The ADA’s ‘Standards of Medical Care in Diabetes’ recommends lowering HbA1c to <7.0% (<53 mmol/mol) in most patients to reduce the incidence of microvascular disease . This can be achieved with a mean plasma glucose of ~8.3–8.9 mmol/l (~150–160 mg/dl); ideally, fasting and pre-meal glucose should be maintained at <7.2 mmol/l (<130 mg/dl) and the postprandial glucose at <10 mmol/l (<180 mg/dl). More stringent HbA1c targets (e.g. 6.0–6.5% [42–48 mmol/mol]) might be considered in selected patients (with short disease duration, long life expectancy, no significant CVD) if this can be achieved without significant hypoglycaemia or other adverse effects of treatment [20, 43]. Conversely, less stringent HbA1c goals—e.g. 7.5–8.0% (58–64 mmol/mol) or even slightly higher—are appropriate for patients with a history of severe hypoglycaemia, limited life expectancy, advanced complications, extensive comorbid conditions and those in whom the target is difficult to attain despite intensive self-management education, repeated counselling and effective doses of multiple glucose-lowering agents, including insulin [20, 44].
The accumulated results from the aforementioned type 2 diabetes cardiovascular trials suggest that not everyone benefits from aggressive glucose management. It follows that it is important to individualise treatment targets [5, 34–36]. The elements that may guide the clinician in choosing an HbA1c target for a specific patient are shown in Fig. 1. As mentioned earlier, the desires and values of the patient should also be considered, since the achievement of any degree of glucose control requires active participation and commitment [19, 23, 45, 46]. Indeed, any target could reflect an agreement between patient and clinician. An important related concept is that the ease with which more intensive targets are reached influences treatment decisions; logically, lower targets are attractive if they can be achieved with less complex regimens and no or minimal adverse effects. Importantly, utilising the percentage of diabetic patients who are achieving an HbA1c <7.0% (<53 mmol/mol) as a quality indicator, as promulgated by various healthcare organisations, is inconsistent with the emphasis on individualisation of treatment goals.
Interventions designed to impact an individual’s physical activity levels and food intake are critical parts of type 2 diabetes management [47, 48]. All patients should receive standardised general diabetes education (individual or group, preferably using an approved curriculum), with a specific focus on dietary interventions and the importance of increasing physical activity. While encouraging therapeutic lifestyle change is important at diagnosis, periodic counselling should also be integrated into the treatment programme.
Weight reduction, achieved through dietary means alone or with adjunctive medical or surgical intervention, improves glycaemic control and other cardiovascular risk factors. Modest weight loss (5–10%) contributes meaningfully to achieving improved glucose control. Accordingly, establishing a goal of weight reduction, or at least weight maintenance, is recommended.
Dietary advice must be personalised . Patients should be encouraged to eat healthy foods that are consistent with the prevailing population-wide dietary recommendations and with an individual’s preferences and culture. Foods high in fibre (such as vegetables, fruits, wholegrains and legumes), low-fat dairy products and fresh fish should be emphasised. High-energy foods, including those rich in saturated fats, and sweet desserts and snacks should be eaten less frequently and in lower amounts [50–52]. Patients who eventually lose and keep weight off usually do so after numerous cycles of weight loss and relapse. The healthcare team should remain non-judgmental but persistent, re-visiting and encouraging therapeutic lifestyle changes frequently, if needed.
As much physical activity as possible should be promoted, ideally aiming for at least 150 min/week of moderate activity including aerobic, resistance and flexibility training . In older individuals, or those with mobility challenges, so long as tolerated from a cardiovascular standpoint, any increase in activity level is advantageous.
At diagnosis, highly motivated patients with HbA1c already near target (e.g. <7.5% [<58 mmol/mol]) could be given the opportunity to engage in lifestyle change for a period of 3–6 months before embarking on pharmacotherapy (usually metformin). Those with moderate hyperglycaemia or in whom lifestyle changes are anticipated to be unsuccessful should be promptly started on an anti-hyperglycaemic agent (also usually metformin) at diagnosis, which can later be modified or possibly discontinued if lifestyle changes are successful.
Oral agents and non-insulin injectables
Important properties of anti-hyperglycaemic agents that play a role in the choice of drug(s) in individual patients are summarised in the text box ‘Properties of currently available glucose-lowering agents that may guide treatment choice in individual patients with type 2 diabetes mellitus’. Ultimately, the aims of controlling glycaemia are to avoid acute osmotic symptoms of hyperglycaemia, to avoid instability in blood glucose over time, and to prevent/delay the development of diabetic complications without adversely affecting quality of life. Information on whether specific agents have this ability is incomplete; an answer to these questions requires long-term, large-scale clinical trials—not available for most drugs. Effects on surrogate measures for glycaemic control (e.g. HbA1c) generally reflect changes in the probability of developing microvascular disease but not necessarily macrovascular complications. Particularly from a patient standpoint, stability of metabolic control over time may be another specific goal.
Metformin, a biguanide, remains the most widely used first-line type 2 diabetes drug; its mechanism of action predominately involves reducing hepatic glucose production [54, 55]. It is generally considered weight-neutral with chronic use and does not increase the risk of hypoglycaemia. Metformin is associated with initial gastrointestinal side effects, and caution is advised to avoid its use in patients at risk for lactic acidosis (e.g. in advanced renal insufficiency, alcoholism), a rare complication of therapy. As noted earlier, there may be some cardiovascular benefits from this drug, but the clinical trial data are not robust.
The oldest oral agent class is the sulfonylurea insulin secretagogues. Through the closure of ATP-sensitive potassium channels on beta cells, these drugs stimulate insulin release . While effective in controlling glucose levels, their use is associated with modest weight gain and risk of hypoglycaemia. In addition, studies have demonstrated a secondary failure rate that may exceed other drugs, ascribed to an exacerbation of islet dysfunction . Shorter-acting secretagogues, the meglitinides (or glinides), stimulate insulin release through similar mechanisms but may be associated with less hypoglycaemia . They require more frequent dosing, however.
Thiazolidinediones (TZDs) are peroxisome proliferator-activated receptor γ activators  that improve insulin sensitivity in skeletal muscle and reduce hepatic glucose production [54, 55]. They do not increase the risk of hypoglycaemia and may be more durable in their effectiveness than sulfonylureas and metformin . Pioglitazone appeared to have a modest benefit on cardiovascular events as a secondary outcome in one large trial involving patients with overt macrovascular disease . Another agent of this class, rosiglitazone, is no longer widely available owing to concerns of increased myocardial infarction risk . Pioglitazone has recently been associated with a possible increased risk of bladder cancer . Recognised side effects of TZDs include weight gain, fluid retention leading to oedema and/or heart failure in predisposed individuals and increased risk of bone fractures [57, 60].
Drugs focused on the incretin system have been introduced more recently . The injectable GLP-1 receptor agonists mimic the effects of endogenous GLP-1, thereby stimulating pancreatic insulin secretion in a glucose-dependent fashion, suppressing pancreatic glucagon output, slowing gastric emptying and decreasing appetite. Their main advantage is weight loss, which is modest in most patients but can be significant in some. A limiting side effect is nausea and vomiting, particularly early in the course of treatment. Concerns regarding an increased risk of pancreatitis remain unresolved. The oral dipeptidyl peptidase IV (DPP-4) inhibitors enhance circulating concentrations of active GLP-1 and GIP . Their major effect appears to be in the regulation of insulin and glucagon secretion; they are weight neutral. Typically, neither of the incretin-based classes cause hypoglycaemia by themselves.
Two agents that are used infrequently in the USA and Europe are the α-glucosidase inhibitors (AGIs), which retard gut carbohydrate absorption , and colesevelam, a bile acid sequestrant whose mechanism of glucose-lowering action remains poorly understood and whose major additional benefit is LDL-cholesterol reduction . Both have gastrointestinal effects, mainly flatulence with AGIs and constipation with colesevelam. The dopamine agonist bromocriptine is only available in the USA as an anti-hyperglycaemic agent . Its mechanism of action and precise role are unclear. The amylin agonist, pramlintide, is typically reserved for patients treated with intensive insulin therapy, usually in type 1 diabetes mellitus; it decreases postprandial glucose excursions by inhibiting glucagon secretion and slowing gastric emptying .
The glucose-lowering effectiveness of non-insulin pharmacological agents is said to be high for metformin, sulfonylureas, TZDs and GLP-1 agonists (expected HbA1c reduction ~1.0–1.5%) [1, 69, 70], and generally lower for meglitinides, DPP-4 inhibitors, AGIs, colesevelam and bromocriptine (~0.5–1.0%). However, older drugs have typically been tested in clinical trial participants with higher baseline HbA1c, which is itself associated with greater treatment emergent glycaemic reductions, irrespective of therapy type. In head-to-head studies, any differential effects on glucose control are small. So agent- and patient-specific properties, such as dosing frequency, side-effect profiles, cost and other benefits often guide their selection.
Due to the progressive beta cell dysfunction that characterises type 2 diabetes, insulin replacement therapy is frequently required . Importantly, most patients maintain some endogenous insulin secretion even in late stages of disease. Accordingly, the more complex and intensive strategies of type 1 diabetes are not typically necessary .
Ideally, the principle of insulin use is the creation of as normal a glycaemic profile as possible without unacceptable weight gain or hypoglycaemia . As initial therapy, unless the patient is markedly hyperglycaemic and/or symptomatic, a ‘basal’ insulin alone is typically added . Basal insulin provides relatively uniform insulin coverage throughout the day and night, mainly to control blood glucose by suppressing hepatic glucose production in between meals and during sleep. Either intermediate-acting (neutral protamine Hagedorn [NPH]) or long-acting (insulin glargine [A21Gly,B31Arg,B32Arg human insulin] or insulin detemir [B29Lys(ε-tetradecanoyl),desB30 human insulin]) formulations may be used. The latter two are associated with modestly less overnight hypoglycaemia (insulin glargine, insulin detemir) than NPH and possibly slightly less weight gain (insulin detemir), but are more expensive [75, 76]. Of note, the dosing of these basal insulin analogues may differ, with most comparative trials showing a higher average unit requirement with insulin detemir .
Although the majority of patients with type 2 diabetes requiring insulin therapy can be successfully treated with basal insulin alone, some, because of progressive diminution in their insulin secretory capacity, will require prandial insulin therapy with shorter-acting insulins. This is typically provided in the form of the rapid insulin analogues, insulin lispro (B28Lys,B29Pro human insulin), insulin aspart (B28Asp human insulin) or insulin glulisine (B3Lys,B29Glu human insulin), which may be dosed just before the meal. They result in better postprandial glucose control than the less costly human regular insulin, whose pharmacokinetic profile makes it less attractive in this setting.
Ideally, an insulin treatment programme should be designed specifically for an individual patient, to match the supply of insulin to his or her dietary/exercise habits and prevailing glucose trends, as revealed through self-monitoring. Anticipated glucose-lowering effects should be balanced with the convenience of the regimen, in the context of an individual’s specific therapy goals (Fig. 1).
Proper patient education regarding glucose monitoring, insulin injection technique, insulin storage, recognition/treatment of hypoglycaemia, and ‘sick day’ rules is imperative. Where available, certified diabetes educators can be invaluable in guiding the patient through this process.
Initial drug therapy
It is generally agreed that metformin, if not contraindicated and if tolerated, is the preferred and most cost-effective first agent  (Fig. 2 and electronic supplementary material [ESM] Figs). It is initiated at, or soon after, diagnosis, especially in patients in whom lifestyle intervention alone has not achieved, or is unlikely to achieve, HbA1c goals. Because of frequent gastrointestinal side effects, it should be started at a low dose with gradual titration. Patients with a high baseline HbA1c (e.g. ≥9.0% [≥75 mmol/mol]) have a low probability of achieving a near-normal target with monotherapy. It may therefore be justified to start directly with a combination of two non-insulin agents or with insulin itself in this circumstance . If a patient presents with significant hyperglycaemic symptoms and/or has dramatically elevated plasma glucose concentrations (e.g. >16.7–19.4 mmol/l [>300–350 mg/dl]) or HbA1c (e.g. ≥10.0–12.0% [86–108 mmol/mol]), insulin therapy should be strongly considered from the outset. Such treatment is mandatory when catabolic features are exhibited or, of course, if ketonuria is demonstrated, the latter reflecting profound insulin deficiency. Importantly, unless there is evidence of type 1 diabetes, once symptoms are relieved, glucotoxicity resolved, and the metabolic state stabilised, it may be possible to taper insulin partially or entirely, transferring to non-insulin anti-hyperglycaemic agents, perhaps in combination.
If metformin cannot be used, another oral agent could be chosen, such as a sulfonylurea/glinide, pioglitazone, or a DPP-4 inhibitor; in occasional cases where weight loss is seen as an essential aspect of therapy, initial treatment with a GLP-1 receptor agonist might be useful. Where available, less commonly used drugs (AGIs, colesevelam, bromocriptine) might also be considered in selected patients, but their modest glycaemic effects and side-effect profiles make them less attractive candidates. Specific patient preferences, characteristics, susceptibilities to side effects, potential for weight gain and hypoglycaemia should play a major role in drug selection [20, 21]. (See ESM Figs for adaptations of Fig. 2 that address specific patient scenarios.)
Advancing to dual combination therapy
Figure 2 (and ESM Figs) also depicts potential sequences of escalating glucose-lowering therapy beyond metformin. If monotherapy alone does not achieve/maintain an HbA1c target over ~3 months, the next step would be to add a second oral agent, a GLP-1 receptor agonist or basal insulin [5, 10]. Notably, the higher the HbA1c, the more likely insulin will be required. On average, any second agent is typically associated with an approximate further reduction in HbA1c of ~1% (11 mmol/mol) [70, 79]. If no clinically meaningful glycaemic reduction (i.e. ‘non-responder’) is demonstrated, then, adherence having been investigated, that agent should be discontinued, and another with a different mechanism of action substituted. With a distinct paucity of long-term comparative-effectiveness trials available, uniform recommendations on the best agent to be combined with metformin cannot be made . Thus, advantages and disadvantages of specific drugs for each patient should be considered (Text box ‘Properties of currently available glucose-lowering agents that may guide treatment choice in individual patients with type 2 diabetes mellitus’).
Some anti-hyperglycaemic medications lead to weight gain. This may be associated with worsening markers of insulin resistance and cardiovascular risk. One exception may be TZDs ; weight gain associated with this class occurs in association with decreased insulin resistance. Although there is no uniform evidence that increases in weight in the range observed with certain therapies translate into a substantially increased cardiovascular risk, it remains important to avoid unnecessary weight gain by optimal medication selection and dose titration.
For all medications, consideration should also be given to overall tolerability. Even occasional hypoglycaemia may be devastating, if severe, or merely irritating, if mild . Gastrointestinal side effects may be tolerated by some, but not others. Fluid retention may pose a clinical or merely an aesthetic problem . The risk of bone fractures may be a specific concern in postmenopausal women .
It must be acknowledged that costs are a critical issue driving the selection of glucose-lowering agents in many environments. For resource-limited settings, less expensive agents should be chosen. However, due consideration should be also given to side effects and any necessary monitoring, with their own cost implications. Moreover, prevention of morbid long-term complications will likely reduce long-term expenses attributed to the disease.
Advancing to triple combination therapy
Some studies have shown advantages of adding a third non-insulin agent to a two-drug combination that is not yet or no longer achieving the glycaemic target [83–86]. Not surprisingly, however, at this juncture, the most robust response will usually be with insulin. Indeed, since diabetes is associated with progressive beta cell loss, many patients, especially those with long-standing disease, will eventually need to be transitioned to insulin, which should be favoured in circumstances where the degree of hyperglycaemia (e.g. ≥8.5%) makes it unlikely that another drug will be of sufficient benefit . If triple combination therapy exclusive of insulin is tried, the patient should be monitored closely, with the approach promptly reconsidered if it proves to be unsuccessful. Many months of uncontrolled hyperglycaemia should specifically be avoided.
In using triple combinations the essential consideration is obviously to use agents with complementary mechanisms of action (see Fig. 2 and ESM Figs). Increasing the number of drugs heightens the potential for side effects and drug–drug interactions, raises costs and negatively impacts patient adherence. The rationale, benefits and side effects of each new medication should be discussed with the patient. The clinical characteristics of patients more or less likely to respond to specific combinations are, unfortunately, not well defined.
Transitions to and titrations of insulin
Most patients express reluctance to beginning injectable therapy, but, if the practitioner feels that such a transition is important, encouragement and education can usually overcome such reticence. Insulin is typically begun at a low dose (e.g. 0.1–0.2 U kg–1 day–1), although larger amounts (0.3–0.4 U kg–1 day–1) are reasonable in the more severely hyperglycaemic. The most convenient strategy is with a single injection of a basal insulin, with the timing of administration dependent on the patient’s schedule and overall glucose profile (Fig. 3).
Although extensive dosing instructions for insulin are beyond the scope of this statement, most patients can be taught to uptitrate their own insulin dose based on several algorithms, each essentially involving the addition of a small dose increase if hyperglycaemia persists [74, 76, 88]. For example, the addition of 1–2 units (or, in those already on higher doses, increments of 5–10%) to the daily dose once or twice weekly if the fasting glucose levels are above the pre-agreed target is a reasonable approach . As the target is neared, dosage adjustments should be more modest and occur less frequently. Downward adjustment is advisable if any hypoglycaemia occurs. During self-titration, frequent contact (telephone, e-mail) with the clinician may be necessary. Practitioners themselves can, of course, also titrate basal insulin, but this would involve more intensive contact with the patient than typically available in routine clinical practice. Daily self-monitoring of blood glucose is of obvious importance during this phase. After the insulin dose is stabilised, the frequency of monitoring should be reviewed .
Consideration should be given to the addition of prandial or mealtime insulin coverage when significant postprandial glucose excursions (e.g. to >10.0 mmol/l [>180 mg/dl]) occur. This is suggested when the fasting glucose is at target but the HbA1c remains above goal after 3–6 months of basal insulin titration . The same would apply if large drops in glucose occur during overnight hours or in between meals, as the basal insulin dose is increased. In this scenario, the basal insulin dose would obviously need to be simultaneously decreased as prandial insulin is initiated. Although basal insulin is titrated primarily against the fasting glucose, generally irrespective of the total dose, practitioners should be aware that the need for prandial insulin therapy will become likely the more the daily dose exceeds 0.5 U kg–1 day–1, especially as it approaches 1 U kg–1 day–1. The aim with mealtime insulin is to blunt postprandial glycaemic excursions, which can be extreme in some individuals, resulting in poor control during the day. Such coverage may be provided by one of two methods.
The most precise and flexible prandial coverage is possible with ‘basal-bolus’ therapy, involving the addition of pre-meal rapid-acting insulin analogue to ongoing basal insulin. One graduated approach is to add prandial insulin before the meal responsible for the largest glucose excursion—typically that with the greatest carbohydrate content, often, but not always, the evening meal . Subsequently, a second injection can be administered before the meal with the next largest excursion (often breakfast). Ultimately, a third injection may be added before the smallest meal (often lunch) . The actual glycaemic benefits of these more advanced regimens after basal insulin are generally modest in typical patients . So, again, individualisation of therapy is key, incorporating the degree of hyperglycaemia needing to be addressed and the overall capacities of the patient. Importantly, data trends from self-monitoring may be particularly helpful in titrating insulins and their doses within these more advanced regimens to optimise control.
A second, perhaps more convenient but less adaptable method involves ‘pre-mixed’ insulin, consisting of a fixed combination of an intermediate insulin with regular insulin or a rapid analogue. Traditionally, this is administered twice daily, before morning and evening meals. In general, when compared with basal insulin alone, pre-mixed regimens tend to lower HbA1c to a larger degree, but often at the expense of slightly more hypoglycaemia and weight gain . Disadvantages include the inability to titrate the shorter- from the longer-acting component of these formulations. Therefore, this strategy is somewhat inflexible but may be appropriate for certain patients who eat regularly and may be in need of a simplified approach beyond basal insulin [92, 93]. (An older and less commonly used variation of this two-injection strategy is known as ‘split-mixed’, involving a fixed amount of intermediate insulin mixed by the patient with a variable amount of regular insulin or a rapid analogue. This allows for greater flexibility in dosing.)
The key messages from dozens of comparative insulin trials in type 2 diabetes include the following:
Any insulin will lower glucose and HbA1c.
All insulins are associated with some weight gain and some risk of hypoglycaemia.
The larger the doses and the more aggressive the titration, the lower the HbA1c, but often with a greater likelihood of adverse effects.
Generally, long-acting insulin analogues reduce the incidence of overnight hypoglycaemia, and rapid-acting insulin analogues reduce postprandial glucose excursions as compared with corresponding human insulins (NPH, Regular), but they generally do not result in clinically significantly lower HbA1c.
Metformin is often continued when basal insulin is added, with studies demonstrating less weight gain when the two are used together . Insulin secretagogues do not seem to provide for additional HbA1c reduction or prevention of hypoglycaemia or weight gain after insulin is started, especially after the dose is titrated and stabilised. When basal insulin is used, continuing the secretagogue may minimise initial deterioration of glycaemic control. However, secretagogues should be avoided once prandial insulin regimens are employed. TZDs should be reduced in dose (or stopped) to avoid oedema and excessive weight gain, although in certain individuals with large insulin requirements from severe insulin resistance, these insulin sensitizers may be very helpful in lowering HbA1c and minimising the required insulin dose . Data concerning the glycaemic benefits of incretin-based therapy combined with basal insulin are accumulating; combination with GLP-1 receptor agonists may be helpful in some patients [97, 98]. Once again, the costs of these more elaborate combined regimens must be carefully considered.