the market in the early 1980s, health maintenance organizations (HMOs) came close to this concept. The main drawback of having no cost sharing, however, is over utilization of services, because there is no incentive for a member to be a conscientious consumer. As member point-of-service cost sharing increases, premiums generally decrease and the corresponding Gini coefficient moves toward 1.0 (greater inequality). For the employer, the dilemma is how to balance the benefit provided to employees and retirees with the premium required to fund such benefits. A certain amount of cost sharing is required to control over utilization of services and make for affordable premiums. However, too much cost sharing can have an adverse impact on certain individuals, resulting in additional employer costs. Employer decisions about benefits often focus on the average member cost and premium. It is important to be aware that subtle benefit changes that have a marginal effect on the average member may have a significant impact on the higher-cost member. It is also important to differentiate between high-cost members with serious healthcare conditions and those without. The amount of members annual cost sharing would not be as critical if income were increasing at the same rate for high-cost utilizers. However, the cost-of-living increase provided earlier indicates a significant erosion of income. Thus, balancing employee and retiree benefits and premiums is important to all members, but significantly more important to those with high-cost healthcare conditions. 10 M I L L I M A N November 2007

15 Some recent benefit changes that create concern for these individuals are: Increased deductibles, copayments, and coinsurance for medical services. Increased OOP limits. Change from prescription drug copayments to coinsurance. Use of a fourth tier in drug benefit designs with coinsurance on high-cost drugs. Varying cost sharing based on healthcare annual expenditures or conditions. Decreases in the ratio of member premium to total member OOP expenses (i.e., less uniform spreading of costs across all participants). Increase in benefit coverage exclusions. To quantify the impact of some of these benefit changes; let us compare an average member to an average member with cancer or rheumatoid arthritis (RA). TABLE 7 Healthcare Benefit Cost Comparison Example Average Member with Cancer Average Member with RA Average Member 2005 Average Healthcare Cost $3,467 $22,396 $13, Average Medical Cost 2,641 19,896 9, Average Rx Cost 826 2,500 4,863 Current Medical: 20% Coins., $1,500 OOP 528 1,500 1,500 Current Rx: $30 per Script, No OOP (est.) ,050 3 Current Member Cost Sharing 798 2,190 2,550 Current Member Premium Change Medical to 25% Coins ,500 1,500 Change Medical OOP to $2, ,500 2,282 Change Rx to 25% Coins ,216 Revised Member Cost Sharing 867 3,125 3,498 Revised Member Premium (est.) Impact on Member 3.2% 31.3% 28.2% 1 Assume 9 prescriptions per year 2 Assume 23 prescriptions per year 2 Assume 35 prescriptions per year 4 Member Premium = (total cost member cost share)/0.90 x 25% member contribution rate Source: Milliman 2005 proprietary data for a commercial population November 2007 M I L L I M A N 11

16 Table 7 shows that simple benefit-design changes that have little impact on the average member may have a significant impact on members with more serious healthcare needs. The benefit change presented increases the average member s OOP expense by 3.2%, however, the increase for members with cancer or rheumatoid arthritis is 31.3% and 28.2% respectively. Since members with high cost conditions are responsible for a large percentage of the total healthcare expenditure, it is important to look beyond the immediate cost savings impact of any benefit change. 12 M I L L I M A N November 2007

17 IV. HIGH-COST UTILIZERS HIGH-COST CONDITIONS People often associate high-cost utilizers of healthcare services with members with high-cost healthcare conditions (e.g., cancer, rheumatoid arthritis, chronic kidney disease, and others). This is partly true, but many individuals with chronic healthcare issues incur significant healthcare expenses annually, whereas high-cost utilizers of healthcare services one year may be low-cost utilizers a year or two later. As evidence of this, the following table isolates individuals with varying levels of annual healthcare costs in 2002 and then looks at those same individuals healthcare costs two years later. TABLE 8 Two-Year Commercial Member: Annual Healthcare Trends 2002 Annual Cost 2002 Average Cost 2004 Average Cost Two-Year Trend Change in Healthcare Cost $0 $0 $1,232 Infinite $1,232 $1,000 $291 $1, % $1,423 $5,000 $1,073 $2, % $1,677 $10,000 $1,600 $3, % $1,622 $10,000+ $24,928 $15,191-39% ($9,737) $50,000+ $95,982 $36,976-61% ($59,006) $100,000+ $174,591 $54,658-69% ($119,933) $500,000+ $694,780 $108,062-84% ($586,718) Source: Milliman proprietary data for a commercial static population The table shows that the members with lower healthcare costs in 2002 (i.e., below $10,000) exhibited two-year trends of over 100%. Alternatively, members with higher healthcare costs in 2002 exhibited negative two-year trends of 39% or more. This pattern indicates some regression toward the mean (i.e., the average annual healthcare expenditure per person), since a significant portion of healthcare cost is random from year to year. If we take out cancer, rheumatoid arthritis, and chronic kidney disease patients, for example, the two-year trends would be even more dramatic because the members with these conditions exhibit less cost variation from year to year. November 2007 M I L L I M A N 13

18 V. WHY THE FOUR-TIER DRUG BENEFIT? No part of healthcare benefit design has undergone a bigger change over the past five to ten years than prescription drugs: from flat copayments to two- and three-tiered copayments to deductibles and coinsurance and, more recently, to the four-tier design. It is unlikely that the four-tier design s full impact will be seen until it is more universally adopted as the benefit of choice by employers. This has not been the case as of the time of this study. The four-tier benefit is primarily a reaction to specialty pharmacy programs introduced to the prescription drug benefit by pharmacy benefit management organizations. Specialty pharmacy programs target high-cost biotech products, which are often injectable and expensive. The purpose of the specialty pharmacy program is to create a distribution channel under the pharmacy benefit rather than the medical benefit, to better control pricing, delivery, and patient care management. The four-tier drug benefit moves these products on a voluntary patient purchase basis to the fourth tier, typically with a 25% member coinsurance. Although currently available data does not reflect the full impact of the four-tier benefit design, we can model the impact by reviewing the recent average wholesale prices for a short list of fourth-tier drugs under the evolving benefit design changes. TABLE 9 Sample Four-Tier Drug List Drug NDC* AWP** $30 Copay 25% Coinsurance Humira $1,440 $30 $360 Avonex $1,076 $30 $269 Aranesp $2,134 $30 $534 Enbrel $720 $30 $180 Epogen $1,412 $30 $353 Neulasta $3,217 $30 $804 Neupogen $2,507 $30 $627 Gleevec $2,938 $30 $735 Procrit $1,752 $30 $438 * NDC= National Drug Code identifier ** AWP=average wholesale price Some benefit designs have even higher coinsurance than 25%, and some have a maximum copayment value, such as $250, for example, or an annual OOP limit on the fourth-tier drugs only. In the absence of lower drug alternatives with similar clinical outcomes, it is evident that member OOP costs for high-cost patients will increase further. 14 M I L L I M A N November 2007

19 The Debate Over the Four-Tier Drug Benefit Design At first glance, charging higher cost sharing for the most expensive drug products seems to make sense, because it reduces plan costs while charging each member a more consistent percentage across all drug products. Upon closer inspection, however, there are some significant, debatable issues with current four-tier benefit designs. Currently, the biggest value of specialty pharmacy programs is that the purchase price for drugs can be 15% to 30+% lower than when drugs are purchased by the physician s office. However, a plan cannot maximize these savings if the member is better off getting the drugs from the physicians office because the member cost sharing is lower. Members would be foolish to voluntarily use the specialty pharmacy if it costs them more OOP. Specialty pharmacy products are prescribed for people with cancer, rheumatoid arthritis, chronic kidney disease, multiple sclerosis, hepatitis C, and other complicated diseases. Individuals with these conditions do not choose between varying-cost brand medications and generics, they choose between very high-cost brand medications and other very high-cost brand medications. The increased cost sharing does not act as a steering mechanism to lower cost alternatives (as it does in the three-tier benefit design). Disease management programs for medical and prescription drug services have always focused on higher-cost health conditions. The key ingredients of these programs have so far achieved better clinical outcomes, intervention, education, and patient care management. The Let s just charge these people more out OOP approach does not seem to fit these programs. Medicare has addressed the huge disparity between medical and pharmacy pricing for specialty drug products by moving the medical benefit reimbursement to average sales price (ASP) + 6%, or using the BioScrip competitive acquisition program. If followed by commercial payers, this move may diminish the specialty pharmacy and four-tier benefit design movement. There are employers who have favored the four-tier design as a cost management strategy, and those such as the University of Michigan who actually charge less for these high-cost drugs to promote patient access to a more complete complement of care options. November 2007 M I L L I M A N 15

20 VI. FINDINGS Prudent use of cost sharing can play an effective role in making patients more sensitive to the rising cost of medical care without discouraging appropriate and preemptive care. However, employers should consider reasonable alternatives to high-oop programs for populations at highest risk. It is critical that future benefit analysts understand the unique behavior of severely ill patients, recognize their greater variability in individual patient-risk characteristics, and strive to avoid the danger of applying a rigid administrative solution in the absence of clinical evidence. Recent changes have seen a greater portion of specialty drugs moving from the medical to the pharmacy benefit. An important unanticipated outcome of this administrative move is potentially an uncontrolled OOP burden for specialty drugs that were previously delivered under the medical benefit. Thus, future benefit designs must take account of the potential impact on patient access and care. Under current circumstances, the most costly patients will quickly hit their maximum OOP limit early in a treatment year; this means that controlling unnecessary costs will depend critically on improvements in the efficiency of delivery and treatment. High-utilization patients have relatively low rates of disenrollment, represent a small percentage of the population, and are easy to identify; for these reasons, improving long-term cost containment among them can be best served through a proactive engagement of programs like targeted disease management, individual case management, and early disease identification and intervention. Ultimately, savings from the more efficient coordination of drug-related cost sharing, when balanced against total clinical risk and overall efficiency of health services utilization, would be passed back to the employer in the form of stabilized or reduced payer premium rates. When considering future corporate benefit policy, employers should think very carefully before creating an employee benefit environment that discourages patients from seeking expensive but appropriate care or therapeutic options that have proven to be effective in helping control overall medical costs. Current research focusing on the underutilization of treatment has shown that even modest increases in cost sharing can significantly reduce appropriate utilization and treatment compliance. The result is increased total spending as members use more costly services, including hospitalization and emergency services. It is easy to lose sight of the human impact of benefit design decision-making, since it may have a significant impact on only one out of five individuals. However, as healthcare benefits become more complex, it is important to realize that targeting services, diseases, or products that affect only a small percentage of the population may have a significant, and potentially adverse, effect on costs. People with high-cost conditions are more likely to be sensitive to healthcare benefit changes, and although cost management is a good and effective strategy, cost shifting may not be. 16 M I L L I M A N November 2007

21 ACKNOWLEDGEMENTS This research was supported by Amgen, Inc. The authors are solely responsible for the manuscript s content. By prior contractual agreement, neither Amgen nor any other agency has authority over the design and conduct of this study; the collection, analysis, preparation, and interpretation of the data; and final preparation of the manuscript. Milliman is an independent firm of actuaries and healthcare consultants. The final editing and content of this study are based on Milliman proprietary data and national healthcare consulting experience. November 2007 M I L L I M A N 17

22 REFERENCES (1) Hsu, J., et al. Unintended Consequences of Caps on Medicare Drug Benefits. New England Journal of Medicine. 354; 22 June 1, 2006: (2) Goldman, D, et al. Varying Pharmacy Benefits with Clinical Status: The Case of Cholesterol-Lowering Therapy. American Journal of Managed Care. Vol. 12, No. 1, January 2006: (3) Goldman, D, et al. Benefit design and Specialty Drug Use. Health Affairs. Vol. 25, No. 5, Sept/Oct 2006: (4) Banthin, J, et al. Changes in Financial Burden for Healthcare. Journal of the American Medical Association. Vol. 296, No. 22, Dec 14, 2006: (5) Chernew, M, Rosen, A, and Fendrick, A. Value-Based Insurance Design. Health Affairs. Jan 30, 2007: w195-w M I L L I M A N November 2007

23 M I L L I M A N November 2007

24 November 2007 M I L L I M A N